Report No. 29468-ET Ethiopia Well-Being and Poverty in Ethiopia The Role of Agriculture and Agency July 18, 2005 Poverty Reduction and Economic Management 2 (AFTP2) Country Department for Ethiopia Africa Region Document of the World Bank M&E Monitoring and Evaluation MoE Ministry of Education MoFED Ministry of Financeand Economic Development MoH Ministry of Health MP Member of Parliament MTNSP MediumTermNational Statistical Program NGOs Non-Governmental Organizations ODA Official Development Assistance Po, PI, p2 Poverty headcount, poverty gap, poverty gap squared PADETES Participatory Demonstrationand Training Extension System PAN Poverty Action Network PER Public Expenditure Review PPA Participatory Poverty Assessment PPD MoH Programand Planning Department,Ministry of Health PPP PurchasingPower Parity PRSC Poverty Reduction SupportCredit PSCAP Public Service Delivery CapacityBuilding Program PTR PupiliTeacher Ratio SDPRP Sustainable Development and Poverty ReductionProgram SDPRS SustainableDevelopment andPoverty Reduction Strategy SG 2000 Sasakawa-Global2000 SNNPR SouthernNations Nationalities andPeopleRegional State SSA Sub-SaharanAfrica TFR Total Fertility Rate UNOCHA UNOffice for the Coordination ofHumanAffairs VCT Voluntary Counseling and Testing WB World Bank WDI World Development Indicators WMS Welfare Monitoring Survey WMU Welfare MonitoringUnit Country Director: Ishac Diwan Sector Director: Paula Donovan Sector Manager: Kathie L.Krumm Task Team Leader: Luc Christiaensen TABLE OF CONTENTS EXECUTIVE SUMMARY............................................................................................................................ 1 Introduction ......................................................................................................................................... 1 Part I: How Well hasthe EthiopianPopulationFared?......................................................................... 5 Chapter 1: Monetary Dimensions of Well Being-A Utilitarian Perspective .............................................. 5 1.1 The Evolution of Povertyand Inequality inEthiopia since 1992.................................... 5 1.1.1 8 Micro evidence ................................................................................................. Macro evidence................................................................................................... 1.1.2 13 21 1.2 The Face of Hunger and Deprivation............................................................................ 1.1.3 Emergingpicture ofpoverty and inequality since 1992................................... 24 1.2.1 24 24 The geography ofpoverty................................................................................. Livelihoods and poverty ................................................................................... Education and poverty ...................................................................................... 1.2.2 1.2.3 27 1.2.4 Pastoralists ........................................................................................................ 27 1.3 Risk,Vulnerability and Poverty..................................................................................... 28 Chapter 2: Non-Monetary Dimensions of Well-Being-A Capability Perspective................................... -33 2.1 The Evolution of Human Capabilities ........................................................................... 33 2.2 The Status of Empowerment inEthiopia....................................................................... 37 2.2.1 The disadvantagedpositionof women ............................................................. 2.2.2 The role of informal institutions amongpastoralists and other social groups ..40 2.2.3 State-society relations-the role of information and traditional institutions......46 49 Chapter 3: People's Well-Being-Concluding Remarks............................................................................ 59 Part 11: Determinantsof MonetaryWell-Beingand Poverty .............................................................. -63 Chapter 4: Endowments. Returns and Risks............................................................................................. 64 4.1 64 Returnsto EndowmentsAcross Time and Space.......................................................... People's Endowmentsand their Distribution inEthiopia.............................................. 4.2 79 4.2.1 Returns across time........................................................................................... 79 4.2.2 Returns across space......................................................................................... 86 4.3 88 4.4 Riskand Poverty............................................................................................................ 91 Geography and Poverty ................................................................................................. Chapter 5: Performance and Potentialof the Agricultural Sector............................................................. 97 5.1 Stagnating Agricultural Performance-A Macro Perspective......................................... 99 5.1.1 Stylized facts..................................................................................................... 99 5.1.2 102 5.2 Enhancing Staple Crop Productivity-A Micro Perspective......................................... Spatial differentiation inproductivity response.............................................. 109 Chapter 6: Growing out of Poverty: The Role of Agriculture and Aid................................................... 121 6.1 A Macro Perspective.................................................................................................... 6.1.1 The role of agriculture in achieving growth with poverty reduction..............122 122 6.1.2 Aid and aid absorption.................................................................................... 126 6.2 Toward a Feasible Pro-Poor Agricultural Growth Strategy ........................................ 131 6.2.1 132 Patterns of agricultural growth and poverty reduction ................................... The theoretical case for agriculture leddevelopment ..................................... 6.2.2 135 6.3 Toward an Optimal Public InvestmentMix-A Micro Perspective ............................. 153 Chapter 7: EnhancingPeople's Income-Concluding remarks................................................................ 159 Part111:DeterminantsofNon-MonetaryWell-Being ......................................................................... 167 Chapter 8: Child Malnutrition. Food Security and Economic Growth ................................................... 168 8.1 A Profile of ChildMalnutritioninEthiopia................................................................. The Nutrition-PovertyTrapAnother Lost Generation inthe Making........................ 168 8.2 173 8.3 Determinants of ChildMalnutrition ............................................................................ 176 8.4 Policy Actions Neededto Halve the Prevalence of Pre-School Stunting.................... 178 8.5 ConcludingRemarks ................................................................................................... 180 Chapter 9: Living and Growing inthe Face of Disease .......................................................................... 183 9.1 Health and Poverty ...................................................................................................... 183 9.2 Health Profile of the Ethiopian People ........................................................................ 184 9.2.1 HIV/AIDS....................................................................................................... 186 9.2.2 Malaria............................................................................................................ 187 9.2.3 Maternal mortality .......................................................................................... 187 187 9.3 Determinants of Child Mortality.................................................................................. 9.2.4 Child mortality................................................................................................ 188 9.4 Policy Actions Needed to Reach the Child Mortality MDG....................................... 193 9.5 Concluding Remarks ................................................................................................... 195 Chapter 10: Toward an EducatedPeople.................................................................................................. 197 10.1 Literacy, Development and Ongoing Policy Challenges inEthiopia.......................... 197 199 10.3 Determinants of Primary Schooling inEthiopia.......................................................... 10.2 EducationalProfile inEthiopia.................................................................................... 203 10.3.1 Household and community characteristics ..................................................... 204 10.3.2 Risk and shocks .............................................................................................. 10.3.3 School characteristics ..................................................................................... 206 206 10.3.4 Determinants of fifth-grade completion inEthiopia....................................... 208 10.4 PolicyActionsNeededto Attain the EducationMDG................................................ 210 10.5 ConcludingRemarks ................................................................................................... 211 Chapter 11:Strengthening People's Agency-Concluding Remarks......................................................... 213 APPENDICES ...................................................................................................................................... -219 Appendix 2: Supplementary Figures........................................................................................................ 225 Appendix 1: Ethiopia's ActionPlanto StrengthenSDPRPMonitoringandEvaluation......................... 219 Appendix 3: Supplementary Tables......................................................................................................... 252 Appendix 5: Price Fluctuations, Substitution, andMarketActivity Per Cereal....................................... Appendix4: The Theoretical, Empirical,andHistoricalCase for AgricultureLedDevelopment..........228 257 REFERENCES..................................................................................................................................... 261 LIST OF PICTURES. BOXES. FIGURES AND TABLES Picture Picture 11.1: Billboards against gender based violence by the Gurage Women's and Teachers' Association......................................................................................................................... 217 Boxes Box 1.1: How poor i s Ethiopia really? Different measures. different results? ............................................. 7 9 Box 1.3: Methodologies used to measure the extent o f poverty ................................................................. Box 1.2: Basic assumptions underpinning the growth-poverty scenarios..................................................... 14 Box 1.4: Methodological considerations inobtainingpoverty lines........................................................... 16 Box 2.2: Intermediate indicators o f empowerment: agency and opportunity structure inpractice............38 Box 2.1: Understanding empowerment....................................................................................................... 39 Box 2.3: Ethiopianrape victim pits law against culture ............................................................................. 40 Box 2.4: Asset distribution at divorce......................................................................................................... 45 Box 2.5: Violence and abduction inmarriage............................................................................................. 45 Box 2.6: Gurage traditional law................................................................................................................. 48 Box 2.7: The effect o f traditional Sidama institutions ............................................................................... 48 Box 2.8: Recent reform infiscal resource transfers ................................................................................... 53 Box 2.9: Citizen's relationships with kebele and woreda governments..................................................... 54 Box 5.1: Agricultural policies in Ethiopia since 1992 ................................................................................ 97 Box 5.2: Alternative definitions o f highand low potential areas.............................................................. 103 Box 5.3: Trends incereal yields inlow, medium, and highpotential zones............................................. 104 Box 6.1: The relationshipbetween growth, aid, and institutions .............................................................. 134 Box 6.3: Rural households' cereal market positions................................................................................. Box 6.2: Reflections on four common critiques to agriculture led development in Sub-Saharan Africa .129 139 Box 6.4: Assumptions underpinning the multi-market agricultural growth-poverty model ..................... 147 Figures Figure i-a: Evolution o fpoverty incidence between 1989 and 2004. ........................................ ... iii Figurei-b: Enrollments ingrades 1.12. Ethiopia. 1967-2002 .................................................. v Figurei-c: Incidence o f radio and TV ownership among rural households in selected Sub-Saharan African countries. 1995-2001....................................................................... viii ... Figure0.1: Real gross domestic product per capita. Ethiopia. 1961 to 2003 ........................................... 1 Figure 1.1: Evolution o fpoverty incidence between 1989 and 2004 ..................................................... 10 Figure 1.2: Evolution o f overall. sectoral and government GDP growth rates. 1992-2004 ................... 11 Figure 1.3: 12 Producer price o f Arabica coffee (US ct. per lb).................................................................. Consumption growth ratios between 1995 and 1999 per consumption decile..................... Evolution o f sectoral shares intotal government expenditures. 1992-2004 ........................ Figure 1.4: 18 Figure 1.5: 26 Figure 1.6: Long run average rainfall and rainfall variation across woredas in Ethiopia....................... 29 Figure 1.7: Nominal and deflated cereal prices - Addis Ababa wholesale ............................................ 30 Figure 1.8: Maize price structure -Addis Ababa and Nekempt............................................................. 50 Figure 1.9: Share o f population at risk o f malaria in Ethiopia................................................................ 31 Figure2.1: Enrollments in grades 1-12. Ethiopia. 1967-2002................................................................ 33 Figure2.2: Evolution o f child stuntingin Ethiopia 1983-00 .................................................................. 35 Figure 2.3: Evolution o f child wasting in Ethiopia 1983-00 .................................................................. 36 Figure 2.4: Xncidence o f radio and TV ownership among ruralhouseholds in selected Sub-Saharan African countries. 1995-2001............................................................................................... 49 Figure 2.5: Governance indicators in Ethiopia....................................................................................... 51 Figure 2.6: Governance inEthiopia compared with similar categories (2002) ...................................... 51 Figure 4.1: 84 Effect o f D A P equivalent nutrient loss due to dung collection on expenditures.................-91 Correlation between altitude and expenditures .................................................................... Effect o f fertilizer use on per adult equivalent expenditures................................................ Figure 4.2: 90 Figure4.3: Figure 4.4: Effect o f rain shocks on expenditures .................................................................................. 92 Figure5.1: Agricultural performance inEthiopia during 1990-2004................................................... 100 Figure 5.2: 101 Trends inagriculturalyields, 1980-2000 (quintalslhectare) ............................................... Trends inthe value o f agricultural output and food production 1990-2004....................... Figure5.3: 101 Figure 5.4: Commercial fertilizer use per ha per crop over time .......................................................... 102 Figure 5.5: Food security potential inEthiopia .................................................................................... 106 Figure 5.6: Trends incommercial fertilizer use by food security index between 1994 and 2001........106 Trends incereal yields by food security index between 1994 and 2001............................ Figure5.7: 107 Figure 6.1: Economic Growth Scenarios and Poverty Reduction ........................................................ 127 Figure6.2: Net official development assistance per GDP for Ethiopia and Sub-Saharan Africa 1981- 2002..................................................................................................... .......................... 128 Figure 8.1: Percent o f children wasted and stunted in Ethiopia, 1983-2000 ........................................ 173 Figure 8.2: Child stunting and wasting in urban vs.rural areas, 1996-2000 ........................................ 174 Figure 8.3: Child wasting and stunting inEthiopia by region, 2000.................................................... 175 Figure 8.4: Prevalence o f child stunting inEthiopia............................................................................. 179 Figure 9.1: Under-five child mortality by wealth quintile .................................................................... 188 Figure9.2: Regional variation in infant mortality and under five mortality rate.................................. 189 Figure 9.3: Achieving MDG goal for reduction in child mortality....................................................... 193 Figure 9.4: Effect o f selected interventions on rural under-five childhood mortality .......................... 194 Figure 10.1: Trends inpupil-teacher and pupil-sectionratios and innon-salary public spending per primary and secondary student, Ethiopia, 1990-2001........................................................ 199 Figure 10.2: Grade-specific completionrates, Ethiopia, 1995-96, 2000, and 200112............................ 200 Figure 10.3: Gross primary enrollmentrates by gender and wealth quintile, 1996................................ 201 Figure 10.4: Gross primary enrollment rates by gender and wealth quintile, 2000................................ 202 Figure 10.5: Primary completionratesby gender and wealth quintile, 1996......................................... 202 Figure A.1.1: Ethiopia- Percent o f Total Territory Exposedto Malaria...................................... 225 Figure A.1.2: Ethiopia-Percent 226 Figure A.2.1: Educational attainment by gender (%)............................................................ o f population vulnerable to malaria........................................ 226 Figure A.2.2: Employment status by gender...................................................................... 227 Figure(A4)l: Fadnon-farmlinkages and leakages inthe rural economy.................................. 253 Tables Table ia: Cereal yield and input use in food deficit. food balanced and food surplus areas. 2001/02 ......xi Table 1.1: Per capita GDP growth inkey economic sectors. 1992-2004.............................................. 10 Table 1.2: Evolution o f poverty between 1995 and 1999...................................................................... 14 Table 1.3: Evolution o f inequality inEthiopia ...................................................................................... Table 1.4: ERHS panel data evidence on the evolution of welfare inrural Ethiopia, 1994-1999......... 19 EUHS panel data evidence on the evolution o f welfare in urban Ethiopia, 1994-2000.......18 Table 1.5: 19 Table 1.6: Ownership o f livestock and consumer durables inEthiopia between 1995 and 1999.........21 Table 1.7: Index o freal government expenditure and share inpercent o f GDP .................................. -23 Table 1.8: Incidence o fpoverty by education in 1999 .......................................................................... 24 Table 1.9: Consumption and poverty incidence by sector o f employment o f household head in 1995 and 1999............................................................................................................................... 25 Table 1.10: Ethiopia, incidence o f poverty by livelihood ....................................................................... 25 Table 1.11: Evolution o f poverty among coffee and chat growers, 1995-1999 ...................................... 26 Table 1.12: Poverty incidence in Ethiopia by administrative region 1995-1999 .................................... 27 Table 1.13: Poverty dynamics in ERHS between 1994 and 1999........................................................... 28 Table 2.1: Selected health indicators for Ethiopia, 1984-199912000 .................................................... 34 Table 2.2: Sources o f drinking water and use o f waste disposal facilities, 1995-1999......................... 37 Table 2.3: Women's opinions on wife beating, Ethiopia 2000.............................................................. 42 Table 2.4: Capacity constraints for decentralization in Ethiopia.,......................................................... 52 Table 2.5: 56 Endowment base and risk factors o f Ethiopian households in 1995 and 1999.................... Strengths and weaknesses of informal organizations........................................................... Table 4.1: 65 Table 4.2: Endowment base and risk factors o f Ethiopian households across rural and urban areas in 67 70 Livestock ownership inEthiopia between 1995 and 1999................................................... Land holdings per person and land inequality...................................................................... 1995 and 1999...................................................................................................................... Table 4.3: Table 4.4: 71 Table 4.5: Parameters o f soil degradation, agro-ecology, population density and malaria incidence in 74 75 Endowment base and risk factors across four regions.......................................................... Biomass use for cooking fuel ............................................................................................... Ethiopia ................................................................................................................................ Table 4.6: Table 4.7: 76 Table 4.8: Parameters o f soil degradation, agro-ecology, populationdensity and malaria incidence across four regions ............................................................................................................... 78 Table 4.9: Estimatedeffects o f household and public endowments on consumption........................... 79 Table 4.10: Variations inreturns to endowments across space............................................................... 87 Table 4.11: Effects o f agro-ecology, population density, soil degradation and shocks on consumption......................................................................................................................... 88 Table 5.1: Characteristics o f food secure and food insecure (rural) zones in 2000............................. 108 Table 5.2: Estimated determinants o f the value o f cereal output in 2000 by food security potential o f the area ............................................................................................................................... 110 Table 5.3: Cereal area and yield by modern input use, 1997198-2001/02 ........................................... 113 Table 5.4: Cereal yield and input use in food deficit, food balanced and food surplus areas, 2001102............................................................................................................................... 113 Table 5.5: Potential for yield gains under improved management....................................................... 114 Table 5.6: Average yield data from SPA practice compared to traditional practice ........................... 114 Table 5.7: The impact of market accessibility on output for different crops ...................................... 116 Table 5.8: Outputsupply elasticities withrespect to own price.......................................................... 117 Table 6.1: Growth, poverty, and inequality under alternative sectoral growth scenarios ................... 123 Table 6.2: 136 Income per capita and income sources across several communities.................................. Cereal market position o f rural households in 1995/96 ..................................................... Table 6.3: Table 6.4: Market participation profile for rural households in selected Afncan locations ................137 Table 6.5: Profile o f market position and degree o f concentration by net sales quintiles ...................140 141 Table 6.6: Net market positionby welfare level in 1995 .................................................................... 141 Table 6.7: Estimated cereal demand and supply elasticities inEthiopia.............................................. Table 6.8: Growth and poverty reducingeffects o f different agricultural growth patterns.................143 148 Table 6.9: Simulated effects on consumption and poverty reduction o f different policy interventions ....................................................................................................................... 154 Table 8.1: Measures o f nutritional status ............................................................................................ 171 Table 9.1: Self reported incidence o f health problems duringthe past two months............................ 185 Table 10.1: Gross primary enrollment in Ethiopia................................................................................ 200 Table 10.2: Net primary enrollment inEthiopia ................................................................................... 200 Table 10.3: Determinants o fprimary school enrollment inEthiopiaby gender and location...............205 Table 10.4: Primary school completion regression-rural and urbanEthiopia ...................................... 209 Table A.1.1: Ethiopia. Price indices in 1999 (at 1995/96 constant prices) .............................................. 228 Table A.1.2: Ethiopia: Poverty lines per reporting area. 1995-1999 ...................................................... Table A.1.3: Ethiopia, Growth in consumption by expenditure decile. 1995-1999................................. 229 230 Table A.1.4: Ethiopia: Reporting adult equivalent total food consumption and share o f food, 1995-1999 ............................................................................................................................................ 231 Table A.2.1: Prevalence o f female circumcision ................................................................ 232 Table A.2.2: Elements o f regional and local government., ..................................................... 233 Table A.2.3: Focus group rankings o f top five most important institutions inpeople's lives, rural sites (Dessie Zuria Woreda, Amhara Region) ......................................................... 234 Table A.2.4: Focus group rankings o f top five most important institutions inpeople's lives, urban sites ..................................................................................................... 235 Table A.4.1: Estimated effects o f radio ownership across location 1995-1999.............................. 236 Table A.4.2: Estimated effects o f radio ownership across location in 1999, with additional controls for wealth effect 1999 ................................................................................... 237 Table A.6.1: Profile o f food aid distribution across market position......................................... 238 Table A.8.1: Estimated child, household and community determinants o f child height for age (pooled sample) ............................................................................................... 239 Table A.8.2: Child malnutrition alleviating potential o f different policy interventions .................... 241 Table A.9.1: Diarrhea incidence and care-seeking ............................................................. 242 Table A.9.2: Burden o f HIV/AIDS inAfrican countries ...................................................... 243 Table A.9.3: Summary statistics o f variables included inthe child mortality regression (rural specification) ......................................................................................... 244 Table (A5)l: Intra-annual cereal price fluctuation in 1995-96................................................ 257 Table (A5)2: Incidence o f cereal market transactions across net cereal buyers/sellers between October 1995 and September 1996 .......................................................................... 258 Table (A5)3: Number o f cereal purchase transactions conditional on the sale o f a cereal .................259 Table (A5)4: Crop level market participation characteristics by season ..................................... 260 ACKNOWLEDGEMENTS Poverty i s as complex as its manifestations are clear. It has many faces and many causes and understanding it requires a multi-disciplinary and context specific approach. To deepen our understanding o f poverty in Ethiopia and sketch pathways out o f poverty the Poverty Assessment Team has drawn on field visits inTigray, Oromiya, Amhara, Harar, and SNNPR, and numerous interactions and dialogues with the Government, development partners, Ethiopian academia and Ethiopian citizens over the course o f 2003-2004. The key themes o f the document were identified during early discussions with the Government and an earlier draft was presented to the Welfare Monitoring Technical Committee, the counterpart in the Government for this study. Earlyversions o f the chapters have also been discussed during the Poverty and Development seminar series organized in collaboration with Addis Ababa University, at the annual 2004 Ethiopian Economic Association conference, the urban poverty workshop organized by the Ethiopian Economic Policy Research Institute, and other seminars and conferences inAddis Ababa. While the Poverty Assessment i s a World Bank ledwork, it i s especially a product o f these many interactions and the Team i s grateful for the numerous occasions it was provided with to deepen its understanding, to address competing arguments, and to sharpen and nuance the resulting insights. The Poverty Assessment Team would like to thank H.E. Ato Mekonnen Manyazewal, State Minister, Ministry o f Finance and Economic Development (MoFED) and Ato Getachew Adem, Head o f the Development Planning and Research Department, MoFED for their support o f the study, the Central Statistical Authority and the Education Management Information Systeminthe Ministryo f Education, for making the survey data available and for their dedicated and graceful assistance in addressing our never-ending follow up queries, and the Biomass project and the Meteorological Institute for making data available on soil fertility and rainfall. The Team would also like to thank the staff o f the Economics Department in Addis Ababa University, the Ethiopian Economic Association, and the Ethiopian Economic Policy Research Institutefor insightfbldiscussions at different stages and occasions. The Core Team o f the Poverty Assessment includes Luc Christiaensen (task team leader), Ruth Alsop, Nazmul Chaudhury, Andrew Dabalen, Stefan Dercon, Berk Ozler, and Martin Ravallion. Most important inputs were given by Vivian Hoffman, Bryan Kurey, Andrew Gache Mude, and Kathleen Withers. Excellent research, editorial and administrative assistance by Meron Assefa, Mesfin Girma Bezawagaw, and Senait Kassa Yifru i s gratefully acknowledged. Invaluable guidance and suggestions throughout the preparation were given by Ishac Diwan, Fred Kilby, Robert Blake and Kathie K " m . Comments and suggestions from the peer reviewers: Jeni Klugman, Giovanna Prennushi, and Kaspar Richter helped improve earlier versions of the document substantially. Many helpful inputs and comments were also received from Harold Alderman, Abebaw Alemayehu, Antoine Bommier, Mulat Demeke, Pierre Dubois, Eyerusalem Fasika, Louise Fox, Madhur Gautam, Caterina Laderchi, Arianna Legovini, David Nielson, Gebreselassie Okubagzhi, Jemal Mohammed Omer, Michelle Phillips, John Riverson, Meera Shekar, Abebe Shimeless, and Jee-Peng Tan. We especially thank Navin Girishankar and Karim El Aynaoui for insightful discussions and moral support throughout the process. EXECUTIVESUMMARY 1. Study objective. A decade and a half of relative peace and political stability, broad economic reforms, and far-reaching political decentralization have brought Ethiopia back from one o f its lowest levels o f income per capita to one o f its highest levels over the past forty years. A t the same time, GDP per capita today i s still only slightly above the levels reached in the early 1 9 7 0 ~underscoring the deep-rooted and complex nature o f poverty in ~ Ethiopia. The positive developments at the macro level obviously beg the questions o f how well the Ethiopian people themselves fared during this period and what can be done to improve their lives further. These are the two overarching questions this study attempts to address, with a larger emphasis on the analysis o f the relative importance o f the different determinants o f people's well-being and its policy implications in light o f the upcoming revision o f Ethiopia's Sustainable Development and Poverty Reduction Strategy (SDPRP). In particular, the study seeks to identify areas o f intervention to improve people's well-being which currently appear promising at the margin and to provide the micro behavioral foundations for developing sector specific policies. For a comprehensive treatment o f the challenges and opportunities related to the different sectors, the study refers to the World Bank Country Status Reports on Health, Education and Rural Development. The particular role o f population growth in poverty reduction and the importance o f effective population interventions are addressed in a separate piece o f Economic Sector Work. 2. Methodologicalapproach. To examine how Ethiopians fared over the past decade and a half, the report takes both a utilitarian and capabilities approach, and explores progress on monetary and non-monetary indicators o f well-being, i.e. measures o f income or consumption, human capabilities and empowerment. It looks inparticular at people's current poverty status, their vulnerability, i.e. their prospect o f being poor in the future, and the evolution o f inequality in Ethiopian society. The findings reported here are grounded in careful economic and micro-econometric analysis o f mostly nationally representative, but also purposively sampled household data, along with secondary data, much o f which has been made available for the first time by the government. In doing so, the report seeks to complement the Poverty Profile o f Ethiopia published by the Ministry o f Finance and Economic Development, Government o f Ethiopia (GoE) in 2002 and also hopes to illustrate how the extremely rich information base in Ethiopia can be further used to address important and pressing policy questions. 3. Information base. The study has particularly drawn on the 1995 and 1999 Household, Income, Consumption and Expenditure Survey (HICES), the accompanying Welfare Monitoring Survey (WMS), the 1994-2000 Agricultural Sample Surveys, the 2000 Demographic and Health Survey and the 1999 National Labor Force Survey. The GoE i s currently undertaking another round o f the HICES and WMS surveys and has launched a national participatory poverty assessment inthe fall o f 2004. Unfortunately, the insights from both surveys were not available in time for inclusion in this report, but will provide the basis for follow up analysis. The insights from the available nationally representative surveys are complemented with information from secondary sources and the findings from analyzing a panel o f 1,500 rural households spanning 1994-1999 in 15 purposively selected villages and a panel o f 1,500 urban households in the major urban centers covering the same time period. Both surveys have been conducted by Addis Ababa University and its collaborating institutions. The study further builds upon and cross references several other economic studies conducted by World Bank teams such as the World Bank Country Economic Memorandum and the World Bank Country Status Reports mentioned above. 4. Striking statistics. While it i s neither our intent, nor indeedpossible, to capture the complex reality o f people's livelihoods in a series o f statistics, the following broad findings, each o f which need further qualification, are worth highlighting,before discussing the overall insights and recommendations o f the report. They epitomize key features of people's environment and their daily struggle to survive: Risk permeates life in Ethiopia, and shocks can have long lasting damaging effects. "Households that reported to have suffered substantially more during the 1984/85 famine continued to experience two to three percent less annual growth per capita between 1989 and 1997 compared to those that suffered substantially less." 0 Remoteness defines daily life inrural Ethiopia. "Only 13 percent of the rural population has a radio, three times less than the Sub Saharan African average, and 87 percent of the rural population are not exposed to mass media (radio, TV, newspaper) at least once a week. Households are on average 10 kilometers away from a dry weather road and, more importantly, 18 kilometers from public transport services. Eighty-five percent of the rural population has been continuously residing in the same district (woreda) as they were born, i,e. without ever having lived elsewhere. Clearly, physical and informational isolation combine to exclude the Ethiopian population from exposure to new ideas and injuences. " 0 Soil nutrient depletion continues at a fast rate. "Preliminary estimates suggest that the annual phosphorus and nitrogen loss nationwide due to dung removal is roughly equivalent to the total amount of commercialfertilizer annually applied." 0 A significant number o fpoor Ethiopianhouseholds arenet cereal buyers. One infive rural Ethiopian households lives on less than 0.08 ha per person', which yields on average only slightly more than half the daily cereal caloric needs per person, given current cereal production technologies used in Ethiopia. A land poor class living on hunger plots is rapidly emerging, with evidence of engagement in multiple off$arm activities as a coping strategy. " 0 Gender inequalities are pronounced. "Rural girls are 12percent less likely than boys to be enrolled in school. In other words, one million people are denied schooling merely on the basis of their gender. Domestic violence is a deeply rooted, culturally accepted practice, with 85 percent of Ethiopian women believing that a husband isjustijied in beating his wifefor at least one of thefollowing reasons: burningfood, arguing with him, going out without telling him, neglecting the children, or refusing sexual relations. '' 0 High returns to education especially for women and access to information provide hope. "An additional grade of primary schoolingfor male and female adults would increase consumption by 1.4 and 1.7percent respectively. This corresponds to one year of economic growth at its average pace since 1991. The recent increase in investment in primary schooling for boys and girls by the government thus holds hope for people's future well-being. There also appear important returns to access to information ' This average excludes SNNPR where many farmers grow and consume enset as well as coffee, and average land size per household is even lower. 11 with rural households with a radio estimated to be 17.5percent better off than those without. Moreover, an increase in the proportion of households in a community owning a radio by ten percentage points is estimated to increase households' consumption in these communities by 3.9 percent. This points to the existence of important externalities of access to information. " How well have the Ethiopianpeoplefared? 5. The micro and macro evidence paint a picture of limited to no decline in consumption poverty incidence since 1992 (see Figure i-a). There is a growing consensus that poverty incidence in urban areas i s increasing, while rural poverty incidence may have decreased slightly, by one or two percentage points. Overall, consumption inequality in Ethiopia remains low (Gini=0.29), though inequality in urban areas i s on the rise. The reasons behind these broad trends are largely found in the disappointing performance of the agricultural sector, which barely kept up with rural population growth. Whatever poverty reduction occurred in rural areas probably resulted from improved access to services and infrastructure. The government estimated rural poverty incidence in 1999 at 45 percent. Figure i-a: Evolution of povertyincidence between 1989and 20042'3 TheMDG for basis assumedpower 1995 HICES/WMS SuNey 1999 HICESNMS survey 3 25-.~................. ~... ~ .... .... . ~ .... ~...~~ ......... . ~ . ~ ....~ ~ ~.......... ~ . . ~ ~ . ~ ~~ 20 ............... . . ..................................... ~ ~ ~ .....~~ ..............~~~ ......... ~ ~ ~ ~ ~ ~ . . . . ~ . . . . . . a 5 g 15 - ...............~ ~ ~ ~~.~.~~ ............... ~~~~~~~.~~........... ~~~.~~ .. ~ ~ ~ .... . . . ................~~ ............ ~ . ~ ~ .......~ .~ .. ..............~~~.~ ~ ...... ~~ ~ ~ . ............~~~~.~ . 10 ~ ~ ....... ... .~~ ................... ~ ................. ......................... ~.~~ ..........~.~ ........~ ~ ~ ~ . . ~ . . . ~ . . . ~ ~ ~ ...... ....... 0 1 , Source: Own calculations 6. This still leaves the question why poverty Inurban areas increased despite substantial growth in the service sector. Overall economic growth was indeedlargely fuelled by growth This evolution is obtained by applying the historical sectoral growth rates from the national accounts to the 1995 household consumption levels from the 1995 HICES survey when the poverty headcount was estimated at 38.3 percent. To do so, each household has been classified across the agricultural, industrial or service sector based on the main employment sector o f the household head. The poverty levels in 1999 obtained using this methodology are similar to those found in the 1999 HICES survey, which provides confidence in the methodology. The poverty headcount levels reported in this figure are based on revised poverty calculations. While the reported levels are slightly different from the official numbers, the reported trends across both the 1995 and 1999 survey are similar irrespective o f the methodology used. 111 ... in the service sector (estimated at seven percent annually) during this period, though its benefits for the urban population were also substantially eroded by urban population growth (estimated at 4.7 percent) following rural-urban migration. Moreover, growth in the service sector was mainly driven by government expansion, with an increase inmilitary expenditures between 1995 and 1999 to finance the border war with Eritrea. While this was followed by a shift out o f defense into poverty sectors since 2000, it was probably too early to already feel the positive effects o f these more recent investments in the poverty sectors (e.g. doubling o f expenditures on education). Also, some o f these investments had a deliberate rural bias, consistent with the potential small decline in rural poverty in the absence o f agricultural per capita growth. Finally, the observed increase in urban inequality suggests that the benefits from growth inthe service sector were unevenly distributed during this period. Urbanpoverty incidence was officially estimated at 37 percent in 1999. 7. The slow pace of poverty reduction does not mean that everyone everywhere endured the same fate. Indeed, the averages hide a substantial amount of chuming. Many people move in and out o f poverty, often in tandem with annual rainfall patterns. This underscores the immediate impact o f rainfall on households' current consumption. Moreover, there are clear signs that the negative effect o f severe rainfall shocks often persists over time. For example, households who reported to have suffered substantially more during the 1984/5 famine experienced on average 16 percentage points lower growth in the 1990s, and 10 percent lower than average rainfall in a single year was found to reduce consumption growth rates by one percent four to five years later. Harvest failure proves to be especially harmful for child growth and female enrolment, thereby permanently damaging the earning prospects o f the next generation. 8. In addition to droughts, HIVIAIDS, and also malaria, continue to present substantialrisks to the Ethiopianpopulation. The prevalence o f HIVIAIDS is estimated at 4.4 percent o f all adults infected (2.6 percent inrural and 12.6 percent inurban areas) in2003. The latest estimates further suggest that the rate at which the HIVIAIDS epidemic is progressing, declined over the past years, especially in urban areas, which i s in line with observed changes inbehavior. Nonetheless, with about 1.5 million people currently infected and an estimated 539,000 AIDS orphans, HIVIAIDS continues to threaten future development and poverty reduction in Ethiopia. While largely neglected, malaria is also a major contributor to the disease burden in Ethiopia, third only to acute respiratory illness and perinatal disease. HIVIAIDS and malaria account for 6.2 and 4.5 percent o f child deaths respectively. 9. Poverty is more prevalent among the uneducated and among agriculturalists. Poor Ethiopians tend to live in large households and households with highdependency ratios. Households with older heads also tend to be poorer. Urban female headed households appear worse off, though the evidence regarding female headed households in rural areas i s somewhat ambiguous. Perhaps the most startling finding i s the very strong correlation between educational achievements and poverty, with the marginal returns to education positive and high, both for male and even more so for female adults. Poverty incidence among cash crop producers (especially chat growers) i s estimated to be substantially lower than average. However, households growing coffee appeared to be equally poor as the rest o f the nation duringthe 1995 - 1999 period. This is surprising, especially giventhe coffee price iv peaks in 1997 and 1998. More generally, the well-being and livelihoods o f coffee growing households is poorly understood and deserves further investigation given the numerical importance o f this group (about 30 percent o f the rural population grows some coffee) and in light o fthe recent collapse ininternationalcoffee prices. 10. Human capabilitieshave substantiallyimproved, albeit from extremely low levels and with a lag in impact on income poverty. Starting from a low base, Ethiopia's enrolment expansion at all levels o f education has been impressive, with the number of studentsinthe first level o fprimary school (grades 1-4) almost tripling since 1994 (see Figure i-b). National primary gross enrollment in grades 1-4 was estimated at 83 percent in 2000- 2001, up from 30 percent in 1994-1995. Yet substantial gender and regional differences remain. Despite some progress, at around 55-60 percent, pre-school child stunting remains among the highest in Sub-Saharan Africa, imposing a substantial drag on the development o f the next generation and fbture economic growth. Under-five child mortality dropped from 216 per 1,000-live births to about 169 between 1984 and 1999-2000. The population share with access to safe water increased from 21 percent to 27 percent between 1995 and 1999. Yet about one-third of households continue to rely on open rivers and lakes as their main source o f drinking water. Finally, it should be noted that the difference in progress between people's human assets and their well-being in monetary terms i s likely related to the time discrepancy, i.e. the observed progress inhuman capabilities i s not only quite recent, it mainly concerns Ethiopia's children (primary enrollment rates, child mortality, child malnutrition). To strengthen people's human capabilities more rapidly, the government is also promoting adult training and it has recently substantially expanded the number o f agricultural and health extension agents. Similarly, adult education (e.g. through adult literacy programs) deserves more attention. Figure i-b: Enrollmentsin grades 1-12, Ethiopia, 1967-2002 k o 4.0 D h z 3.0 2.0 1.o 0 Source: World Bank, 2004c V 11. Women (but also pastoralists) (40 million people in total) appear particularly disempowered--"The husband's beatingstick is like butter."4The widespread acceptance o f violence against women by women themselves epitomizes the deeply rooted existence of pronounced gender inequalities. Results from a nationally representative household survey conducted in 1999 by the Central Statistical Authority o f Ethiopia indicate that 85 percent o f women believe that a husband i s justified inbeatinghis wife for at least one o f the following reasons: buming the food, arguing, going out without telling, neglecting the children, and refusing sexual relations. Moreover, these attitudes seem to be associated with development outcomes: the under-five mortality rate for children o f women who do not accept any o f the given reasons as justification for abuse is 154 out o f 1,000 live births, while for those accepting at least one reason the rate i s over 192. The same survey reveals that 60 percent o f all women support female circumcision. Women consistently have lower educational attainment than men, with over 75 percent o f women having received no education at all (compared to 50 percent o f men). With only 6 percent o f all rural women exposed to mass media (radio/TV/newspaper) at least once a week (compared to 20 percent o f rural men), they are virtually excluded from any outside information and thus other perspectives on life. They have little representation in decision making positions. Nonetheless, as illustrated in Picture 11.1, p. 217, sporadic civic reaction to violence against women i s emerging. While there i s less quantitative evidence on pastoralists' disadvantaged position, it i s recognized. 12. More broadly, the full effects of government reform and action to empower citizens have yet to be fully felt by citizens and sub-regionalgovernments. The GoE has committed itself through the Constitution and the SDPRP to the empowerment o f citizens through decentralization. In addition, specific laws, policies and initiatives have been launched to address the position o f women and pastoralists. However, transitional processes have moved slowly, meaning that the potential effects o f the GoE's efforts are not yet fully realized. Partly as a result o f this, the strength of informal, traditional practices persists at the community level, where resistance to change represents a tremendous impediment to opening opportunities, particularly for traditionally disempowered people such as women and pastoralists. Moreover, the shift from an historically deep-rooted political culture favoring a strong role for the federal level government to devolving real power to local government and people is difficult to bring about. As a result, opportunities for meaningful participation o f Ethiopian citizens in governance and political life to shape their lives are currently few. This i s further exacerbated by the few resources Ethiopian citizens can claim access to, and their immobility and isolation from the outside world, limitingthe extent to which they are able to take action to improve their own lives. Households startfrom an extremely low endowment base 13. Ethiopia's private endowment base is extremely low and has largely remained so over the past decade. While the educational status o f Ethiopia's population has beensteadily improving over the past decade, educational attainment remains limited. Male adults completed on average only 1.8 grades; female adults only 0.88 grades. Disease and malnutrition further erode labor productivity. Landpressure has increased tremendouslyover the past decades, with average landholdings declining from 0.5 ha per person in the 1960s to Saying inAmharic. vi 0.11 ha per person in 1999. While about 40 percent o f the population have at least one ox, only 30 percent have two, the number necessary for an ox-span to plough the fields. 14. A land poor class living on "hunger" plots has emerged. Since the land reform under the Derg regime, Ethiopian agriculture is essentially characterized by smallholder farming. Nonetheless, land inequality i s high (Gini=0.47). This does not follow from the existence o f a small class o f large landowners, but i s rather the result o f the continuous fragmentation o f landholdings and the emergence o f a rapidly growing group o f people living on "hunger plots". For example, given current technology, about one fifth o f all rural households (excluding SNNPR) do not manage to produce half o f their annual cereal caloric needs from their plots, despite beingmainly dependent on agriculture. 15. Soil nutrient depletion and environmental degradation appear to be substantial. The most important source o f cooking fuel is firewood, used by almost 75 percent o f households. About one insix households uses mainly dung cakes as a source o f cooking fbel, resulting ina continuous depletion o f the soil at alarming rates. Preliminary estimates suggest that the annual phosphorus and nitrogen loss nationwide due to dung removal i s about equivalent to the total amount o f commercial fertilizer annually applied. While the extent to which the nutrient content o f dung translates into actual alimentation o f the soil with phosphorus and nitrogen depends critically on the handling o f the dung and the cultivation methods, these results are sufficiently large to warrant a much closer investigation o f the effect of dung collection on agricultural productivity and poverty. Inaddition to soil nutrient depletion, soil run-off also poses an important environmental threat, with experiments in Tigray indicating that soil loss could be substantially decreased through stone bunds (on average by 68 percent). 16. Remoteness epitomizes daily life in rural Ethiopia. Rural households are on average 10 kilometers away from a dry weather road and 18 kilometers from any public transport, rendering people immobile. Indeed, 85 per cent o f the rural population live in the same area (woreda) as they were born, without ever having moved. Only 13 percent o f the rural population has a radio, the lowest incidence in Sub-Saharan Africa (see Figure i-c), exemplifying the sheer disconnect o f Ethiopian's population from the rest o f the world, not only interms o f market access, but also interms o f access to information. vii Figure i-c: Incidenceof radio and TV ownership among rural householdsin selectedSub-Saharan African countries, 1995-2001 110 100 90 80 C 70 0 3 60 .c 0 3 50 0 40 r 0 a, a 30 v) 20 I c 10 s0 0 Note: Incidence of radio ownership is added to incidence of TV ownership. As a result, total incidence can reachup to 200 % maximum. Source: Own calculationsfrom 1995-2001 DHS surveys 17. Risk and drought shocks havesevere and longlastingeffects on poverty. Ceteris paribus, households in areas characterized by larger rainfall fluctuations were found to be poorer. The effects o f crop damage due to droughts, pests, insects, frosts, or other causes on consumption were substantial. For example, it was estimated that 2.7 percent o f consumption per adult equivalent was lost in 1999 due to crop damage, a year characterized by average rainfall. This corresponds to about 1.5 years o f GDP growth per capita at its historic rate. Moreover, not only are households unable to protect their consumption from continuously recurring shocks, but the effect o f these shocks are often long lasting. Micro-econometric evidence shows that households that suffered substantially during the 1984-1985 drought, which resulted in a nationwide famine, continued to experience two to three percentage points less annual per capita growth during the 1990s than those who weren't hit as hard. Although there i s little evidence that short-term illness affects incomes and consumption, serious illness episodes tend to reduce consumption substantially. 18. Fluctuations in consumption due to shocks also have important gender dimensions, with men typically experiencing less severe fluctuations in their welfare than women. Furthermore, the experience o f an income shock leads parents to disinvest in their children's schooling, especially their daughters'. Households with half o f their plot area damaged were found to be 2.6 percentage points less likely to send their daughter o f school- going age to school. viii Livelihoods are agriculture based, but labor productivity in agriculture is low 19. Rural households continue to rely heavily on low input, low output, subsistence oriented, rainfed agriculture and agriculture related activities. Agriculture is responsible for 85 percent o f employment, 45 percent o f national income, and more than 90 percent o f exports. O f the total area under temporary crops in the 1 9 9 0 ~cereals accounted for 88.7 ~ percent. Commercial fertilizer i s applied to approximately 40 percent o f total farmland under cereals over the past years, and i s heavily concentrated on a few cereals (wheat, teff, and maize). Improved seeds are only applied on about five percent o f the total cultivated cereal area and pesticides on about seven percent. Less than one percent o f the total cultivated area inEthiopia is irrigated, despitemassive fluctuations inrainfall. The limited (and only slowly expanding) use o f inputs and modern technology i s consistent with the low average yields in cereal production hovering around 1.1-1.2 tonsha (about one-fifth o f the yields observed in Asia since the Green Revolution), as well as the low marginal productivity o f labor observed inthe data. Ethiopiastill finds itselfat the verybeginningofits structuraltransformation. 20. Evidence suggests significant numbers of net cereal buying poor households in rural Ethiopia, an important finding in discussing the food price dilemma. Micro evidence suggests that in 1996, 53 percent o f all rural households were net cereal buyers. Moreover, the bulk o f the marketed surplus i s produced by a minority o f producers. This i s consistent with the evidence from other poor SSA countries. Poorer households are more likely to engage inmultiple activities as a coping strategy. Especially livestock products, but also business activities (collection o f water and fuel wood, artisanal activities, grain trading,) and off-farm wage work (especially food and cash for work) provide sources of cash income to buy food. Inaddition to the urban (poor) population, these rural households will gain from a (gradual) decline in cereal prices. This i s in contrast to the smaller set o f poor rural households which are net cereal sellers. 21, Substantial potential remains for increasing productivity in staple crop production through agricultural intensification,especially in food secure areas. Based on a review o f the available evidence, rough estimates suggest that doubling cereal yields in the more food secure areas and increasing cereal yields by 50 percent in the food insecure areas lie well within the realm o f the possible. Much could be gained from better cultivation techniques, broader and joint adoption o f both fertilizer and improved seeds, and increasing market access (better access to roads and rural towns) in the more food secure areas, while wider adoption of fertilizer-improved seed packages will have to be especially complemented with the promotion of soil conservation and better water and risk management techniques as well as improved market access in the food insecure areas. Consistently, the government has recently complemented its longer standing extension program, geared at promoting modern input use, with soil conservation programs and rapid expansion of water harvesting. It is also experimenting with broad bed ploughing o f vertisols permitting two crops per year and thus potentially a substantial increase inyields. 22. Nonetheless, to realize the unrealizedpotential in staple crop production, it will be important to further our understanding of the major factors constraining wider adoption and diffusion of land saving technologies. The role o f political and land tenure security in the adoption o f irrigation and environmentally sustainable cultivation practices i x respectively, as well as the role o f having effective risk management strategies in adopting modern inputs deserve further investigation. Anecdotal and empirical evidence suggests that the demand by poorer households for fertilizer-improved seed packages is limitedbecause of the downside risks involved. Similarly, input delivery systems may have to be made more efficient and demanddriven for these strategies to be more effective. 23. In addition, the low marginal value of labor in terms of additional agricultural income from cereal production,given current landholdingsize, and the higher marginal value of expanding landholdings, holding labor input constant, suggest the need for complementary policy routes to expand households' income. As noted, one option is raising marginal productivity o f labor and land through agricultural intensification in cereal production. Second, the land frontier could be pushed further and new areas could be exploited, 1.e. agricultural extensification. Third, labor productivity could be increased through diversification into non-cereal (tradable) agricultural production. Fourth, land pressure could be reduced and labor productivity enhanced through diversification and migration out o f agriculture into highly remunerative non-agricultural activities. The optimal combination will obviously differ across space depending on the region's comparative advantage in terms o f agro-ecological potential, and market access as determined by population density and access to infrastructure. Further work to help identify and quantify these optimal combinations across products and space i s called for. A balanced road out ofpoverty - therole of agriculture and agency 24. Given the findings noted above, a balanced approach to reduce poverty presents itself. Macroeconomic projections suggest that reaching the MDG o f halving poverty incidence by 2015 from its 1990 level will not be possible without buoyant agricultural growth. Simulations indicate that a decade o f 4.1 percent agricultural growth would reduce the current poverty headcount by about one third. While such growth in agriculture will be necessary to sustain the required growth in the non-agricultural sectors and facilitate the structural transformation with labor shifting out o f agriculture into industry and services over time, it will not be sufficient. Enhancement o f individual agency and empowerment, both in the economic and social space, and increased foreign aid will be equally necessary to achieve growthwith poverty reduction. Agriculture and rural development 25, The contribution of increased agricultural productivity to economic growth and poverty reduction works through consumption and productionlinkages. Inparticular, a productivity increase innon-tradable activities such as cereal production leads to lower prices, effectively increasing consumers' real incomes. There will be important direct gains through decreased food prices for all net cereal buyers, which in most years make up the majority o f the Ethiopian population. However, the greatest benefits are to be expected from the consumption linkages, whereby increased demand for locally produced goods and services following the decline o f food prices inresponse to a productivity increase, generates off-farm employment and subsequently increases the demand for food as well. To generate sizeable multiplier effects, the income elasticity for locally produced non-food non-tradables must be large local, supply elastic, and labor intensive. The productivity gain must concern a non- X tradable with a high average budget share such as cereals, which constitute 30 to 40 percent o f total expenditures among the poor inEthiopia. Yet, net cereal sellers could lose if demand i s inelastic. 26. Production linkages can occur when increased productivity or higher prices in the productionof tradables positivelyaffect the incomes of producers. The direct poverty reducing effect may be substantial if assets are equally distributed and access to complementary inputs (e.g. fertilizer, improved seeds) is universal. To maximize the poverty reducing effect, promoted technologies should be scale-neutral and labor intensive. Multiplier effects through backward linkages (increased demand for inputs) are usually limited, since inputs are generally capital intensiveand imported. Nonetheless, important externalities may exist through increased local availability o f inputs for other (non-cash crop) agricultural activities. 27. Agricultural extensification has so far been the key force behind growth in agriculture, though its potentialto further boost agriculturalgrowth is limited. Not only has the ongoing expansion o f the land frontier not been sufficient for growth in agriculture to keep up with population growth, let alone to sustain growth o f 4.1 per cent, the scope for further agricultural extensification is rapidly decreasing, especially in light o f continued population growth o f 2.5 to three percent per year. Some potential for area expansion remains, in the lowlands, when they become more accessible through malaria and tse-tse fly eradication, and better public services, which are potentially costly. This approach is consistent with the philosophy behind the ongoing resettlement program, though sustainable population movements could be equally achieved through well-functioning labor markets, an area which i s still poorly documented and understood inthe Ethiopian ~ o n t e x t . ~ Table ia: Cereal yield and input use infood deficit, food balanced and food surplus areas, 2001/02" Food deficit Foodbalance Food surplus Cereal yield (toniha) 1.08 1.19 1.44 Cereal yield usingfertilizer only 1.24 1.25 1.44 Cereal yield usingfertilizer &improved seed 1.65 2.20 2.63 Absolute difference between using fertilizer & improved 0.57 1.01 1.19 seed compared to average cereal yield (tonha) % difference between usingfertilizer & improved seed 53 85 84 compared to average cereal yield (tonha) Fertilizer use rate in cereals (% area) 29.12 26.40 56.13 Fertilizer combined with seed rate (% area) 3.08 3.15 4.88 1)Calculated from Agricultural Census, 2001102 (Federal Democratic Republic of Ethiopia, Central Statistical Authority). Source: Diao et al., 2004 This is being addressed in the ongoing Labor Market Study undertakenby the World Bank in collaboration with the GOE. x i 28. Agricultural intensification in cereal production through the promotion of increased modern input use will continue to play an important role in raising incomes and reducing poverty. Given the large share o f cereal consumption in (poor) people's budget, and the existence o f a substantial group of (poor) net cereal buyers (rural and urban), it will be important to sustain the focus on increasing cereal productivity to prevent cereal prices from rising rapidly which would hurt the poor (as well as many non-poor). Evidence suggests that there i s still significant scope for improvement, in particular through increased use o f combined modem input packages (fertilizer and improved seeds), and especially in the food secure areas (see Table ia). However, agricultural intensification in cereal production through increased modern input use alone will clearly not suffice and the constraints to further technology adoption (risk management, input delivery systems, market development) must be better understood. 29. Increasingcereal productionin the face of priceinelasticcereal demandmay lead to large cereal price declines. When increased production is due to reversible productivity increases (e.g. through the use o f fertilizer and improved seeds), large (and undesirable) cereal price fluctuations may follow. Or, when productivity increases are due to irreversible investments (e.g. infrastructure facilitating better water and soil management), it may put cereal producers on a price treadmill whereby producers see the gains from their investments being eroded by lower cereal prices, which will force them to either withdraw from agriculture or engage in another round o fproductivity increases. 30. However, these risks must be put in the right context and can also be managed. First, for the 2000-2002 experience, it is generally agreed that the observed collapse in (especially) maize prices was compounded by food aid mismanagement, with food aid being imported while it could have been locally procured to help support local cereal prices. Clearly there i s a role for more effective and non-distortive management o f food aid in preventing undesirable cereal price movements: this is currently being pursued under the productive safety nets program. Second, in contrast to maize, the demand for other cereals such as teff and wheat i s more sensitive to prices (as i s their supply). The observed price collapse in 2000-2002 was indeed the largest for maize. Third, it is important to recognize that for those (rural) poorer households who are net cereal buyers as well as the urbanpopulation, declining cereal prices increase their real incomes. 31. To avoid large price fluctuations and/or a price treadmill in cereals, better food aid management and complementary actions especially in market development and agricultural diversification are needed. A simultaneous increase in the production and productivity of non-staple tradables (livestock, traditional and non-traditional agricultural export crops), in addition to increased cereal production, can foster production linkages, help generate off-farm employment, and generate demand for food which will prevent food prices from collapsing. A more balanced agricultural growth pattern will also facilitate migration out o f the food insecure areas and maximize the linkage effects and thus poverty reduction. To do so, focused interventions in staple and non-staple agriculture such as agricultural research and extension will need to be complemented with market development, i.e. proper incentives for farmers and traders, a facilitating institutional environment, and infrastructure to improve market connectivity. xii 32. In addition to diversification across agricultural products, agricultural growth strategies will also need to be spatially diversified. While there is still some scope for intensification o f foodcereal production in the northern dry lands o f Ethiopia, this should be complemented with promoting livestock production and tree planting. This underscores the desirability o f the broader extension packages currently pursued by the government. Moreover, to successfully intensify food production, promotion o f the use o f fertilizer and improved seed packages (e.g. for sorghum) will need to be complemented with the adoption o f soil conservation structures (e.g. stone terraces) and sustainable land management practices as well as better water and risk management. The current food for work programs when complemented with sufficient technical assistance could be usefully mobilized to help build this infrastructure. 33. Inthe highpotential cereal producing areas ofthe central and northwesternhighlands, a continued focus on intensive cereal production through increased use o f input packages i s warranted given the existing scope for substantial yield increases. This strategy could be usefully complemented with the development of dairy production in areas closer to urban markets. Continued focus on intensification o f food production through improved seeds and fertilizer use, and concerted efforts to increase productivity of coffee production and marketing efficiency hold promise to foster agricultural growth and reduce poverty in the humid high potential perennial zones in the southern and western highlands. In addition to further intensification of cereal crop production, development o f the dairy industry in peri- urban areas, tree planting, and the promotion o f non-traditional agricultural exports, including floriculture and horticulture products, also hold promise in Central Ethiopia around Addis Ababa, initially without much additional public investment. 34, Many public investments benefit both agricultural and non-agricultural growth. While an agricultural led growth strategy does require continued investments in agricultural research and extension and other agriculture specific investments which are not necessarily costly (e.g. soil conservation through stone terracing), many o f the necessary investments to foster agricultural growth (both in staple and non-staple crop production) actually lie outside agriculture, such as investments in infrastructure, risk management, education, health, and access to information. As usually both agricultural as well as non-agricultural activity stand to benefit from these investments, the debate about agriculture versus non-agriculture i s often illconceivedfromthisperspective. 35. Promoting market connectivity through improved access to roads in addition to access to informationwill be key to stimulate and distribute the benefitsfrom increased agricultural and non-agricultural production. The micro-evidence shows that access to markets as proxied by distance to urban centers and roads contributes greatly to increased agricultural production and diversification o f agriculture into non-food production. For example, in regions which are overall better connected, the estimated elasticity o f cereal output to market accessibility (defined as the population size o f the nearest town or big market divided by the road distance to this town or market) usually exceeds 1.2. However, the estimated elasticities are much lower for the remote northern regions covered in the same study, which may suggest the existence of infrastructure threshold effects below which the returns to connectivity are limited. Roads bring direct short term employment, generate access to markets and services, facilitate migration and exchange o f information and ideas, xiii and bring long term off-farm employment opportunities. Continued emphasis by the govement on expanding especially the rural road network as envisaged in the current road development plans appears warranted. The potential existence o f threshold effects should be explored further and subjected to empirical scrutiny. 36. Policies to strengthen households' asset bases should be supplemented with promoting a broad range of ex-ante and ex-post risk managementstrategies, also in the food secure areas. Ex ante options includebetter water management. The micro evidence on the large immediate and persistent effect of shocks on people's welfare underscores the large benefits o f containing any crisis and the critical importance to support those affected by a crisis well beyond the initial crisis period. Moreover, both in food secure and food insecure areas, risks and the absence o f efficient tools to cope with shocks ex post may prevent many poorer people fkom adopting more productive, but higher risk, production technologies such as fertilizer and more remunerativecrop portfolios. 37. A combination of small scale irrigation, weather based insuranceand productive safety nets hold promise to help farmers better manage their risks. Strengthening farmers' capacity to effectively manage their risks could be done either by helping them to better mitigate the effects o f shocks (e.g. irrigation, pest and plant disease management) or by increasing their capacity to cope with shocks ex post (e.g. better targeting o f food aid in response to shocks, development o f weather index based insurance schemes, strengtheningo f the existing informal insurance schemes, or through contingent transfer programs such as foodcash for work or for education). The cost effectiveness and implementation constraints o f each o f these interventions needs to be investigated further. The following three interventions deserve particular attention: 0 Supplemental irrigation. There are compelling reasons for Ethiopia to focus on irrigation in general, and small scale irrigation in particular, for poverty reduction. First, unreliable rainfall i s the leading cause o f harvest failure and hunger. Second, the availability o f new irrigation technologies (low cost drip systems) make small scale irrigation possible, and open up new opportunities for water conservation. Finally, the country already has successful experience with such a strategy: over 66,000 Ethiopians are reported to enjoy higher crop yields due to small scale irrigation through the Ethiopian Social Reconstruction Development Fund. The government has recognized Ethiopia's irrigation potential and has identified water harvesting and small scale irrigation as a key instrument for reducing vulnerability and poverty in the SDPRP. A recent evaluation o f the experience inTigray suggeststhat household incomes could be substantially increased through investment in rainwater harvesting ponds provided that they are close to the homestead, that they are properly constructed, that households grow high value crops (such as vegetables), and that households receive adequate extension support. Given that poorly planned irrigation programs also introduce their own risks (e.g. increased malaria incidence), the current water harvesting program should be closely monitored and their impact further evaluated. The internal constraints to wider adoption o f irrigation techniques should also be better understood, as recent developments in the multi-country dialogue on the development o fthe Nile Basin open up important opportunities. xiv 0 Weather based insurance: Irrigation will not be possible for many farmers and rainfed agriculture will continue to be at the core o f their livelihoods for years to come. An innovative low-cost risk management tool, which i s much less prone to the usual moral hazard issues, i s to insure farmers against drought risk through formal contracts with private insurance companies or public institutions. Such contracts insure the contracting party against a specific and objectively verifiable rainfall outcome, e.g. drought, and may be entered into by farmers directly, by credit institutions, or by governments. Such schemes are already available to poor farmers in India, Mexico, and South Africa. There are many ways to deliver this insurance. The contract could stipulate a cash payment to participants upon the realization o f the event, or it could (partially) forgive loan repayment on an input (e.g. fertilizer) if the rains fail and thus foster technology adoption. Weather based insurance schemes are currently piloted inEthiopia. 0 Productive safety nets: Existing safety nets have saved lives but have been largely unproductive and often not well targeted. Yet they can continue to serve their vital insurance function while beingmade more productive through a mix o fprograms aimed at building productive physical and human assets. In particular, guaranteed multi-annual transfers to households in return for participation inpublic works and targeted health and education programs can: (1) encourage risk-taking behavior among small-holder farmers by insuringagainst downside risk o f consumption loss; (2) buildpublic infrastructure and maintain community assets, which provide complementary inputs to private inputs and improve the productivity o f individuals; and (3) promote market development by increasing demand inplaces that are otherwise too poor. 38. Remaining institutional and resource obstacles to the generation of off-farm employment and private sector growth must be removed. While agricultural intensification will generate demand for locally produced goods and services, and thus create local employment, institutional constraints in factor markets must be further addressed to facilitate an appropriate off-farm supply response and maximize the linkage effects. Private sector growth in off-farm activities will also be necessary to sustain robust growth inthe non- agricultural sector which has so far been largely fueled by government expansion. While the government has recently introduced new policies in favor o f the private sector and has been very active in improving its dialogue with the sector, significant obstacles remain. The World Bank Country Economic Memorandum and the upcoming Investment Climate Assessment present a detailed review o f both progress in and continuing constraints to private sector development and off-farm employment generation. A particularly critical issue, which has also been raised during participatory poverty assessments, relates to the availability and cost o f urban land to set up a business, both in urban areas and rural towns. While the reform of urban land i s now in full swing, progress has been slow and substantial residual uncertainties for private investors persist. Agency 39. To further people's well-being it will also be critical to continue to unlock their potential and strengthen their agency and opportunity structure, as indicated in the SDPRP. Strengthening people's agency will require large improvements in their human xv capabilities (being well educated, healthy, and well-nourished) as well as in their access to informational resources, over and above the need to substantially increase their incomes and material and financial resources. In the face o f continued poor performance on voice and accountability and government performance indicators, the GoE has recently indicated a strong commitment to professionalize its governance apparatus and empower its citizens through political, fiscal and administrative decentralization. However, these necessary changes to improve people's opportunity structure are relatively recent and, as in other countries, processes o f change o f such magnitude are slow. Great strides have already been made in terms o f decentralizing to certain woredas and in urban areas, but government representatives and citizens report that progress toward making the state apparatus more responsive and accountable i s limited as o f yet. With formal institutions o f government in a state o f transition, there i s evidence that many citizens rely on their own informal forms o f organization and norms o f behavior to manage everyday life rather than those o f government. It will be important to minimize the risk that these informal practices exclude marginalized individuals and groups, thereby reducing their capacity to make effective choices about their own development. 40. The critical importanceof educationand especially female education for people's well being cannot be sufficiently underscored,and warrants a continuedfocus on public investment in education, even though the benefits may only be fully felt over time. One o f the most robust and most striking empirical findings o f this report i s the huge effect o f education and particularly female education on consumption poverty as well as on human development outcomes and people's ability to shape their own lives. The empirical simulations suggest that bringing all female adults up to at least a 4th grade education could reduce poverty incidence by 12 percent. Education will be necessary to help households adopt new technologies and thus enhance their agricultural productivity. It will facilitate migration out o f agriculture and marginal areas into off-farm employment and rural towns, thereby reducing land pressure among the remainingpopulation. 41. Female adult education is one of the most critical determinants of all human development outcomes. Giving the average mother in rural areas four years of education would increase children's survival by 5.1 percent, and bringing at least one female adult inthe household to 6thgrade would reduce child stuntingby up to 12 percent. Male adult education also has substantial positive effects on consumption and child malnutrition, though these are generally slightly smaller than the effects o f female adult education. Moreover, educational attainments o f parents and the community at large are critical determinants o f the education o f the next generation. For each additional year o f educational attainment by the household head, the probability that a child is enrolled in school increases on average by 1.1percentage points inrural areas and by 1.6 percentage points when it concerns a girl's enrollment. 42. The empirical analysis suggests a series of promising interventions to enhance primary school enrollments and completion. Merely enhancing households' income through overall economic growth will not suffice to reach the education MDG (universal primary completion by 2015) and specific actions will be required: 0 specific policy interventions to increase rural enrollment and completion rates should focus on: (1) increasing accessibility to schooling, especially by locating new schools in xvi unserved and underserved areas; (2) improving sanitation facilities and the availability o f water in schools; (3) using innovative pedagogical methods such as multi-grade teaching to promote school quality; and (4) strengthening the set o f risk management tools available to farmers, which would especially help inraising girls' educational attainment; actions to increase urban enrollments should focus more on improving the quality o f schooling by: (1) reducing student-teacher ratios, particularly in the early grades, where dropping is concentrated; (2) increasing the deployment o f female teachers; and (3) improving sanitation facilities and the availability o f water in schools; While attainment of the education MDGs presents a daunting challenge, the foregoing interventions would accelerate Ethiopia's progress, bringing the intermediate objective o f 100 percent gross enrollment inGrades 1-4 within reach in the near future6. 43. The potential of adult literacy programs deserves further attention. Given the critical role o f adult education for both economic development and human development outcomes, the potential for adult literacy programs which were common during the Derg regime, should be explored further. This could timely complement the government's ongoing efforts to foster primary school enrollment rates. The large externality effects on primary school enrollment o f the average educational attainment in the community, and o f female adult literacy in particular, further suggest that adult literacy programs would help increase enrollment rates as well. Finally, a more detailed ethnographic investigation o f why households are less likely to invest ingirls' education i s called for. 44. In addition to formal maternal education, specific health and nutritional knowledge also play a critical role in reducing child malnutrition and child mortality. The empirical evidence on the determinants o f child stunting suggests that child growth monitoring and maternal nutritional education programs could play an important complementary role to other development actions such as promotion of food security, income growth, and more general parental education, which are already underway. The role o f other direct nutrition interventions such as micronutrient supplements, promotion o f exclusive breast-feeding, and appropriate complementary feeding should be equally considered. Moreover, while it will take a considerable amount of time before these other development actions substantially affect pre-school child growth faltering, child growth monitoring and nutritional education programs as well as complementary feeding and micronutrient supplementation could take effect immediately, as illustrated by the successful ongoing Vitamin A supplementation program. It i s anticipatedthat the new Health Extension Program will take on some o f these challenges. The most promising (non-health) interventions to reach the MDGo f reducing child mortality by two thirds by 2015 from the 1990 level are enhancing maternal education and increasing access to safe drinking water. Given that 24 percent o f under-five child deaths are attributed to diarrhea, maternal health knowledge and behavioral change will be equally critical. ~~ An indepthanalysis of the educationsector inEthiopiais providedinthe 2005 World BankEducationCountry Status Report. xvii 45. Given the critical importance of early childhood malnutrition for future economic growth, reducing child malnutrition should be of great concern to the Ministry of Finance and Development. This will require increased awareness about the long term detrimental effects o f early childhood malnutrition on future economic growth. At the institutional and policy level, a comprehensive and coherent multi-sectoral nutrition policy will need to be developed, and the institutional responsibilities o f the various ministries, and mechanisms for coordination o f their actions in the field o f nutrition will need to be clearly delineated. 46. More broadly, a better understandingof the existence of synergies between, and the appropriate sequencing of, interventions is needed to inform a multi-sectoral approach toward improving human capabilities. The empirical analysis shows that irrespective o f the particular human development outcome (malnutrition, mortality, or education), important opportunities for improving these outcomes are to be found outside the particular sectoral realm. The institutional implications o f these findings must be further investigated. For example, in the case o f child stunting there are indications that income and community nutritional knowledge act as substitutes, suggesting that there may substantial gains from imparting nutritional knowledge, a relatively easy to implement and low cost intervention, even if people remain very poor. Yet, further investigation o f the potential substitution between income and nutritional knowledge i s necessary. There may also be thresholds below which these determinants begin to act as complements. Identifying the relationships between different determinants, and the potential thresholds below or above which they act either as complements or as substitutes will be critical in designing effective multi-sectoral interventions to reach the MDGs. 47. There is tremendous scope to enhance people's capability to aspire and expand their choice horizons, as well as to reducemonetary poverty by increasingtheir access to information for example through wide dissemination of radios and mass civic education programs. Information is a powerful transformer, and community radio programs are a commanding medium to transmit sound information for example about better farming techniques, hygiene and nutritional practices, market and weather conditions, etc., especially when the majority o f the population is illiterate and physically isolated, as in Ethiopia. It is estimated that providing the poorest quintile o f the population with a radio would increase average consumption by five percent, reduce poverty incidence by 11percent, and reduce the poverty gap by about 40 percent. 48. Radio programs are major dialogue initiators, often empowering individuals and fostering societal change. This is consistent with the important empirically observed extemality effects o f radio ownership within communities on poverty, and i s exemplified by the deeply disturbing story o f Woineshet, a 13-year old rape victim in southern Ethiopia whose father's decision to bring her case to court was prompted by his exposure while working in Addis Ababa to radio announcements and bus ads urging the prosecution o f rape cases (see Box 2.3, p. 40). Moreover, the marginal beneficial effects o f radio ownership are real, empirically robust, and high compared to the marginal costs, making investment in increasing access to information through community radio programs a cost effective and timely intervention to better connect rural Ethiopia with the rest o f the world, facilitate civic engagement and reduce their poverty. To do so will not only require increasing people's xviii access to radios but also providing an enabling legal framework fostering open debate and supporting citizens' rights to information, includingplurality o f radio licensing. 49. In addition to enhancing people's agency (Le. their human capabilities, their access to (and production of) information, their financial and material assets), people's opportunity structure should also be strengthened to foster empowerment of citizens in generaland women (and pastoralists)in particular. To improve the position o fwomen in Ethiopian society actions are recommended inthe legal, social and economic spheres: 0 Following the recent increase in the legal age at marriage to 16 years old, further support i s required to hone and deepen government strategies supporting equal legal protection o f women. This includes: (1) better aligning the penal codes and application o f existing laws to make them consonant with the word and spirit o f Article 25 o f the constitution and the National Policy on Women; (2) providing training on gender sensitization to judges, lawyers and other members o f the legal profession; (3) establishing a watchdog to track changes in the law and its application; and (4) supporting legal advocacy groups and providing legal aid, women's advisory centers and shelters for abused women. 0 The entrenched social norms and practices that discriminate against women should be addressed. The billboard actions against gender based violence undertaken by the Gurage Women and Teacher's Association are encouraging signs o f civic engagement in this context (see examples o f billboards inPicture 11.1, p. 217). Other entry points include: (1) ensuring that gender issues are appropriately addressed in all development interventions and government programs; (2) organizing training events for women parliamentarians and other champions o f women's issues on communication skills, computer literacy, gender budgeting, planning, monitoring and evaluation for women MPs; and (3) continuing the focus on girls' education and promoting the inclusion o f gender sensitivity programs in education curricula. 0 Women's participation in the formal economy should be increased by providing: (1) incentives to businesses to hire women; (2) business management training and follow-up support to women; and (3) expanded credit availability to female entrepreneurs. 50. To further facilitate the ongoing transition from traditional norms to national legal frameworks, actions should focus on continued support to existing programs o f decentralization and support to the development o f independent civil society. Particularly crucial to foster empowerment through these programs would be: (1) a continuous emphasis on capacity buildingat the woreda and kebele level to ensure effective use o f block grants for poverty reducing purposes; (2) the enhancement of the interface between kebele and woreda councils, and between citizens and both o f these organizational entities; (3) the increased involvement o f citizens inthe formulation o f kebele plans, budgeting and monitoring; and (4) establishing functional mechanisms o f accountability, including annual performance appraisals. 51, Despite encouraging signs of a deceleratingincreasein the spread of HIVIAIDS, the current focus on reversingthe spread of HIVIAIDS must be sustained, if Ethiopia's aspirations for economic growth and poverty alleviation are to be met. Ethiopia's HNIAIDS xix epidemic is generalized, having spread far beyond the original high-risk subpopulations. Continued progression o f the epidemic will simply undermine any current and future development efforts. Inparticular, successful implementation o f the innovative multi-sectoral HIVIAIDS program, which i s now in full swing, will require unrelenting support from the highestpolitical levels. An accurate databasewill also need to be urgently established. 52. In sum, a balanced approach to poverty reduction focused on agriculture and agency holds promise. A geographically differentiated focus on increasing labor productivity in cereal and non-cereal agvicultuve, including concerted efforts to strengthen people's ability to manage risks, will be needed Inaddition, to maximize the linkage effects and reduce poverty substantially, connectivity to urban growth centers will equally need to be improved and the investment climate ameliorated to foster off-farm employment generation and facilitate the structural transformation. Actions to strengthen people's ability to make effective choices, both in the economic and social space, will further be key. This will help unlock people's innate potential and will require enhancing their agency (Le. asset base) and fostering a conducive opportunity structure (i.e. institutional environment). To do so, a strong emphasis on female education and increased access to information will be needed. Empowering people in the economic space will also be necessary to effectively absorb the current and future resource flows, and issues o f pacing o f aid and the government's administrative absorption capacity will increasingly have to be at the center o f our attention as aid flows further increase. Finally, any prospect for poverty reduction in Ethiopia i s pre- conditioned on adequate containment o f HIVIAIDS. xx INTRODUCTION 1. The last decade and a half have ushered in a period of optimism in Ethiopia, pre- conditionedby several important internalsocial developments. The first o f these was the end o f the civil war in 1991. Except for the two years (1998-2000) o f the border war with Eritrea, which was costly both in lives and in financial resources, the fifteen year period since 1990 has been a period o f relative peace inthe recent history o f the country. This has allowed the govemment to channel its resources to more productive activities. Second, broad economic reforms were introduced by the new government in the first half o f the 1990s, which included shifting the economy from central planning to market mechanisms, keeping inflation under control, and improving the management of public expenditures. Third, the new govemment also fundamentally changed its governance structure. Its program o f political decentralization, which i s still ongoing, introduced a process o f empowerment o f sub-national governments and diluted one o f the sources of past conflicts--concentration o f power at the centre. Figure 0.1: Real gross domestic product per capita, Ethiopia, 1961to 2003 Real GDP per Capita, in Birr, 1961-2003 -+GDP/Capita Real - Trend I Source: World Bank, 2004c 2. As a result,the 1990s and the early 2000s are sometimes viewed as a period of broad recovery. One o f the first signs o f this recovery has been the overall output response o f the economy. Between 1991 and 2004, GDP per capita has risen from its lowest level to one o f the highest inthe past 40 years (see Figure above). To be sure, even at its peak in 2001, GDP per capita was only at a level last reached in 1972. Moreover, stark volatility o f growth has remained, and agricultural growth has continued to lag far behind the other sectors (industry and services) despite the prominence and attention it received under the reformist government. The good news i s that per capita GDP has risen by about 1.7 percent per year since the Ethiopian People's Revolutionary Democratic Front (EPRDF) assumed powere7 About 1.1 percentage points o f this growth can be attributed to the policy changes and public investments o f the reformist government, which boosted total factor productivity, while the rest i s partly the peace dividend (or catch up growth).' 3. These positive developments at the policy, institutional and macro-level beg two major questions: (1) How well did the Ethiopian people fare during this period of economic recovery? and (2) What can be done to improve their lives further? This study seeks to shed light on these two broad, overarching and complex questions which continue to underpinthe policy dialogue inEthiopia. The arguments and policy implications presented in this report follow from confronting insights and hypotheses put forward during elaborate consultations with the government and development partners with both development theory and rigorous empirical analysis o f Ethiopia's rich informationbase. 4. The report is comprehensive in scope, though some topics such as the role o f agriculture, risk and empowerment in poverty reduction are dealt with in more depth given their prominence in the ongoing policy debates in Ethiopia. The study will draw upon and cross reference several other economic sector works which have been undertaken in parallel with this study by other World Bank Teams, such as the World Bank Country Economic Memorandum and the World Bank Country Status Reports on Education and Health. As the report seeks to complement and extend the Poverty Profile o f Ethiopia published by the Ministry o f Finance and Economic Development (MoFED), Government o f Ethiopia (GoE) in 2002, it addresses people's well-being in a broader sense and includes issues related to empowerment and vulnerability. It further places a larger emphasis on analyzing the determinants of people's well-being and its policy implications in light o f the upcoming revision Ethiopia's Sustainable Development and Poverty Reduction Strategy (SDPRP). 5. To examine how Ethiopians fared over the past decade and a half, the report explores both monetary and non-monetary dimensions of well-being. This can also be referred to as a utilitarian and a capability approach.' In particular, the report explores progress on measures o f income or consumption to capture the monetary, or utilitarian, perspective, and focuses on such issues as nourishment, health, literacy, and the extent to which people in Ethiopian society are empowered to capture the non-monetary, or capability, approach. The report hrther distinguishes three different, though related, dimensions o f well- being, i.e. poverty, vulnerability and equality. In particular, the report examines whether people have enough o f what i s considered valuable compared to some external benchmark, i.e. it assesses their poverty status. It further explores the extent to which people can be sure today that they will have enough o f what i s considered valuable in the future, i.e. it studies how vulnerable they are. Finally, it investigates how people fare relative to their neighbors, i.e. how equally or unequally assets and well-being are distributed. Following this picture o f the evolution o fmonetary andnon-monetary well-being, the report proceeds with an analysis o f their determinants. 'Average growth rates between 1992 and 2004; an estimate and a projection were used for 2003 and 2004 respectively. Easterly, 2002. 9 Sen, 1985. 2 6. To address these different perspectives and dimensions of people's well being as well as their determinants, the report draws on a rich informationbase collected during the 1990s by the GoE and Addis Ababa University in collaboration with other international institutions and universities. In addition to the national accounts at the macro level, the GoE conducted a series o f nationally representative household surveys. The 1995 and 1999 Household, Income, Consumption and Expenditure Survey (HICES) and the accompanying Welfare Monitoring Survey (WMS) provide the key information base used in this study to examine consumption poverty and its determinants as well as the evolution o f non-monetary indicators. However, it will be shown that caution i s warranted in drawing too strong conclusions regarding the evolution o f poverty based on only two observations intime, given the rain dependence o f the Ethiopian economy. Other important sources o f information includethe 2000 Demographic and Health Survey and the 1999 National Labor Force Survey. The GoE has recently also launched another round o f the HICES and WMS surveys and is conducting a national participatory poverty assessment. The information derived from these surveyswill unfortunately not be available intime for inclusion into this report. 7. The insights from these nationally representative surveys are complemented with information from the Ethiopian Rural Household Surveys (ERHS), a panel o f 1,500 rural households in 15 villages conducted by Addis Ababa University along with Oxford University and the International Food Policy Research Institute (IFPRI), its collaborating institutions. The villages were purposively selected to capture the agro-ecological diversity o f Ethiopia, Households were interviewed 5 times between 1994 and 1999 and a new round has just been completed, though its results are not yet available. In parallel, another group at Addis Ababa University in collaboration with Goteborg University conducted the Ethiopian Urban Household Surveys (EUHS), tracking a sample o f 1,500 urban households from the seven major urban centers o f Ethiopia during 1995-1999, with a new round being completed in2004. The findings from boththese data sets provide important additional insightsinto the evolution o f rural and urban poverty and its determinants in Ethiopia. These survey data are further augmented with secondary information on agro-ecological and demographic conditions (rainfall, population density, and soil erosion) as well as insights obtained from qualitative and participatory case studies providing key contextual information in the dynamics o f social interaction, empowerment and the evolution o f poverty. 8. The report begins by exploring the evolution o f well-being in Ethiopia in Part I.It will zoom in on the evolution o f monetary well-being in its different dimensions (poverty, inequality and vulnerability) (Chapter 1) followed by a discussion o f the levels and evolution o f people's capabilities, i.e. their educational attainments, health and empowerment status (Chapter 2). Part Iconcludes with a summary o f the emergingprofile o f people's well being in Ethiopia and a discussion o f continuing knowledge gaps and data needs (Chapter 3). An overview and assessment o f the current poverty monitoring and evaluation system is included in Appendix 1. Part I1 explores the determinants o f monetary well-being. Taking a livelihoods approach, it sketches the endowment base o f Ethiopia's people and the risk factors they face, and using econometric analysis it subsequently analyzes how the returns to these endowments and risk factors differ across endowments, time and space (Chapter 4). Given that Agricultural Development-Led Industrialization (ADLI) forms the comer-stone o f the government's current poverty reduction strategy, and given that the large majority o f the poor 3 live in rural areas and that almost the entire rural population is primarily employed in agriculture, Chapter 5 subsequently sketches the performance o f the agricultural sector and empirically analyzes the relative importance o f the different determinants o f agricultural income. Chapter 6 discusses the continuing strategic importance o f raising the performance of the agricultural sector (both food and non-food) as well as the need for scaling up foreign aid to reach the MillenniumDevelopment Goal (MDG) related to poverty inEthiopia. It also presents micro-econometric simulations to shed light on the optimal sectoral composition o f public investment (education, infrastructure, health, agriculture) to reach the poverty MDG. Part I11 explores the determinants o f non-monetary well-being, in particular the relative importance o f different factors in reducing child malnutrition and child mortality and the respective role o f demand and supply side factors in increasing male and female primary enrollment rates (Chapters 8-10), Chapter 11 concludes by identifying key policy implications to enhance nutritional, health and educational outcomes, and explores altemative ways to enhance people's agency and empowerment. 4 Part I:How Well has the EthiopianPopulation Fared? This report takes a multi-dimensional approach and addresses the broad overarching question o f how the Ethiopian population has fared over the past decade and a half both from a utilitariadmonetary and a capabilityhon-monetary perspective. Chapter 1 begins by examining the evolution o f well-being, using monetary indicators such as income and consumption--the most common approach. We build on the 2002 poverty profile o f Ethiopia preparedby MoFED, and report new results on levels and changes in consumption poverty and inequality for the years 1995/96 and 1999/2000 along socio-economic, spatial and temporal dimensions. We also briefly review how vulnerable people are to becoming poor in the future. Chapter 2 addresses the same overarching question from a capabilityhon-monetary perspective and reports on the evolution o f human capabilities (education, health, nutrition), followed by a discussion o f people's ability to make effective choices, i.e. their level o f empowerment. In contrast to the discussion o f the monetary indicators and people's human capabilities, which heavily draws on quantitative statistics, the evidence base for gauging people's status o f empowerment i s more qualitative in nature. Chapter 3 summarizes the main findings and concludes by identifying continuing knowledge gaps in our understanding of people's well-being in Ethiopia and the corresponding data needs. An overview and assessment o f the current poverty monitoring and evaluation system i s provided in Appendix 1. CHAPTER 1. MONETARY DIMENSIONS OF WELL BEING-A UTILITARIANPERSPECTIVE 1.1 The chapter first sketches the evolution o f monetary poverty and inequality since 1992 as suggested by both the macro and micro data, and interprets the emerging picture within the evolution o f the broader economic and social context during that period. This i s followed by a brief characterization o f poverty in Ethiopia. The chapter concludes by discussing people's prospects for becoming poor inthe future, i.e. their vulnerability. 1.1 The Evolutionof Povertyand Inequalityin EthiopiaSince 1992 1.2 Tracing trends in household consumption and poverty over time is harder in practicethan is suggestedby theory. While the difficulties surrounding the measurement of household consumption and consumption poverty are well known," these difficulties are compounded when one tries to compare household consumption over time. First, this task requires full comparability o f the sample and questionnaire design, which i s often not the case, especially when repeated cross sections are used. Second, the longer intervals between surveysallow sufficient time for households to adjust their preferences renderingcomparisons across time more challenging. Third, there i s the thorny issue of which deflators to use in loIssues include: the valuation o f certain goods and services such as housing, durables and the market value of own-grown foods for which markets (and thus prices) are often incomplete or even totally missing; the need for geographical price deflators to make expenditures comparable across space; and measurement error related to recallperiods (Deaton and Grosh, 2002; Sahn and Stifel, 2003). 5 order to make expenditures between periods (and areas) comparable. Few countries have reliable and regularly updated consumer price indexes that account for spatial differences in prices. 1.3 This poverty assessment encounters all of these problems. The questionnaire design for non-food expenditures changed somewhat between the national HICES survey rounds, which are four years apart. Further, in the period between rounds important changes have taken place in the weather, the external environment (e.g. a border war), and prices (e.g. a collapse o f coffee prices), all o f which may have led people to adapt their consumption bundles. Finally, and most importantly, we find that there are several deflators, which steer the conclusions in widely divergent directions. All these problems are further exacerbated when comparing consumption andpoverty levels across countries (see Box 1.1). 1.4 T o address these challenges, we draw on different data sources and triangulate the evidence emerging from each of these sources. We begin by linking the sectoral components o f overall GDP growth with household consumption from surveys, based on the household head's sector o f employment, to explore how poverty evolved over a longer time period. While limited in its own way, this methodology enables one to track poverty over time and allows a more fine-grained tracking o f poverty than i s possible usingthe aggregate growth rate. Second, we compare consumption and poverty across the two HICES surveys conducted in 1995 and 1999. The HICES surveys allow presentation o f a nationally representative picture based on micro-data, and therein lies their major strength. Yet, as will be further illustrated, it i s difficult to draw firm conclusions about longer term trends based on two data points when the economy experiences large annual fluctuations in its overall value. Thus, and third, we complement the evidence from the national surveys with those from purposively sampled rural and urban panel data which are also more consistent in survey and questionnaire design over time, as well as with qualitative case studies and other contextual evidence. Finally, given that large improvements and deteriorations in welfare should eventually be reflected in asset holdings, we also examine the evolution o f key assets over time. Taken together, the evidence suggests that consumption poverty during the 1990s has largely stagnated, with indications that urban poverty and inequality have increased while rural poverty may have slightly declined. 6 Box 1.1: How poor is Ethiopia really? Different measures, different results? H o w poor i s Ethiopia when compared to other countries? Alternative monetary measures and methodologies often seem to yield different answers. We explore this paradox and compare living standards in Ethiopia to those in other countries using monetary measures o f average income and monetary measures o f poverty, which are sensitive to the lower end o f the distribution, and find that the results are broadly internally consistent. Usingboth sets o f statistics Ethiopia emerges as an equal, but equally poor society. Some o f the differences emerging from comparing different measures arise from the technical solution adopted to express monetary variables in a common scale. Conversion o f the monetary variables in the same currency (usually U S dollars) using market exchange rates offers an intuitive answer. Market exchange rates reflect, however, a host of factors which might not be relevant for the comparison o f welfare across countries, such as speculative movements and government interventions on the exchange rate. These factors can alter the exchange rate, without really affecting how people's purchasing power varies from country to country. Purchasing Power Parity (PPP) adjusted exchange rates form a useful alternative. PPPs are exchange rates that take into account the cost and affordability o f a basket o f common items in different countries, usually expressed in the form o f U S dollars. The first two columns in table 1 below illustrate how the use o f different conversion methodologies may lead to seemingly widely different conclusions when evaluating people's welfare ina country. Table B1.1.1: Average monetary resources inEthiopia inUS dollars GDPicapita GDPicapita Private consumptiodCapita ConsumptiodCapil (2003 US$) (2003 PPP US$) (national accounts, 1999)* a (1999 Survey) Level 97 716 80 126 Ranking out o f 153 148 154 countries Average exchange rate for I999 was 7.942 Source: World Bank, 200413, and FDRE,2001 (1999 HICES), own calculations The large increase in income per capita shown by adopting PPPs rather than market exchange rates i s not a peculiarity o f Ethiopia. When in country A prices are lower than in country B, values expressed inPPP terms will be higher. In general, there is a broad association between price levels and income o f a country, so that inpoorer countries PPP values tend to be higher than in dollar terms. Price differences often follow from differences in wages, which tend to be much higher in richer countries. Yet despite the large increase in GDP per capita which may suggest that Ethiopia i s not so poor after all, its relative position as one o f the poorest countries in the world remains largely unchanged when moving to PPP exchange rates. Rankings are indeed much less affected than levels by the use o f a dollar or a PPP indicator-the rank correlation between ordering countries either on the basis o f GDP in PPP or GDP inU S dollars is 0.97. Inthe last columns o f Table 1, we consider only those resources which are available for private consumption, and contrast the data o f the national accounts with those o f the HICES household survey to gauge internal consistency. While consumption per capita according to the national account i s almost 40 percent less than what i s observed in the household surveys, at 126 US dollars per capita the bottomline that Ethiopia is among the poorest countries in the world remains. Discrepancies between national accounts and household surveys in estimating private consumption are very common. In the former, private consumption i s obtained as a residual, while household surveys use self-reported consumption to measure consumption directly. Factors driving the discrepancies include problems with respondents' recall and effective sample coverage in the case o f household surveys, difficulties in accounting properly for informal and subsistence activities inthe national accounts, differences inthe definition o f "private sector" between household survey and national accounts, and different treatment o f income in kind, immted rents and financial services (Ravallion. 2003). 7 Turning to distributionally-sensitive measures o f welfare, we compare Ethiopia's 1 and 2 dollar a day poverty incidence numbers with those o f other countries, and explore how these poverty numbers compare with the poverty incidence derived from locally estimated poverty lines, again to gauge internal consistency. International comparability in this case i s ensured using comparable household surveys and by adopting an internationally comparable poverty line (Le. a poverty line whose level can be considered meaningful across different countries) expressed inPPP dollars. The choice o f an intemationally comparable poverty line has received much attention. 1 dollar a day (or more exactly US$ 1.08 a day) in 1993 PPPs has been identified as a good approximation to the national poverty lines, given consumption levels adopted by low income countries. It is, however, a conservative estimate, and an alternative poverty line, set to be twice as much as the first one (US$ 2 a day), has been adopted as well to provide a less stringent standard. To obtain comparable poverty estimates based on these international poverty lines, the intemational line is translated into local currency in 1993 (by using the PPP for 1993) and the local consumer price index is then used to take into account the difference in local prices between 1993 and the present. The results for Ethiopia are inTable 2. Table B1.1.2: Poverty incidence inEthiopia 1US$/day Ranking across 2 US$/day Ranlung across National poverty ~- _ _ ~ _ _ _ J P E _ L- 96 countries_ _ ~ L 9 6 c 0 u " t r i e s _ _ _______ line 1999 Poverty 26.3 73 80.7 84 44.2 headcount (%) Source: World Bank, 2004g, based on HICES, 1999 The table above shows that estimates o f poverty based on the international poverty lines differ from estimates based on the national poverty line. Ths i s not surprising as in Ethiopia the national poverty line adopted by the government is equivalent to 1.50 dollars in 1993 PPP (even though this corresponds to only 0.31 actual 1999 US$) (Kakwani, 2004). This implies that the estimated incidence o f poverty based on the national poverty line will lie between the incidence obtained with the 1 and 2 dollar a day lines and the poverty numbers thus appear internally consistent. Its exact position depends on the distribution o f consumption. These considerations hrther underline the importance o f using different poverty lines as well as the need to take into account the spread o f the distribution. For example, Senegal, despite having a GDP per capita almost 6 times hgher than Ethiopia in 2003, had exactly the same incidence o f US$ 1 a day (in PPP terms) poverty, 26.3 percent according to the latest available estimates, though only 68 percent o f the population were found to be below US$ 2 a day as opposed to 81 percent in Ethiopia. This results from higher overall inequality in Senegal, with the Gini coefficient being 0.41 inSenegal compared to 0.29 inEthiopia. Indeed, wealth is rather equally distributedinEthiopia comparedto other countries in the world with only 16 countries out o f 126 with a lower Gini coefficient than Ethiopia. This is also reflected inthe different ranking Ethiopia takes depending on the US$ 1 and 2 a day poverty lines. While it ranks 731dout o f 96 countries when using the US$ 1a day poverty line, it ranks 84" when using the U S $ 2 a day poverty line. Together the different monetary indicators suggest that Ethiopia is equal, but equally poor. Ethiopia's ranking inthe Human Development Index (170 out o f 177 countries) and the Human Poverty Index (92 out o f 95 countries), which also include non-monetary indicators o f poverty, confirm this picture. ~~ ~~ Sources: World Bank 2004g; International Comparison Project (www.worldbank.otddata/icp), Ravallion,2003; Kakwani, 2004 1.1.1. Macro evidence 1.5 T o explore poverty trends over the past 15 years, the historical information on sectoral GDP growth rates from the national accounts i s linked with information on the sector o f employment o f the household and household consumption from the 1995 HICES. More specifically, it i s assumed that GDP and private consumption grow at similar rates. The population's employment share in each sector i s derived from the 1995 HICES based on the sectoral employment of the household head or other male adults incases where the household head was inactive. Inaddition, within-sector growth i s assumed to be distribution neutral. 8 1.6 The evolution o f headcount poverty for Ethiopia from 1989 through 2004 obtained by applying this macro-micro simulation method to the 1995 HICES i s presented in Figure 1.1. The fact that the 1999 head count poverty rate generated through this simulation i s similar to the one obtained by direct calculation from the 1999 HICES data (see Section 1.1.2, Table 1.2 using the lower poverty line) provides confidence in the empirical validity o f our methodology.l1When examining the evolution o f head count poverty portrayed inFigure 1.1, there are three important observations to note. Box 1.2: Basic assumptions underpinningthe growth-povertyscenarios A number o f steps and assumptions underpin the growth-poverty scenarios presented in this report. First, the poverty reducing effect o f the different growth scenarios i s obtained by applying observed (sectoral) GDP growth and population rates to the income distribution observed in the 1995 HICES. Household income was approximated by total expenditure per adult equivalent with imputed rent, excluding energy expenditures given the intricacies involved in imputing expenditures related to firewood and dung collection, a common fuel source inthe rural areas. Inusing GDP growth rates to predict the evolution o f household consumption, we assume that private consumption and GDP grow at similar rates. While the large amounts o f food aid Ethiopia receives could arguably affect the levels o f private consumption as well as smooth out its fluctuations over time, this should not affect longer t e r m average growth rates, unless food aid reception grows at a substantially different pace than GDP. Closer inspection o f the national accounts indicates that nominal private consumption and GDP have historically followed similar growth paths.I2 This i s corroborated by rural household surveys where consumption is found to closely track the movements in agricultural income.I3 Finally, for 2003, we only considered a decline inagricultural GDP o f 7 percent inour simulations, corresponding to the reported decline in private consumption, as opposed to the reported 12 percent decline in agricultural GDP. This way we account for the massive influx o f food aid in 2003, in response to the huge slump in agricultural income in 2003 following the widespread drought and the substantial drop in commercial fertilizer use compared to 2002. The exceptional influx o f food aid has partly offset the drop in agricultural income and has helped the country steer clear o f a disastrous famine. Second, we use the lower full national poverty line discussed in Table 1.2. Third, the overall population in 1995 in Ethiopia i s estimated at 54 million. By operating the actual population and sectoral growth rates on the 1995 income distribution, we estimate that 24.3 million people or 36.2 percent o f the Ethiopian population were poor in 2004. Fourth, the population's employment share in each sector i s derived from the 1995 HICES based on the sectoral employment o f the household head. In cases where the household head was inactive, we looked at the employment sector o f the other male adults in the household. In 1995, 85 percent of the Ethiopian population was employed in agriculture, one percent in industry and 14 percent in the service sector. According to the 1999 HICES, 84 percent o f the household heads were employed in agriculture, six percent in industry and 10 percent in the service sector. Clearly, Ethiopia i s still overwhelmingly an agricultural society. 1.7 The first important observationis that poverty declinedonly marginallybetween 1990 and 2004. Despite 1.7 percent per capita overall growth (Table 1.1), the simulation method suggests that poverty incidence declined only marginally between 1990 (the starting point for the poverty MDG) and 2004 (from 38.4 to 36.2 percent). To shed further light on this surprising and seemingly contradictory result, Table 1.1 also presents a brief review of the sectoral growth rates (both in overall and per capita terms) from 1992 to 2004. Overall and sectoral growth rates during 1995-1999, the period spanned by the HICES surveys, are also included and reviewed for reference when we discuss the micro evidence below. To I' For a more elaborate discussion o f the assumptions and the validity of this simulation method, see Box 1.2. l2 Xiao Ye and Alan Gelb (2004) also indicate that real private consumption and GDP growth have started to diverge since 1995. They attribute this to a divergent evolution o f the CPI index and the GDP deflator. They further indicate that reported CPI evolution since 1995 is highlyunlikely compared to other existing evidence o f the evolution o f food and non-food market prices inEthiopia. l3 Dercon, 2004. 9 obtain sectoral per capita growth rates, we divide agricultural GDP growth by rural population growth and industrial and service GDP growth by urbanpopulation growth. About 96 percent of the rural population was employed inagriculture. Figure 1.1: Evolution of poverty incidence between 1989 and 2004 EPRDF assumed The basis for MDG power 1995 HlCESiWMS survey 1999 HICES/WS I I 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Year Source: Own calculations 1.8 Economic growth per capita since 1992 has been completely driven by growth in the service sector, with virtually no (or even slightly negative) annual growth per capita in the agricultural sector (Table 1.1 and Figure 1.2). Giventhat 85 percent o fthe population lives in rural areas and that virtually everybody in rural areas derives their livelihood directly from agriculture or agriculture related activities, it should not come as a surprise that poverty stagnated or declined only marginally over the past 15 years. Based on these findings, we hypothesize and confirm below using micro-data that there has been no or only limited poverty reduction inrural areas. Nonetheless, this still leaves the question why service sector growth has not translated into more (urban) poverty reduction. Table 1.1: Per capita GDP growth in key economic sectors, 1992-2004 Sectors 1992-2004 1995-1999 Overall GDP per capitagrowth (annual %) 1.73 2.53 Real annual agricultural growth (%) 2.04 2.35 Real annual industrial growth (%) 5.13 4.07 Real annual services growth (YO) 6.80 7.83 Real annual growth (YO)inpublic administration, 10.2 11.3 education, health and defense Real annual agricultural growth per capita(%) -0.25 0.05 Real annual industrial growth per capita(%) 0.45 -0.66 Real annual services growth per capita(%) 2.11 3.11 Source: World Bank, 2002b; an estimate and a projection were usedfor 2003 and 2004 respectively 1.9 Analysis of service sector growth indicates that this was largely determined by rapid government expansion (Table 1.1 and Figure 1.2). As a result, the government's share o f GDP increased from about 12 percent in 1992 to around 20 percent in2004. Service sector expansion was especially rapid during the 1995-1999 period, which can be largely attributed to the military build-up for the border war with Eritrea. In 1998 and 1999, the 10 government sector (as derived from the national account^)'^ expanded by 25 and 16 percent respectively compared to about 4.8 and 6.6 percent in 1995 and 1996. Nominal military expenditures (as per reported government outlays) rose fivefold and eightfold between 1996 and 1998 and 1999. Economic growth fueled by military expansion i s however unlikely to have reducedpoverty. We will explore this further with the micro data which cover the 1995- 1999 period. Figure 1.2: Evolutionof overall, sectoral and government GDP growth rates, 1992-2004 25.0 Gov't expenditures 20.0 $s h v 15.0 8- i$ 10.0 9 E 4 *E 5.0 Ed 0.0 -5.0 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 Source: Own calculations 1.10 The composition of government expenditures has shifted substantially from defense to poverty related sectors since 2000 (agriculture, road construction, education and health) with spending increasing most rapidly in the education sector (Table 1.2, Section 1.1.3, Figure 1.3). This increase inpublic investment inpoverty related sectors holds promise for future poverty reduction. While this i s consistent with the predicted decline in poverty since 1999 in Figure 1.1 above, it may take some time before the effects o f this enhanced public investment will be fully felt in terms o f poverty reduction. This holds especially for the expansion o f the education sector, but also to some extent for the rehabilitation and expansion o f the roadnetwork." ~~ l4Expendituresmentionedhere exclude expenditures on educationandhealth. IsFor a more detailed analysis of the evolution of the composition of the Ethiopian economy, government expenditures,and domestic demand, see World Bank, 2004c. 11 Figure 1.3: Evolution of sectoral shares in total government expenditures, 1992-2004 40.0 .......................................................................................... 35.0 4- 5 20.0......................... B 30 15.0 ..................... rr I 8 i 8 5.0 (II Rcad Heahh 0.0J I 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 Source: Own calculations 1.11 Harvest failure in 2003 likely undermined positive effects of poverty-focused public investment. However, the anticipated poverty reducing effects o f the increase in poverty-focused public expenditures have likely been substantially undermined by the massive harvest failure in 2003, which threatened about 14 million people with famine and most likely pushed many people back into poverty, as suggested by the sharp increase in poverty incidence in 2003. Analysis o f the upcoming 2004 HICES (not included in this report) will help get a more accurate estimate o f the net effect on poverty o f these opposing forces (enhanced poverty focused public investment versus droughts). These events also illustrate the broader point that while analysis o f the national accounts coupled with household survey data is very informative in generating hypotheses about the broad trends in poverty incidence and its underlyingcauses, micro data are needed to get a more detailed and accurate picture o f the evolution across locations and livelihoods. Micro data on consumption will also be better able to capture income from the informal economy. We will thus use the available micro-data to further gauge the empirical validity o f the emerging picture of limited decline inpoverty incidence over the past 15 years, as well as to shedlight on the evolution of poverty among the rural and urban population and different livelihoods in light of the observed trends in government spending and the performance o f the different sectors o f the economy. Before doing so, we make two more observations regarding the evolution of poverty incidence inEthiopia. 1.12 The second observationregardingthe evolution of poverty incidence in Ethiopia i s that poverty incidence i s highly volatile. As can be seen, for example, from Figure 1.1 above, poverty incidence declined sharply between 1994 and 1997, but bouncedback strongly 12 between 1997 and 1999. While food aid, which i s not taken into account in these figures, would presumably smooth the fluctuations a little bit, it i s safe to say that high volatility o f poverty incidence characterizes the evolution o f poverty throughout the 1990s. Delays in procurement and distribution often prevent food aid from acting as an effective insurance mechanism. The volatility in poverty incidence follows from the high dependence o f the Ethiopian economy on rain-fed agriculture and the inability o f the Ethiopian population to insure themselves adequately against these shocks (see hrther Chapter 4). The bi-variate correlations between contemporary and lagged average annual rainfall and poverty are estimated at -0.37 and -0.30 respectively.16 1.13 Third, and closely relatedto the second observation,the conclusion regardingthe direction and rate of poverty reduction differs significantly depending on which two years are compared. For example, a comparison between 1992 and 1997 shows a high rate o f poverty reduction, while a comparison o f 1995 and 1999 suggests stagnation. Similarly, when comparing 1994 with 1997 it could be concluded that poverty has substantially declined, while a 1994-2000 comparison suggests a much smaller decline. This calls for caution inthe interpretation o fpoverty trends derived from a small number o f data points over time, and the development of better methodologies for monitoring poverty between survey rounds. It also underscores the need to take a longer term perspective when analyzing trends of poverty inEthiopia. 1.1.2. Micro evidence 1.14 Inthis section, we examine ifthe broadpicture of limiteddecline inpoverty incidence during the past 15 years emerging from the macro data is confirmed in the micro data. The micro data allow us to obtain more accurate and more disaggregated poverty numbers by location and livelihoods, though care mustbe taken ininterpreting trends emerging from these data given that they are collected much less frequently, and in light o f high volatility in the Ethiopian economy. We begin with evidence from the two nationally representative household consumption surveys, followed by evidence from purposively sampled surveys. Evidencefrom the HICES surveys 1.15 Table 1.2 reports three sets o f poverty ratios. The first (estimates from MoFED) reproduces the results from the Government's own estimates o f the three most common measures o f poverty-the head count (PO),the poverty gap (PI) and the poverty gap squared (P$, a measure o f the severity o f poverty. The next four columns report estimates o f poverty which assume a common basket across time for all households and are based on a lower and an upper poverty line. This way, we test the robustness o f the results to the choice o f the poverty line. The last four columns report the same measures but allow for variation in consumption baskets across reporting levels and across time for the same reporting level. Put differently, the findings in the last columns allow for the possibility that households or individuals living in different locales consume different baskets o f goods and services at different prices. The differences between our methods and those o f the government are ~~~ l6A correlation coefficient o f one indicates that both variables move perfectly in parallel, while a correlation coefficient o f zero indicates that both variables change independent from each other 13 discussed in more detail in Box 1.2. The construction of the poverty line is discussed in Box 1.3. Table 1.2: Evolution of poverty between 1995 and 1999 Estimates by Poverty estimates usingone poverty line Poverty estimates using local poverty lines MoFED (No substitution) Lower Poverty Lines Upper poverty lines Lower poverty lines Upper poverty lines - PO 0.47 0.45 0.40 0.38 0.59 0.58 0.37 0.29 0.54 0.39 LZ P1 0.13 0.12 0.10 0.10 0.19 0.18 0.10 0.07 0.16 0.10 P2 0.05 0.05 0.04 0.04 0.08 0.08 0.04 0.02 0.06 0.04 PO 0.33 0.37 0.31 0.37 0.46 0.53 0.32 0.46 0.47 0.70 0.1 0.1 0.09 0.10 0.15 0.17 0.09 0.14 0.16 0.29 -2 '9 P1 P2 0.04 0.04 0.04 0.04 0.07 0.08 0.04 0.06 0.07 0.15 0.46 0.44 0.38 0.38 0.57 0.57 0.36 0.32 0.53 0.43 .% PO 2 P1 0.13 0.12 0.10 0.10 0.18 0.18 0.10 0.08 0.16 0.13 P2 0.05 0.05 0.04 0.04 0.08 0.08 0.04 0.03 0.06 0.05 Source: Ministry of Finance and Economic Development, 2002, and own calculationsfrom HICES data Box 1.3: Methodologies used to measure the extent of poverty The poverty numbers obtained by the government and this report differ in three ways: 1) the underlying expenditure variables; 2) the poverty lines; and 3) the deflators used." First, our approach uses expenditures with imputedrents as opposed to rents estimated during the interview by the enumerators, and excludes energy variables whose measures were left to the discretion o f the enumerators and therefore subject to a lot o f noise. Second, while the Government report uses one poverty line, based on a national average consumption bundle, we report results using a lower and an upper poverty line. While in essence similar to the government's approach, we convert each households' consumption figures into Addis Ababa prices and calculate their poverty status by relating these converted consumption figures to the 1995 Addis Ababa urban poverty line. We also explore how the use o f region- and time-specific poverty lines as opposed to one national poverty line may affect the poverty estimates (last four columns in Table 1.2). These are obtained through direct comparison o f the expenditures in each reporting levevregion to the reporting level specific lower and upper poverty line, which has been separately calculated for each survey year. In doing so, we account for differences in preferences across regions (corrected for income differences) as well as potential substitution by households in response to changes in relative prices over time Third, we use different deflators than the government. Following standard practice, we first present results which do not allow for substitution in consumption over time inresponse to relative price changes (column 4-8, Table 1.2). No deflators are needed for our second set o f results which are based on location and time specific poverty lines. For geographic price deflators, we adjust expenditures by the ratio o f food poverty lines in each reporting level relative to the Addis Ababa urban food poverty line (for details on the calculation o f the food poverty line, see Box 1.4). Specifically, we scale both 1995 and 1999 expenditures by the same spatial deflators (ratio o f each reporting level o f food poverty in 1995 to Addis Ababa food poverty in 1995), and then adjust the temporal price difference using the Addis Ababa food consumer price index calculated by the Central Statistical Authority (CSA) o f Ethiopia. So only location specific food poverty lines are used to calculate geographic price deflators. Incontrast, the government report uses food and non-food prices to compute geographic deflators and uses national average food and non-food baskets. Although it i s technically correct to include the price o f food and non-food in the index, we find that using food CPI rather than the General CPI makes little difference l iThis report uses the same adult equivalent scales as inMOFED, 2002. 14 because during the period under consideration (1995-1999), there is very little difference between these two measures (see Appendix 3, Table A.1.1). However, by using location specific food poverty lines as deflators, we also control for differences in the dietary patterns across regions. This could be important as some regions largely derive their calories from cereals while others are more enset dependent. 1.16 Followingstandardmethods to measure poverty, there appearsto havebeen little reductionin poverty between 1995 and 1999. The estimated poverty rates reported in the middle four columns o fTable 1.2 are consistent with those reportedbythe government. Ifthe lower poverty line is taken as the poverty line, then roughly 38 percent o f Ethiopians could not meet their basic needs in 1995. By 1999, the same proportion would have been considered poor. The government's estimates put the national poverty head count at 46 percent in 1995 and slightly less, 44 percent, in 1999. The severity o f poverty and its depth as measured by P1 and P2, also held steady throughout. At the upper poverty line, about 57 percent o f Ethiopians would have been counted as poor inboth 1995 and 1999. Insum, there was hardly any change inpoverty rates at the national level. 1.17 Further decomposition suggests a slight decline in rural poverty. Using the lower poverty line, about 40 percent o f the rural population was estimated to be poor in 1995, but I poverty incidence declined to 38 percent by 1999. Similarly, a small reduction in poverty, from 59 percent in 1995 to 58 percent in 1999, i s observed when upper poverty lines are used. These results are again consistent with the pattem in rural poverty reported by the government. 1.18 But urban poverty increased. Among urban residents, poverty incidence was estimated at 31percent in 1995 and rose to 37 percent in 1999. When usingthe upper poverty line even larger increases inpoverty rates are observed, from 46 percent in 1995 to 53 percent in 1999. The substantial increase in urban poverty seems largely driven by a substantial increase in poverty in Addis Ababa. Those counted as poor in Addis Ababa rose from 50 percent to 57 percent, and this increase accounts for 71 percent o f the increase in urban poverty. The government numbers suggest an increase in urban poverty incidence o f 4 percentage points. 1.19 Allowing for poverty lines to vary by locale, the observed tendencies are intensified. When we allow households to substitute their food consumption bundles in response to price changes, rural poverty substantially declines and urban poverty starkly increases, resulting in an overall decline o f the poverty head count by 4 to 10 percentage points depending upon the poverty line used. While the findings from the conventional approach are clearly more consistent with those obtained from the macro-evidence, it would be useful to explore this further with the upcoming 2004 HICES. We refer to Appleton (2003) for a more detailed discussion o f the different normative assumptions underpinningthe use o f one national vis-a-vis location- and time-specific poverty lines. 15 Box 1.4: Methodological considerations in obtaining poverty lines The cost of basic needs approach: The most common method for obtaining poverty lines starts from the assumption that every individual has basic caloric needs to survive and perform daily activities (to work, to think, etc.). On the basis o f this assumption, the first goal becomes identifying a minimumbundle o f goods that will satisfy this basic need. Since there are numerous possible bundles, the simplest strategy is to pick the consumption bundle o f a group o f individuals, typically ranked in the lower part o f the consumption distribution (say in the 30" or 40" percentile), and treat this as containing the minimumbundle o f goods that would be required to meet the basic needs o f individuals. Once the basket is identified, the next step i s to cost it by calculating the minimum expenditure that would buy the minimum calories needed. Specifically, the idea is to estimate the price o f a calorie (dividing expenditures by calories) and multiply this by the required calories per day (inthis case, we used 2,200 calories per day per adult). Computing what it costs to buy such a bundleestablishes the food poverty line. However, the poor also have non-food basic needs. They need minimumstandards o f clothing, shoes, etc. in order to participate in social activities, and to maintain their dignity. So after establishing the basic food expenditures needed, the next step is to obtain basic non-food spending. This i s done by calculating the average non-food spending o f the individuals identified as poor. Adding the food poverty line and the non- food basic needs spending establishes the full poverty line. The standard practice i s to calculate one food poverty line and one full poverty line, each at the national level. The most recent analysis o f poverty conditions in Ethiopia by the Government o f Ethiopia (MoFED, 2002) does this, as does the last poverty report by the World Bank (1999). In both o f these studies, a consumption level o f 1,075 Birr in 1995 prices is established as the poverty line. Multiple poverty lines and substitution effects: Like the analyses mentioned above, this report uses the cost o f basic needs approach. However, there are some differences between the preceding studies and what i s done here. First, rather than compute one national food and one national full poverty line, we compute multiple food and full poverty lines, one for each o f the 32 groups o f zones that the Ethiopia Central Statistical Authority uses for sampling o f households. We do this for 1995 and again for 1999, so that altogether there are 64 food and full poverty lines. There are several reasons for this choice. One is that defining a common national basket is neither accurate nor desirable because individuals living in different parts o f Ethiopia may have different preferences and tastes. Another i s that a common and nationally defined basket o f goods does not allow easy introduction o f new goods and household substitution o f goods that are confined to a single or a few geographic spaces. Indefining multiple baskets and letting the contents o f these baskets change, we allow for differences inpreferences, and substitution effects inresponse to relative price changes over time. A major objection to such an approach derives from the fact that preferences may differ by income level. Consequently, if government spending across space is sensitive to poverty incidence, it may in effect remunerate "expensive" tastes. T o control for differences in dietary patterns, and allowing for substitutions in response to relative price changes while still controlling for income effects, we follow a practical way suggested by Ravallion (2003).'* Inparticular, to obtain the food poverty lines, we first choose a group o f individuals who are around the 45" percentile o f the national consumption distribution. We choose the 45" percentile because the Government's report estimated that about 45 percent o f people are oor S ecifically, we take individuals within a 20 percent band around this percentile, or those within the 35t to 55 r .ttpercentiles. From this group o f individuals, we obtain the prices they paid and the quantities they purchased, as reported in the survey data. We then group individuals in the 35" - 55" percentiles into geographic reporting levels, and use a reporting level-specific basket o f food to calculate location-speiific food poverty lines. To obtain full poverty lines, we use non-parametric methods discussed in Ravallion (2003). Rather than one full poverty line, we calculate both a lower and an upper bound to check that the poverty rates and the evolution o f poverty are robust to the choice o f the poverty line. T o obtain the lower poverty line we first calculate the mean per adult equivalent non-food expenditures for individuals in a reporting level whose per See Appleton (2003) for an alternative way o f obtaining income adjusted regional poverty lines as well as a discussion o f the normative assumptions underlying the use o f one versus multiple poverty lines. 16 adult equivalent total expenditures lie within a small interval (say plus or minus 10 percent) o f the per adult equivalent food poverty line for that reporting level. The mean o f the non-food expenditures is obtained by taking the mean of the means obtained through stepwise expansion o f the interval around the food poverty line (e.g. first one percent around the food poverty line, then two percent, etc.) This mean per adult equivalent non- food component, when added to the food poverty line, establishes the lower full poverty line. For the upper poverty line, we follow a similar procedure but use the average per adult equivalent non-food expenditures for individuals whose food expenditures lie within a small interval o f the per adult equivalent food poverty line for each reporting level (for more details see Appendix 3, Table A.1.2). Note that the individuals included inthe sample to calculate the food poverty line are not necessarily the same as those used to calculate the full poverty lines. In the first case, the individuals are drawn from the group defined as the poor-in this case, those inthe 35" -55" percentile o fthe consumption distribution. But when calculating the full poverty lines, individuals are drawn from allthose inthe survey from that reportinglevel. Once the reporting level-specific poverty lines are calculated, one can report poverty rates and the associated severity and depth o f poverty measures by comparing reported consumption per adult equivalent and the reporting level-specific poverty lines, properly weighted. In this report we present the results using this approach (Poverty rates using local poverty lines, cols 9-12, Table 1.2). However, we also report results from the more common approach (Poverty rates using one poverty line, cols. 4-8, Table 1.2. 1.20 To examine the evolution of inequality over time, we present the evolution of growth in consumption across the different expenditure deciles in Figure 1.4 (see Appendix 3, Table A.1.3 for details). The deciles provide a picture, albeit discrete, o f the growth patternacross the entire distribution of consumption. The 1999 consumption averages have been expressed in 1995 prices. Rural consumption is slightly above one across all deciles except for the top decile suggesting a slight improvement in rural living standards across the rural population. Inurban areas on the other hand, we observe a decline in living standards across all deciles, except for the poorest decile which becomes slightly better off, and the richest decile whose living standards increase by about seven percent, suggesting an increase in inequality. The national pattem reflects the rural pattern with the lowest seven deciles slightly better off andthe richestthree deciles slightly worse. 17 Figure 1.4: Consumptiongrowth ratiosbetween1995 and 1999per consumptiondecile 1.10 1.05 S .-.--..~'~ v) C 8 09 5 . _- ~.~~~ ................................... ................ ...............~. ....... ~-:-. ..........~~ ~ ~~..... ~~. ... ~~~..~. .-5 C * *. 0.go __ .....:. 93 ...... ............~..........~....... :.1.* ...... ~..................... :. ..................... ~.~.~..~. .. :. ~ ~ ..----.__......-- ~ . ~ ~~ .. ~ ~ 0.85 . . .. .... . ...... .............. .......~.~~~~ . ........... ............ ........................ ~ ~...... ~ . ~~~~ 0.80 - I , ,, 1 1 I 1 1 1-4- - 1 National consumption growth -- Rural consumption growth 1 . . I .._.Urban consumption growth Source: Own calculationsfrom HICES data 1.21 The suggested picture of an increase in urban inequality and slight decline in rural and nationalinequalityis mirroredby the estimatedGini coefficients(Table 1.3).19 The Gini coefficients paint Ethiopia as an equal, but equally poor society consistent with people's self image that holds that "everybody i s equally poor."20 It should be noted, however, that inequality in urban areas i s larger than in rural areas and increasing, possibly fuelled by the influx of unemployed migrants and the emergence o f a small upper class. Table 1.3: Evolutionof InequalityinEthiopia Gini coefficient 1995 1999 Rural 0.284 0.272 Urban 0.350 0.365 National 0.300 0.287 Source: Own calculationsfrom HICES data Evidencefrom purposively sampled urban and rural panel data 1.22 Evidence from the EUHS (urban panel data) suggests a slight increasein urban poverty between 1995 and 2000 and a starker increase in inequality (Table 1.4). The panel survey o f 1,500 urban households from seven major cities and towns in Ethiopia has been conducted by the Department o f Economics, Addis Ababa University, jointly with l 9Intuitively, the Gini index (commonly referred to as "Gini coefficient") of apopulation represents the expected income difference betweentwo randomly selected individuals or households. 2o Rahmato andKidanu, 1999. 18 Goteburg University. The survey started in 1994, and the same households were revisited in subsequent surveys (1995, 1997, 1999, and most recently in 2004). The cities and towns included are Addis Ababa, Awassa, Bahir Dar, Dessie, Dire Dawa, Jima and Mekele. The results regarding the evolution o f poverty and inequality from the urban panel data are broadly consistent with those o f the national surveys based on the more standard calculation methods. Poverty increased slightly in urban areas between 1995 and 2000, with signs o f a continuing increase ininequality over time (Gini increased from 43 in 1995 to 48 in2000. Table 1.4: EUHSpaneldata evidence on the evolution of welfare in urbanEthiopia, 1994-2000 Year Share o f food exp in Realper capita consumption per Headcount poverty total exp month (Birr) (Po) (%) Gini 1994 0.85 103 39 44 1995 0.71 109 37 43 1997 0.62 125 34 47 2000* 0.67* 116* 38* 48* *preliminary results Source: Shimeles, 2004 1.23 The ERHS i s a panel o f 1,500 rural household surveys undertaken in 15 villages by Addis Ababa University, jointly with the Centre for the Study of African Economies at Oxford University and IFPRI. The villages have been purposively sampled from a north- south axis of the country in an attempt to include the diversity o f the agro-ecological zones, farming systems, and socio-economic characteristics. The villages were first surveyed in 1994 (two rounds), revisited in 1995, 1997, 1999, and again in 2004 (data not yet available). 1.24 Evidence from ERHS suggests a drop in rural poverty of six percentage points (Table 1.5). While the decline in rural poverty noted by this survey i s consistent with the decline found in the national surveys using standard methods, at six percent it i s substantially larger. Yet Table 1.5 also shows that real monthly consumption, measured in 1994 prices, increased by 27 percent between 1994 to 1997 and by 11 percent between 1994 and 1999, which translates to about 2.1 percent per year. This i s significantly more than the aggregate per capita agricultural growth and the reported private consumption growth calculated from the nationallyrepresentative household surveys usedfor this report. Table 1.5: ERHS Dane1data evidence on the evolutionof welfare in ruralEthioDia. 1994-1999 Year Real per adult equivalent consumption Head count poverty (Po)(%)*I per month (Birr) Method 1 Method 2 199415 90 53 30 1997'' 115 43 23 1999 100 47 25 In 1997, three villages were surveyed in a different season not appropriate for comparison with the other data. The latest suitable observation for 1995 is used. However, altemative assumptions do not affect these findings. Only households with complete information for all three periods have been included. In total, 1,377 households (from a possible sample of 1,477) have been retained. This implies an overall attrition rate of less than seven percent, which is low for this type of survey. 2,The poverty numbers based on method 1 have been obtained by using similar methodologies as in the urban panel to facilitate comparison. The poverty figures based on method 2 are comparable with those reported in Table 1.11. Source: Dercon, 2004 19 1.25 Given the care with which the ERHS data have been collected and compiled, and given the consistency in the survey and questionnaire design over time, it is unlikely that these differences result from measurement error. Further comparison o f their asset base shows that farmers in the ERHS cultivated on average 1.6 to 1.35 ha in 1994 and 1999 respectively,21 compared to the national average o f 0.99 and 0.95 ha per holder. Also, the reported rainfall in 1994 and 1997 exceeds the national average by about 100 to 130 mm respectively.22Better overall rainfall conditions combined with more landholdings might have resulted in larger consumption growth. Yet the position o f the ERHS households within the national distribution needs to be investigatedfurther. Inclusion of some o f the same questions as inthe WMS inthe 2004 ERHS will allow ERHS researchers to do so. 1.26 Other evidence from the Northeastern highlands (Amhara Region) suggests a slide towards destitution. The study surveyed 2,160 households between November 2001 and April 2002, and examined the evolution o f their well-being through retrospective questions.23 Based on respondents' self assessments o f the evolution o f their well-being over time, the study finds that the percentage o f respondents who considered themselves destitute had increased from 5.5 percent in 1992 to 16.4 percent in2000, falling back to 14.6 percent in 2002, following two good rainfall years in most o f the surveyed communities. At the same time, the percentage of households considering themselves to be "doing well" dropped from 32 percent inthe early 1990s to just three percent at the beginning o f 2000. Most notably, the group o f vulnerable increased from 17 percent 10 years ago to 55 percent in 2002, indicating that 2.5 million people out o f 4 million are at serious risk o f becoming destitute in the near future. These trends, based on self-assessments were largely confirmed by the community level perceptions o f trends in well-being and matched evidence from the more quantitative data collected inthe study. Evolution of assets 1.27 There is no sign of asset accumulationin rural areas, which might have signaled a population experiencing rapid or measurably significant poverty reduction. These assets can take the form o f consumer durables (radios, TVs, etc.) as well as cattle and livestock in general. In rural Ethiopia in particular, it would seem reasonable to expect significant livestock accumulation because o f their consumptive and hedgingvalue. However, there i s no evidence o f growth in livestock holdings during the period spanned by the HICES surveys (Table 1.6). As a matter o f fact, the number o f holders with livestock slightly declined (see Chapter 4, Table 4.4). A look at the changes inthe distribution o f ownership o f assets such as TVs or radios does not show a significant improvement either, though the percentage o f holders with a radio (a small consumer durable) increased. 1.28 Some improvementsin the possessionof consumer durableswere observed in the urban areas. Especially radio, but also TV ownership increased. We also observed some increase inthe ownership o f bicycles and cars. As ownership o f the larger consumer durables 2 'Van den Broeck, 2004; Gabriel and Demeke, 2003. 22 Average rainfall in the ERHS villages reported at 1,102 and 1,181 mm in 1994 and 1997 respectively, comparedto 1,001 and 1,05 1mmnationwide. 23 Sharp, Devereux, and Amare, 2003. 20 (car, TV, bicycle) i s still concentrated among a small minority and associated with larger consumption levels (see Chapter 4, Table 4.9), this evolution i s consistent with the observed increase inurban inequality. Table 1.6: Ownershipof livestockand consumer durables inEthiopia between 1995 and 1999.'' Average Rural Urban 1996 1998 2000 1996 1998 2000 Livestock holdings per holder" # cattle for agricultural purposesioxen 1.15 1.08 cattle for other purposes 2.85 2.88 sheep and goat 2.85 2.36 horses, mules, asses, camels 0.55 0.58 Consumer durables Radio 7.7 8.3 11.2 47.8 56.5 60.0 TV 0.1 0.0 0.0 6.8 8.8 13.7 Bicycle 0.5 0.4 0.4 2.1 2.2 4.0 Car (for private or commercial use) 0.0 0.2 3.O 3.2 ')Figures do not include pastoralist households not cultivating ; figures for livestock actually refer to 1995 and 1999. Source: CSA, 2001, and own calculationsfrom 1995 and 1999Agricultural Sample Surveys, CSA 1.1.3 Emergingpictureof poverty and inequalitysince 1992 1.29 The macro and micro evidence paint a picture of limited to no decline in overall poverty incidencein Ethiopia since 1992. Urbanpoverty seems to be edging upwards while rural poverty remained largely constant with signs o f a potentially limited decline (one to two percentage point). While overall inequality inEthiopia remains low and does not appear to be increasing, inequality within urban areas i s on the rise. Moreover, the overall economy in general, and the Ethiopian population in particular, remain highly vulnerable to rainfall shocks. We provide a more detailed account o f the latter dimension o f people's well-being based on micro data inSection 1.3. Several factors have contributedto this distressing picture o f limitedto no poverty decline inEthiopia. 1.30 First, while overall growth per capita has been positive, agricultural growth per capita during the period has been either negative or zero, consistent with the limited poverty decline observed in rural areas. With 85 percent o f the population living in rural areas and virtually everyone in rural areas earning their livelihoods in agriculture or agriculture related activities, absence o f a larger reduction in national poverty, and rural poverty inparticular, does not come as a surprise. Overall economic growth has been largely drivenby growth inthe service sector. 1.31 Second, the rather strong average annual growth in the service sector (6.8 percent during 1992-2004 and 7.8 percent during 1995-1999) begs the question of why hasn't urban poverty declined further. Rapid urban population growth due to rural-urban migration significantly eroded the benefits from overall growth inthe service sector. Average urban population growth over the past decade and a half i s estimated at about 4.7 percent. As a result, service sector (or urban income) growth inper capita terms i s substantially less (2.1 and 3.1 percent during 1992-2004 and 1995-1999 respectively). According to the 2000 21 Demographic Health Survey (DHS) total fertility rate (TFR) of Addis Ababa (which accounts for about 30 percent o f Ethiopia's urbanpopulation) is currently estimatedat 1.9 childrenper woman and is therefore slightly below replacement level, whilst the TFR for urban Ethiopia as a whole i s only 3.3.24 This suggests that rural to urban migration must be important to explainurbanpopulationgrowth of 4.7 percents2'Furthermore,labor force participationrates in urban areas are observedto be low and declining, suggestingdiscouragedworkers and a growingnumberof urbanresidentsjoining the masses ofthe poor.26 1.32 Third, not only was economic growth in urban areas substantially eroded by urban population growth, the quality and spatial allocation of growth in the service sector over the past decade was not always conducive to urban poverty reduction. Growth in the service sector has largely been driven by Government spending - a boost in defense expendituresduringthe second half of the 1990sand a rapid expansionof spendingin poverty related sectors (especially education) since 2000 (see Table 1,7). While growth in defense spendingis not conducive to rapid povertyreductionin general, it i s also unlikely to leadto rapid urbanpovertyreduction,which helps explainswhy despite3.1 percentper capita growthinthe service sectorbetween 1995 and 1999,urbanpovertydid not decrease. 1.33 Fourth, the more recentincrease in Government spending on poverty sectors had a deliberateruralbias and will affect incomepovertywith a lag (Table 1,7), Inparticular, the Government's focus has beenon increasingthe reach o f agriculturalextension, and access 24For comparison, the national TFR is 5.9 and national population growth is estimated at 2.5 to 3.0 percent dependingon the source. 25Nonetheless, the internal consistency of the available statistics regarding rural and urban population growth and fertility will need to be further investigated. According to the 1999 Labor Force Survey (CSA, 2000), a person is considered a non-migrant if s h e has been continuously residing in the same town or if slhe has been continuously residing in the rural part of the same woreda where he was born. Migrants are hrther divided in long term and recent migrants. Recent migrants are migrants who migrated during the last five years prior to the survey date. While 19.6 percent of the population (Le. 10.7 million people) was considered a migrant in 1999, only 4.3 percent of the total population was estimated to be a recent migrant (or 2.3 million people). Closer investigation of the recent migration patterns show that most recent migration has happened within rural areas (38 percent). It was fiuther found that about 24 percent of all recent migrants (about 0.5 million people) moved from rural to urban areas, while 16percent movedback from urbanto rural areas. This impliesthat net urbanin- migration from rural areas i s only eight percent (24 -16 percent), or only 176,000 people (=2.2 million*0.08) between 1994 and 1999. In 1999, the urban population has been estimated at 7.35 million (and the country's total population at 54.4 million). Given an urban population growth rate of 4.7 percent, this would imply that urban population expanded by 1.5 million people and that only 12 percent (176,00011.5 million) of this expansion was due to net rural in-migration. This seems at odds with the estimated total fertility rate of 3.3 which has been survey based, unless the urbanization rate i s actually muchlower. It i s also possible that urban to rural migration has been overestimated. According to the 1994 population census it was only seven as opposed to 16 percent. In sum, these discrepanciesunderscore the need for an urgent update of the population census andmore in-depthinvestigation o f migrationpatterns in Ethiopia. 26Urban migrants tend to be male, between 20 and 49 years old and mostly work in services and commerce if they are low-skilled, or remain unemployed during the first years of their arrival in urban centers if they are better educated (Admassi, Guta and Ayalew, 2003). As evidence of queuing for better public jobs, workers with diplomas have unemployment rates comparable to those with general education, and unemployment duration tend to be long-a mean of 4 years. Nearly 75 percent of the urban employed hold low skill jobs. In 2002, urban unemployment was estimated to be around 25 percent (Commander, 2004). A more detailed profile of urban labor markets and the position of migrants will be provided inthe upcoming Labor Market ESW prepared by a World Bank Team. 22 to primary education and health, in rural areas. By contrast, urban areas (also known as municipalities) were expected to provide these services from their own revenues, except for rare cases like the city o f Addis Ababa, where investments were insufficient to keep up with the growing population. Therefore, it can be argued that inthe absence o f agricultural growth per capita, any observed poverty reduction in rural areas probably resulted from better access to services and infrastructure, and the slight increase in female adult education. (see Chapter 4, Table 4.2). The increase in urban poverty, together with the observed deterioration in access to public services (health, public transport, drinking water) in urban areas (Chapter 4, Table 4.2), i s consistent with the rural bias in public investment. The latter finding i s also consistent with increasing pressure on public services observed in urban areas due to rural- urban in-migration. Moreover, it must be emphasized that some investments will take time before the effects are felt in terms o f poverty reduction-especially investments related to education, but also investments related to road construction. Insum, the more recent increase in public spending on poverty sectors was unlikely to have heavily affected urban poverty because many o f the additional investments were geared towards rural areas and several o f these investments take time to pay off interms o f income poverty reduction. Nonetheless, the increased spending on poverty sectors holds hope for the future (both for rural and urban poverty reduction) as indicated by the micro-simulations in Chapter 6 about the expected poverty reduction from these investments. Table 1.7: Index of real government expenditure and share in percent of GDP 1999100 2000101 2001102 2002l03 2003104 2004105 Actual pre act pre act preact Con.Budget ProvBudget Index of Expenditure Total GovernmentExpenditure 100.0 93.5 103.5 112.8 116.2 124.5 Defence 100.0 48.9 38.6 32.6 40.0 38.2 Debtservicing 100.0 89.9 83.6 95.6 105.5 112.5 Povertysectors includingFoodSecurity 100.0 135.3 159.5 170.6 229.2 257.3 Povertysectors ExcludingFoodSecurity 100.0 135.3 159.5 170.5 210.2 221.2 Agriculture &natural resource 100.0 123.2 151.3 141.4 155.8 187.3 Roads 100.0 137.4 166.8 183.1 248.2 195.1 Education 100.0 133.0 152.6 193.4 219.6 263.0 Health 100.0 168.1 184.5 152.2 242.9 239.1 Relief relatedaidfoodaid 100.0 91.2 93.8 254.0 73.6 72.4 Other sectors 100.0 126.1 165.2 157.2 154.0 181.5 InpercentofGDP Total Government Expenditure 32.6 29.8 34.0 35.9 32.0 31.9 Defence 12.8 6.1 5.0 4.1 4.3 3.8 Debtservicing 2.3 2.0 1.9 2.1 2.0 2.0 Povertysectors includingFood Security 9.1 12.0 14.5 14.8 17.5 18.3 Povertysectors ExcludingFoodSecurity 9.1 12.0 14.5 14.8 16.0 15.7 Agriculture&natural (water) resource 2.7 3.3 4.I 3.9 3.6 4.0 Roads 2.2 3.0 3.6 3.5 4.6 3.3 Education 3.1 4.0 4.7 5.7 5.7 6.4 Health 1.1 1.8 2.0 1.8 2.2 2.0 Food security 0.0 0.0 0.0 0.0 1.4 2.6 Reliefrelatedaidfood aid 2.0 1.8 1.9 5.1 1.3 1.2 Other sectors 6.4 7.9 10.6 9.8 8.3 9.1 Source: World Bank, 2004d 23 1.34 Fifth, the benefits from economic growth which reached the urban centers were likely concentrated among a limited number of citizens. Indeed, only a small fraction o f the population inurban areas may have benefitedfrom growth inurban areas, which has been largely driven by growth in the Government sector, as only 20 percent o f all urbanemployees above the age o f 10 years work in the public sector.27 And this i s under the extreme and unlikely assumption that growth was uniformly distributed among public employees. This observation i s consistent with the reported rise inurban inequality. 1.2 The Face of Hunger and Deprivation 1.35 The overall picture o f limited decline in poverty masks a wide variety o f different experiences. Inthis section we disaggregate the poverty numbers further, and briefly review some key defining characteristics o f the poorer sections o f the population. We look in particular at how poverty incidence differs across educational attainment o f the household, its livelihood systemand its location. 1.2.1 Educationand poverty 1.36 Despite an overall lack of movement in poverty reduction, there is a surprising diversity among the Ethiopian population, with the less educated having a particularly high likelihood of being poor. Table 1.8 presents the poverty rates among different socio- economic groups in 1999 using the lower poverty lines.28 Clearly, the more educated a household head, the less likely his household i s in poverty. While education i s generally observed to be important in developing countries, the probabilities o f being poor among households with educated versus uneducated households are quite stark. We will return to this inChapter 4. Table 1.8: Incidence of povertyby educationin 1999 Socio-economic characteristics Poverty Ratio Incomplete Primary education 0.32 Completed Secondary 0.21 Completed postsecondary 0.09 Source: Own calculationsfrom I999 HICES. 1.2.2 Livelihoodsand poverty 1.37 Poverty incidenceamong households employed in non-agricultureis substantially lower. Poverty incidence also differs widely across people's livelihoods. As indicated in Table 1.9 households employed in the non-agricultural sector tend to be less poor than those inagriculture. The policy implications ofthis findingwill bediscussed at lengthinChapter 6. 27CSA, 2003. Using the 1995 data or the upper poverty lines yields the same insights. 24 Table 1.9: Consumptionand povertyincidenceby sector of employmentof household head in 1995 and 1999 Sector of 1995 1999 employment of Population Consumption Poverty Population Consumption Poverty householdhead per adult incidence per adult share (%) e uivalent (%I share (%) equivalent incidence (%) Agriculture 85 1592 40 84 1600 38 Industry 1 1980 32 6 1707 43 Services 14 2201 28 10 2113 35 Source: Own calculationsfrom 1995 and 1999HICES. 1.38 Within agriculturalhouseholds,cash crop producersare better off. Eveninrural areas families undertake a variety o f activities. Some grow grains, some keep livestock and other grow crops for export or cash. Table 1.10 illustrates the incidence o f poverty among these various livelihood strategies, in particular the agriculturalists. Overall, cash crop producers have lower rates o f poverty than food crop producers. Further decomposition shows that among the cash crop producers, chat producers were the least poor, though their poverty rates increased substantially between the two periods. Poverty among tea producers however declined. The poverty rates among the coffee producers were even slightly higher than among the rest of the population (Table 1.10). Table 1.10: Ethiopia, incidenceof poverty by livelihoodzg Type of livelihood 1995 1999 Mainly agriculture 0.41 0.38 Mainly cash crop producers 0.29 0.26 Coffee producers 0.42 0.40 Chatproducers 0.19 0.33 Teaproducers 0.41 0.24 Source: Own calculationsfrom 1995 and 1999HICES 1.39 While not representative o f coffee/chat growing households at large-for example, the sample excludes any sites from Keffa and only one site from Hararghe, both places o f high value coffee production-the five year ERHS panel offers a snapshot o f the fortunes o f people engaged in such livelihoods. Two-thirds o f the households inthe ERHS sample grow neither coffee nor chat. About 21 percent grew coffee (of which 80 percent are from three villages: one in Sidamo, one in Kembata and one in Hadiya), seven percent grew chat (mostly in a village in Hararghe) and six percent grew coffee and chat (mostly in Gurage). The results show that coffee growers in this sample are relatively poorer than non-coffee growers, whether they have coffee alone or with chat, consistent with the nationally representative household surveys. Similarly, chat growers appear better off than other farmers.30 Table 1.11 also shows that non-coffee/chat growers have managed to increase mean consumption levels and reduce poverty between 1994 and 1997, but that after this the decline in poverty stagnated. However, for many o f the chat and coffee farmers any gains between 1994 and 1997 appear to have been wiped out. 29Uses lower poverty line. Using upperpoverty line gives same results. 30See Dercon, 2004 for details. 25 Table 1.11: Evolution of poverty among coffee and chat growers, 1995-1999 Real consumption per adult Poverty incidence (%) equivalent per month (Birr) 1994-5 1997 1999 1994-5 1997 1999 Neither coffee or chat (66%) 97 123 110 27 20 18 Coffee only (2 1%) 70 89 67 44 35 44 Chat only (7%) 104 136 115 12 10 20 Chat and coffee (6%) 65 84 80 40 35 38 Group Total 90 115 100 30 23 25 Source: Ethiopian Rural HouseholdSurvey 1.40 While coffee prices did indeed decline substantially between 1995 and 1999, they were also at an historicalpeak in 1995, while prices in 1999 were around their historical average. This may explain why we did not observe an increase in poverty among coffee growers in the national survey. Nonetheless, the high poverty incidence among coffee producers in Ethiopia i s somewhat puzzling. Could it be related to the size o f their landholdings? While coffee production may be a more productive activity than cereal production, landholdings among coffee growers are on average only 66 percent as large as those among non-coffee producers, and more than 30 percent have less than 0.07 ha per person. The reasons behind the large poverty incidence among coffee growers deserves further investigation, especially in light of the collapse in coffee prices over the past few years. The price paid to producers o f coffee dropped from about 70 US ct. per pound in 1999 to an historical low o f about 30 US ct. per pound in2002 (Figure 1S). About 25 to 30 percent o f rural households grow coffee, although some o f them only for home consumption. Figure 1.5: Producer price of Arabica coffee (US ct. per lb) 140.00i 120.00 100.00 80.00 60.00 40.00 20.00 I 0.002 , Source: International Coffee Organization 26 1.2.3 The geography of poverty 1.41 Geographically, poverty is widespread. The majority o f the poor live in the four large regions (Tigray, Amhara, Oromiya and SNNP) and Addis Ababa. Together, these regions account for 85 percent o f the population o f the country. As Table 1.12 shows, the highest rates of poverty among the major regions are found in Tigray and SNNP. These regions are characterized by lower than average arable land per capita, underscoring the role o f land scarcity in determining poverty. But poverty rates are also high in the city o f Dire Dawa, and inthe peripheral, sparselypopulated regions of Benishangul-Gumuz and Gambela. By 1999, only Amhara had experienced a substantial decline in poverty. Information about average expenditures and food expenditure shares by region i s reported inTable A.1.4. Table 1.12: Poverty incidenceinEthiopia by administrative region1995-1999 Region Lower poverty rates Upperpoverty rates 1995 1999 1995 1999 Tigray 0.45 0.49 0.66 0.69 Afar 0.20 0.43 0.26 0.63 Amhara 0.45 0.36 0.65 0.55 Oromiya 0.28 0.32 0.46 0.52 Somali 0.08 0.15 0.18 0.33 Benishangul-Gumuz 0.49 0.54 0.72 0.71 SNNP 0.49 0.48 0.67 0.65 Gambela 0.35 0.66 0.48 0.79 Harari 0.25 0.29 0.43 0.47 Addis Ababa 0.34 0.41 0.50 0.57 DireDawa 0.47 0.49 0.65 0.68 Source: Own calculationsfrom 1995 and 1999HICES. 1.2.4 Pastoralists 1.42 Pastoralists tend to be poorer and more vulnerable. There i s little statistical information on pastoralists, but it i s commonly held that pastoralists tend to be poorer and more food insecure than their highland, humidand sub humid, counterparts, to be less literate, to enjoy less access to public infrastructure and services and to depend disproportionately on food aid.3' Despite the lack o f representative statistical information, the main characteristics o f their livelihoods are well documented in a variety o f qualitative studies. Pastoralists face a myriad o f risks, with drought and losses o f herdbeingpervasive concerns. Lack o f access to economic infrastructure compounds the challenges pastoralists face when dealing with the consequences o f these shocks: markets for livestock are little developed so that markets are fragmented and prices are very variable in space and time. The weakness o f livestock markets also implies that pastoralists lack a rapid way o f replenishing their stocks after a crisis. Violence linked to land access issues and as a stock replenishment strategy i s quite common. 3'Smith, Barrett and Box, 2001, p. 3. 27 1.43 Pastoralists are increasingly entrapped in poverty.. Pressures from population . growth, the increasing share o f land which becomes inaccessible due to either the establishment o f private property rights or agriculture, and more generally institutional limits to herd mobility, are further putting pressures on land. They lead to increased livestock density, land deterioration, and thus increasedrisks. These factors can combine to entrap poor pastoral households: impoverished households whose herdsize becomes too small to support their livelihoods following a shock, move closer to towns. 1.44 ...andare movingto towns. Towns offer food aid distribution centers, easier access to services andpotentially more income earnings opportunities, though these seem to be more easily available to those who have savings that can be used as starting capital. The concentration o f livestock inperi-urban areas increases the likelihood that the landwill not be able to support the animal stock, resulting therefore in further impoverishment. Impoverished pastoralist household end up being increasingly dependent on external assistance, generally provided by food aid. Informal insurance systems no longer seem to offer appropriate help.32 1.3 Risk,Vulnerability andPoverty 1.45 Vulnerability to povertyis high. While there is some consensus on how to measure poverty and inequality, the academic community has only recently started to develop direct and comprehensive measures o f householdvulnerability. The merits and weaknesses o f these different methodologies are still debated and empirical applications are scarce.33 One view considers vulnerability as the probability o f becoming poor inthe future (irrespective o f one's welfare now). An empirical application o f this methodology to cohorts o f the Ethiopia 1995 and 1999 HICES shows that 61 percent o f the Ethiopian population have a larger than 50 percent chance o f falling below the poverty line in the future. This means that, holding current endowments constant, about two out o f three Ethiopians will be poor at least five out o f the coming 10 years.34 1.46 Average poverty numbers hide a substantial amount of moving in and out of poverty. Inthe absence o f direct empirical estimates, household vulnerability is often also gauged by the extent o f movement inand out o f poverty over time. The substantial amount o f churning reported in the HICES cohort study i s also found in the ERHS (Table 1.13). Although a substantial number (22 percent) moved out o f poverty between 1994 and 1999, 16 percent fell into poverty inthis period.35 Table 1.13: Poverty dynamics inERHS between 1994 and 1999 Cell% Non-Poor 1999 Poor 1999 Total Non-Poor 1994 53 16 70 Poor 1994 22 9 30 Total 75 25 100 Source: Dercon, 2004 32Lentz and Barrett, 2004. 33See Christiaensen and Subbarao (2005) for a review of the different methodologies. 34World Bank, 2004e. 35 I t is unlikely that these movements are largely driven by measurement error. See Dercon and Krishnan (2000a) for a discussion. 28 1.47 Ethiopians face multiple risks. To further gauge people's vulnerability, we also explore their risk environment and the extent to which people can cope with it. Risks arise from many sources-natural (droughts, floods), economic (prices), institutional (land tenure insecurity), and political (civil conflict and war). Most importantly, Ethiopian producers have to deal with highly volatile rainfall and frequent droughts. As an example, between 1978 and 1998 alone, there were 15 droughts (and famines) that have led to the displacement, injury or death o f more than one million people.36 Moreover, the burden o f low average rainfall i s often exacerbated by highvolatility inthe rainfall patterns as illustrated inFigure 1.6.37 Figure 1.6: Long run average rainfall and rainfall variation across woredas in Ethiopia , ' 2000 0.4 6 1 - 1500 0.35 1 9 z 1000 500 0.1 I 0 0 S I Zones ~ I Long Run Average Rainfall (mm) +Coefficient of Variation Source: Own calculations 1.48 Drought shocks often lead to severe harvest failure and have long-lastingeffects. Careful analysis o f the evolution o f poverty between 1989 and 1995 among households in six villages in Ethiopia indicated that while poverty overall declined by 29 percent, it would have declined by about 39 percent in the absence o f rainfall shocks. Indeed, about 78 percent o f the households inthe ERHS reported harvest failure, most often caused by rainfall failure, as the most common type of hard~hip.~'Other ways in which drought risks are experienced include loss of livestock (including oxen), and as food and water insecurity. Empirical evidence further suggests that in addition to the immediate negative impacts o f shocks on consumption, the detrimental effects can be long lasting. Again, among the households inthe ERHS data, it has been shown that 10 percent lower rainfall 4-5 years earlier led to a reduction in current growth rates o f one percent. Even more strikingly, the impact o f the 1984/85 drought was found to affect consumption growth throughout the 1990s. Households at the 75thpercentile o f consumption loss duringthe 1984/5 famine ex erienced on average 16 percentage points less growth during the 1990s than those at the 25tr ~ e r c e n t i l e . ~Clearly, ~ not only are households' livelihoods exposed to drought shocks, they are often unable to cope with it ex post, rendering them very vulnerable. We will present further evidence on the pervasive negative effects o f shocks on well-being inthe remainder o f the report (Chapters 4, 8 and 9). 36World Bank, 2000. 37The observed correlationbetween averagerainfall and the coefficient ofvariation is -0.49. 38Dercon, 2002. 39Dercon, 2004. 29 1.49 Intra-annual price volatility can be high. Figure 1.7 shows nominal and deflated cereal prices at the Addis Ababa wholesale market. The main observation i s that cereal prices, which constitute the main staple o f the population and particularly o f the poor, display wide intra-annualvariations. For instance, injust a few months, July 1998 to October 1999, practically all three major cereal prices first increased by over 60 percent (July 1998-July 1999) and then declined by about the same amount. Maize prices are even more volatile (Figure 1.8). Such large swings inprices, especially cereal prices, create large uncertainties in the market, inducerisk-aversion and therefore loss ofproduction andpoverty reduction. Figure 1.7: Nominal and deflated cereal prices - Addis Ababa wholesale 350 0 300.0 250.0 f 2000 '3 a 1500 100 0 50 0 1800 160 0 140 0 i 1200 1000 'gc E 800 eo 0 40 0 20 0 30 1.50 Malaria presents another pervasive risk. About 75 percent o f the 60 million Ethiopians in 2000 were estimated to live in malaria infected woredas (World Bank, Roll Back Malaria, 2003). Figure 1.9 displays the distribution o f malaria risk across regions in Ethiopia. All the populations in Benishangul/Gumuz and in Gambela regions are considered to be at risk since all live inmalarious woredas. This i s followed by Somali and Afar, both o f which have very highproportions (90 percent) o f their populations at risk o f malaria (see also Appendix 2, Figures A.l.1-A.1.2). Malaria i s third only to acute respiratory illness and perinatal disease in terms o f its contribution to the disease burden in the country. Over 10 percent o f the total discounted years o f life lost due to premature death i s from malaria.40 We retumto the estimated effect o fmalaria and illness shocks more broadly on consumption and poverty inChapter 4. Figure 1.9: Shareof populationat risk of malariainEthiopia 1007 40 i O J BenshagGumuz Gambella Somali Afar Amhara Oromiya SNNPR Tigray regions 1.51 Ethiopia is home to one and a half million people living with AIDS.41 The HIV/AIDS infection rate has risen from a low base in the mid-1980s to 4.4 percent in 2003 with recent signs that the increase may be leveling off, especially in urban areas. The HIV/AIDS incidence among adults was estimated at 12.6 percent in urban areas in 2003 compared to 13.4 percent in 1995. It i s estimated at 2.6 percent in rural areas in 2003 compared to 0.8 percent in 1995. There are an estimated 539,000 AIDS orphans. These orphans are a particularly vulnerable group. Only 41 percent o f maternal AIDS orphans living in foster families find feeding conditions with their foster families satisfactory, which is significantly lower than the percentage o f orphans not infected with HIV/AIDS. The percent 40See Tables 1 and 11, in Geresu, 1996. 41FDRE,2004. 31 o f AIDS orphans engaged in income-generating activities increased from nine percent to 23 percent after the mother's death.42 1.52 The geographical and demographic characteristics of the epidemic bode illfor its impact on the economy and poverty. Urban rates of infection are twice those inrural areas, with 15.6 percent prevalence in Addis Ababa. Based on data collected by the Ministry o f Health, about 91 percent o f infections occur among adults between 15 and 49 years old, the most economically productive segment of the population. With (relatively) well-educated urban adults hit hardest by the disease, compounded by the probable negative effects on orphans' education, the impact on the country's stock o f human capital threatens to be severe. As it stands, AIDS imposes, and will continue to impose, a burden on the limited capacity o f Ethiopia's health system, as well as on the surviving families to care for the sick and the orphaned. While awareness about the disease has increased and some change inbehavior has beenobserved over the past couple o f years, HIV/AIDS continues to pose a significant threat to the govemment's development efforts and continued and concerted actions as well as strong leadership will be needed to prevent the HIV/AIDS epidemic from turning into a disaster. 42 Bhargava, 2005. 32 CHAPTER 2. NON-MONETARY DIMENSIONS OF WELL-BEING-A CAPABILITY PERSPECTIVE 2.1 In this chapter we look at poverty from a capability perspective, and examine how people have fared along dimensions which could be considered intrinsically (as opposed to instrumentally) valuable. In particular, we gauge overall progress in people's educational attainments, their nutritional and health status, i.e. their human capabilities (Section 2.l), and explore how able people are inmaking effective choices to shape their lives, i.e. their level of empowerment (Section 2.2). In contrast to lackluster advancement in terms o f monetary measures o f well being and poverty, Ethiopia has made significantly more progress in improving human capabilities, with rapid improvements in education and a slower but positive evolution inhealth and nutrition. Yet the full effect of recent government reform and actions to empower citizens as a means o f poverty reduction have yet to be fully felt by citizens, and the empowerment level o f Ethiopian citizens, especially o f women and pastoralists, remains low. 2.1 The Evolutionof Human Capabilities 2.2 Starting from a low base, Ethiopia's enrolment expansion at all levels of education since 1994 has been impressive. Primary enrolment (grades 1-4) has gone up from less than two million children in 1991 to about six million in 2001-02, almost tripling the number o f students in the first level o f primary school. Inthe same period, enrolment in the second level o f primary (grades 5-8) rose from under one million to over 2.5 million (see Figure 2.1). As a result, gross enrolment in grades 1-8 has increased from 24 percent in 1994 to 62 percent in 2001-02. Similarly, while secondary and higher education start from even lower bases than primary education, their rate o f expansion has also been phenomenal: secondary education rose by 40 percent between 1994 and 2001-02 while university enrolment more than doubled. Figure 2.1: Enrollmentsingrades 1-12, Ethiopia, 1967-2002 7.0 6 0 -'S2 -.- 2 5 0 4 0 23 8 3 0 0 eb 2 0 z E I O Source: World Bank, 2004a 33 2.3 Substantial gender and regional differences in education remain. Clearly, Ethiopia has accomplished much in educational attainment inthe last decade. This is, by far, the country's most impressive and far-reaching achievement in non-monetary measures o f well-being. Key remainingareas for improvement include the large gender gap in enrolment, which while having narrowed, remains substantial. The ratio o f girls' over boys' gross primary school enrollment is now about 0.8, though only still 0.67 for secondary education (grades 9-12) and merely 0.21 for higher education. Note furthermore that the relatively favorable position o f girls today i s the culmination o f progress over more than 30 years with girls in 1967/8 accounting for just 30 percent o fthe pupils ingrades 1-4 to reach 42 percent in 2001/2. The evidence indicates that the main source o f differences in enrolment between boys and girls i s at entry, since once in, the attrition (drop-out) rates between boys and girls are the same.43The large regional differences in educational attainment must also be bridged. Afar and Somali, the two pastoralist regions, bothhave enrolments not exceeding 15 percent. 2.4 Unlike education, progress in most of Ethiopia's health outcomes has been slower. For instance, life expectancy has shown no improvement between 1996 and 2000 (Table 2.1). Instead, it actually declined from 44 years to 42. Levels o f matemal mortality are estimated at 871 deaths per 100,000 live births, potentially a huge improvement since 199544but still among the highest inAfrica. This contributes directly to the observed low life expectancy. While the urban groups have better access to health care, the problem o f few professional birth-attendants persists everywhere in the country. Nearly three-quarters o f pregnantwomen receive no antenatal care. 2.5 Under-five mortality dropped from 216 per 1,000 live-births to about 169 per 1,000 live-births between 1984 and 1999/2000. Effectively, this means that Ethiopia has been successful at reducing the probability o f a child dying before hidher fifth birthday from 21.6 percent in 1984 to 16.6 percent in2000. While this i s a significant achievement, a child mortality rate o f 166 deaths per 1,000 live births i s still unacceptably high. There i s very little difference in outcomes between the very poor (24 percent child mortality) and the very rich (around 22 percent child mortality). Too many children are dyingneedlessly. Table 2.1: Selected health indicators for Ethiopia, 1984-1999/2000 Census Survey Health indicators 1984 1994 1995196 1999100 Overall life expectancy (years) 42a' 45`' 44 42 Total fertility rate (birthsper woman over reproductive lifespan) 7 6.4b' 6 5.9 Maternal deaths per 100,000 live births 1800d' 871 Chld mortalityrate (deaths per 1,000 live-births) 216 169 a) refers to 1982. b) refers to 1990. c) refers to 1992. d) the range of uncertainty for this estimate is particularly wide, from 790 to 3,200, and the reduction by 2000 shouldbe interpreted inthis context. Source: CSA and ORCMacro, 2001; World Bank, 2003a; World Bank 2004b 43World Bank, 2004a. 44Data quality leads us to question the actual size of this improvement 34 2.6 Accessibilityto health care services has improved, but remains poor. Only about 50 percent o f the population live less than seven kilometers away from the nearest health care centre, which i s a big problem considering that roads are not available everywhere and few are useable in all weather conditions, and that many o f the poor do not own transport animals or vehicles to get to these facilities. This i s particularly pronounced in pastoralist areas. According to a World Health Organization report, only about five percent o f the Afar population had access to proper health care, with 2 hospitals inthe entire region.45 2.7 Some progress has been made in reducing child malnutrition. The positive news for Ethiopia i s that the prevalence o f child stunting, measured as an abnormally low height- for-age score, an indicator o f cumulative (long term) poor nutritional status, has dropped below 60 percent from levels as high as 67 percent in 1997 (Figure 2.2). Consistent with other evidence from Sub Saharan Africa, prevalence o f stunting i s lower among boys than among girls46. Furthermore, the average rural child has a much higher chance o f being malnourished over the long run than the average urban child. However, measures o f child wasting as captured by low weight for height scores, a measure o f short runmalnutrition, did not improve (see Figure2.3). Figure 2.2: Evolutionof child stunting in Ethiopia 1983-00 %male stunted 75 T .................................................................... 1 70 65 60 I 55 % female stunted 50 4 ................................................................................................................................. 45 , , ~ I I I I 1983 1992 1996 1997 1998 2000 Year 45CitedinUNOCHA, 2002. 46Svedberg, 1990. 35 Figure 2.3: Evolution of child wasting in Ethiopia 1983-00 Child malnutrition remains among the highest in Sub-Saharan Africa, locking another generation into a nutrition-poverty trap. In 2000, the Ethiopian rates for pre- school child stunting were still 18 percentage points above the average o f 39 percent in Sub- Sahara Africa. Severe malnutrition, especially for children between six and 36 months old, lowers cognitive development and the ability to learn, which in turn decreases the long term accumulation o f human capital and lifetime eamings. Clearly, any strategy for reducing poverty must include measures to reduce child malnutrition as a core element. Ethiopia- specific evidence suggests that imparting maternal nutritional knowledge can be an effective andtimely complementary strategy to reduce malnutrition in addition to raising incomes and increasing female formal education.47 2.9 Progress has been made in providing access to safe drinkingwater, though little improvement has been observed concerning sanitary conditions. The population share with access to safe water (tap or protected well) increased from 21 percent to 27 percent between 1995 and 1999 (Table 2.2). Nonetheless, about one third o f the households continue to rely on open rivers and lakes as sources o f water. Sanitation conditions are worse. In 1995, about 84 percent o f the population disposed o f waste by throwing it away (Table 2.2). Some o f that waste was used as fertilizer, a dangerous practice interms o f spreading disease. The situationhas changed little since 1995. 47ChristiaensenandAlderman, 2004, 36 Table 2.2: Sources of drinking water and use of waste disposal facilities, 1995-1999 % PopulationUsing FacilityiMethod in: 1995 1999 Water Public or private tap 14.87 15.88 Protected well 5.74 11.03 Unprotectedwell 19.06 38.58 Unsafe River or Lake 46.39 33.85 Sanitation Disposalvehicles / containers 2.71 2.57 Dug-outs 3.63 3 Throw away 84.27 47.76 Use as Fertilizer 42.12 Bum 3.35 Other 9.35 1.1 Source: FDRE, 1997 (1995/96 WMS), and FDRE, 2001 (1999/2000 WMS) 2.10 It may be noted that these observed improvements, especially in education but also in health outcomes, seem at odds with the earlier reported limited decline in monetary poverty. In Chapter 4, we do indeed find substantial economic retums for example to education. Yet, most o f the reported improvements in human capabilities are rather recent and concern children (primary enrollment rates, child mortality, and child malnutrition). Thus, it will take some time before the benefits are felt in terms o f poverty reduction, In Chapter 6, we will simulate the poverty gains to be expected from these improvements inhumancapabilities. 2.2 The Status of Empowermentin Ethiopia 2.11 Good governance and empowerment are among the eight major thrusts of the GoE's Sustainable Development and Poverty Reduction Program (SDPRP),48reflecting the comprehensive and multi-dimensional view the government takes of people's well being-r lack thereof. In this view, evaluating people's well-being cannot be reduced to only reviewing their material welfare (as reflected by their income), or their human capabilities, but also includes exploring their status o f empowerment. To address this issue, this section draws largely on qualitative information, inlinewith the nature o fthe topic, but to the extent possible it will also use statistically representative quantitative data to strengthen the insightsobtained from analyzing the qualitative information base. Because data collection systems have traditionally been geared toward collecting information on monetary indicators of well-being and indicators o f human capabilities, the data base to analyze empowerment, though growing, i s unfortunately not as rich and systematic as that available to assess improvements in monetary indicators o f well being and human capabilities. Nonetheless, important insights and hypotheses regarding the empowerment status o f particular groups and 48 "The broad thrust o f the SDPRP [includes]: Improvements in governance to move forward in the transformation o f society, improve empowerment o f the poor and set frameworksiprovide enabling environment for private sector growth and development (p 40) ...To eradicate poverty, correct policies and good governance equipped with efficiency and effectiveness are needed, ...A democratic system, which i s based o n the realization o f people's participation in the development process, and which ensures good governance, is a vital instrument for combating poverty and backwardness (section 6.3)" (FDRE, 2002). 37 people more generally can already be derived from the available body o f information and knowledge. Moreover, this situation i s not unlike the one encountered less than a decade ago when studying well-being from a monetary perspective. The fact that a rich and comprehensive information base to track monetary indicators o f well-being has been established inless than a decade49actually holds promise that more systematic information on people's empowennent situation will rapidly become available. 2.12 Empowermentis the process of enhancing an individual's or group's capacity to make choices and transformthose choices into desiredactions and outcomes. The status o f a person's empowerment i s the outcome o f the interaction between the person's ability to make meaningful choices (Le. his agency), and the context within which the person operates (i.e. his opportunity structure) (see Box 2.1 for a more detailed discussion). A person's agency i s largely indicated by his asset endowments, including material, financial, human, informational, organizational, and psychological assets. A person's opportunity structure i s shaped by the presence and operation of formal and informal institutions, or rules o f the game, i.e. the laws, regulatory frameworks, and norms goveming people's behavior. Box 2.1: Understandingempowerment Ifapersonor group is empoweredthey possess the capacity to make effective choice. This capacity is primarily influenced by two sets o f inter-related factors: agency and opportunity structure. Agency is defined as an actor's ability to make meaningful choices - that is, the actor is able to envisage and purposively choose options- and is indicated in large part by their asset endowment. Opportunity structure is defined as those aspects o f the context within which actors operate that determine their ability to transform agency into effective action. This i s shaped by the presence and operation o fformal and informal institutions, or rules o f the game, including the laws, regulatory frameworks, and norms governing people's behavior. TheRelationship between Outcomes and Correlates of Empowerment Agency )Degree o f Structure T o illustrate the empowerment process by an example, consider an individual whose human assets are improved through completion o f secondary education, while at the same time new opportunities for citizen participation in budget allocations are opened up by the institutionalization o f local level budget planning processes. Usingthe new skills, confidence, and knowledge gained through formal education, and taking advantage o f the opportunities opened up in the planning process, that individual is empowered to effectively participate in local- level decision-making. This corresponds to one "Degree o f Empowerment" noted inthe figure aboves0 Source: Alsop, Heinsohn and Somma 2004 49 The Welfare Monitoring System was established in 1996 and the first comprehensive household consumption survey was carried out in 199516. 50 These different degrees o f empowerment include the existence o f choice, the use o f choice, and the effectiveness o f choice. For further discussion o f these different forms, see Alsop and Heinsohn, 2004. 38 2.13 Assessing a person's empowerment status thus requires information on his agency and his opportunitystructure as well as an assessment of the complex interaction between the two (see Box 2.2 for examples of observable indicators on these two dimension^).^^ An overall review o f people's material, financial and human assets has been given in the previous sections and more detailed accounts will be presented in Part I1o f the report. In this section we will only briefly summarize this information, or refer to it where necessary, and focus on people's informational assets.52 When discussing the opportunity structure, we will especially explore how the persistence o f traditional social practices and norms may limit people's capability to use (1) their (limited) asset base, and (2) the formal opportunities opened up inthe legal and governance structure o f the GoE to make choices and transform these into desiredactions and outcomes. Box 2.2: Intermediate indicators of empowerment: agency and opportunity structure inpractice IndicatodDomain Examples Material Resources Land ownership, tool ownership, housing type FinancialResources Employment, income, indebtedness, access to credit, food expenditures, type o f occupation Human Resources Literacy, numeric skills, education, health status Informational Resources Media access, such as radio, television, newspaper, passable road access, locally functioning post office Organizational Resources Existence and quality o f local organizations, local organization membership Psychological Resources Happiness, self-perceived exclusion, sociability, capacity to envision change Ouuortunitv Structure Indicator/Domain Examples LegaUJustice Protections incivil code, press freedom, protection from political and social oppression, and from domestic violence, statutory rights Government accountability, elections, freedom o f political parties, participation o f Political excluded groups inpolitical processes, availability o f information on government decisions, citizen participation indecision-making, influence o f local elites Availability o f hospitals, education facilities, water points, other basic services, Service Delivery enrollment rates, responsiveness/performance o f service provision agencies, citizen influence odparticipation inservice delivery schemes Household Allocative decision-making within households, division o f labor inhouseholds, domestic violence Existence o f neighborhood associations, cultural and religious traditions and norms, ~ ~~ Community influence o f different groups on community decisions, social interaction, social capital, violence inthe community, informal patterns o f exclusion 2.14 Given that especially women, but also pastoralists, are considered to be particularly disempowered, we begin by discussing their ,agency and subsequently review the GoE's reforms and the continuing effect o f informal institutions on first, women, and second, pastoralists and other social groups. This i s followed by a broader analysis of the government's major vehicle for poverty reduction through citizen empowerment- decentralization. The review finds striking evidence that gender inequalities are extremely pronounced and that women are repressed by traditional gender roles and behavior. " Indicators and measures for empowerment are currently being developed through a five-country study on Measuring Empowerment. For additional information on that initiative, see www.worldbaxkorgiempowerment. 52Systematic information on people's organizational and psychological assets i s currently not available. 39 Pastoralists also emerge as a disadvantaged group. The analysis furthermore suggests both that traditional institutions remain more important to citizens than the formal organizations that are attempting to bring governance and resource control closer to citizens, and that these efforts are also currently inhibitedby informal practices and norms. 2.2.1 The disadvantagedpositionof women 2.15 Gender inequalities are very pronounced and the available data present a grim picture o f women's capability to make choices and transform them into desired actions. This i s most vividly illustrated by the recent experience of a 13-year old rape victim in her elusive quest for justice (Box 2.3). The real significance o f this horrifying, though potentially singular, event becomes especially apparent when augmented with nationally representative evidence on women's attitudes towards domestic violence collected by the Central Statistical Authority o f Ethiopia under the aegis of the Ministry o f Health.53 Box 2.3: Ethiopian rape victim pits law against culture "She rushed through the tangled brusho f onion farms and up the footpaths o f her village. Her shirt was bloody, her clothes were torn and her thighs were bruised a deep shade o f purple, recalled the villagers who were drawn by her screams. Woineshet (...) was running from her rapist. She was abducted one night in March 2001 by four men who hacked down the front door o f her home inthe village o f (..)with a machete. Police and witnesses said she was forced into a nearby shack by the men's leader and raped for two days. She was 13 years old. When the police finally arrived, Woineshet [ran]. The police, who say they have never seen a child covered in so much blood, arrested the suspect. Woineshet's father, (..), 49, who was working and living in Addis Ababa, the capital, went home, looked at his daughter and made [an] unusual decision. For months, he had heard radio announcements and seen bus ads sponsored by the Ethiopian Women Lawyers Association urging the prosecution o f rape cases(. ..) he vowed: This case will go to court. But what happened next made hlm distrust not just justice, but his own common sense. The accused, Aberew Jemma Negussie, was released on bail [and..] abducted Woineshet again, this time hiding her inhis brother's house and raping her for 15 days. She escaped again (...) (...)Woineshet's example highlights an important moment o f change here, as lawyers, police and family members struggle to overcome social taboos and establish a new pattern for investigating and prosecuting rape in Ethiopia. Last year, Woineshet's abductor was taken to court a second time, convicted o f rape and kidnapping and sentenced to 10 years injail. But a judge released him after he had servedjust one month. Woineshet and her father, backed by the Ethiopian Women Lawyers Association and Equality Now, an international women's rights group, are appealing the case to Ethiopia's highest court. (. .) . Woineshet's father recalled that he felt caught between the draw o f the modern world in the capital and the traditions o f the village. H e said he was offered bribes o f cows and cash by local elders to keep quiet. H e also endured pressure from some members o f his family, who thought that Woineshet should marry her abductor. Ethiopian law absolves abductors o f their crime if they marry their victims. Other family members said they also wanted Woineshet to get married because she was no longer a virgin and therefore, they believed, would never find a husband. But her father resisted. "I 'Here Iam, very muchhappy inAddis, and women here are thought, working and smart. They aren't suffering all the time,' I' he said. "Ihave only one daughter. And Ihad that dream for my daughter. That is how I got my courage. Iwanted to see her happy like them.'' 'Maybe They Were Just in Love' "We have a problem here," said Bekissa, the court's president, .,. "The trouble is, this type o f crime happens secretly. You can't gain evidence about her virginity so easily." Bekissa called inJudge Biyo Ukie, who had helped make the decision to allow the accused assailant out on bail. "I don't 53CSA and ORC Macro, 2001. 40 think she was abducted or raped," said Ukie, ......"The health report did not specify that she was a fresh virgin. N o one wants to rape anyone who is not a virgin. Maybe they were just in love. This case has no evidence."Even Woineshet's state-appointed lawyer, Srat Tolch, expressed doubt about the rape charge. "I think Woineshet was like, 'Please rape me.' Maybe he couldn't afford the dowry and they wanted to be together without a formal marriage," he said, shrugging his shoulders. "Culturally, no one rapes a non-virgin. So unless we can prove for sure she was a virgin until the time o f the rape, there is reasonable doubt and the man should just be left alone." A trial was held, and the accused was sentenced to 10 years injail last November. But in December, during a new court session, Ukie, then the judge on the bench, (...) suggested her choices were to marry Negussie or try to send him back to jail.. .Woineshet refused to marry. Her father refused. The police refused. Even members o f the community attending the trial stood up and refused. "I had already made it through the worst nightmares," Woineshet said. "I couldn't have been hurt any more than Ialready was. He raped me. His family beat me. They forced me to be married. Iwanted to speak out. Ihad knownpain for so long. All Iknew was that Ididn't want to be married to my abductor." One month later, for reasons no one is certain of, Ukie let Negussie out o f jail. Ukie said that there were not enough witnesses and that Woineshet was most likely in love with Negussie and ready for marriage. "This family i s only out for revenge," Ukie said in an interview. "Maybe they don't want her to marry him. So they accuse him o f rape." Later, when he was asked about a health report showing severe abuse during the second abduction, Ukie said: "Look, a marriage contract had been signed, and Ithink we should find it. If she wanted to marry him, then ifthere was a rape that makes it legally okay." Then he sighed and said, "Some o f our new laws and ideas o n these matters do not fit with the culture anymore." (,..) Source: Wax. 2004 2.16 "The Husband's Beating Stick is Like B~tter"~~-it found that 85 percent of was women "believe that a husband is justified in beating his wife for at least one of the following reasons," burning food (65 percent agree), arguing with him (61 percent), going out without telling him (56 percent), neglecting the children (65 percent), or refusing sexual relations (5 1 percent) (see Table 2.3). Further exploration indicates that opinions don't change much with age (82 percent o f the 15-19 year olds surveyed still agree with this statement compared to 90 percent o f 45 to 49 year olds), that social acceptance o f domestic violence i s also widely spread in urban areas (69 percent of urban women agreed) and that acceptance only significantly declines with secondary and post secondary education. Nonetheless, even among the more highly educated still 57 percent support the practice. There are no apparent differences in attitude across the regions, with the exception of Harari, Addis Ababa, and Dire Dawa. Yet even in these urbanized centers, the practice carried the support of the majority o f women. 54 This is an example o f the many sayings inthe Ethiopian languages referring to domestic violence inparticular and the disadvantaged position o f women more generally. 41 Table 2.3: Women's opinions on wife beating, Ethiopia 2000 Percentageof women who agree with specific reasonsjustifying a husbandhitting or beating his wife and percentage who agree with at least one of the reasons, by backgroundcharacteristics,Ethiopia 2000 Reasonsjustifying ahusbandhitting or beatinghis wife Goes out Neglects Refuses Agrees with Background Characteristics Burnsthe Argues without at least one food withhim telling the sexual specified Number him children relations Age 15-19 62.6 59.1 52.9 62.8 43.5 82.1 3,710 20-24 64.1 61.4 56.4 64.3 49.5 84.5 2,860 25-29 63.O 61.5 55.4 63.7 52.1 83.9 2,585 30-34 66.7 63.5 57.9 65.2 54.2 85.3 1,841 35-39 64.3 60.9 57.1 67.6 53.5 85.2 1,716 40-44 64.7 60.9 57.6 64.3 56.2 85.6 1,392 45-49 70.2 65.3 61.5 66.9 58.9 89.6 1,264 Location Urban 41.0 39.6 38.2 51.6 29.0 69.0 2,79 1 Rural 69.7 66.1 60.2 67.4 55.7 87.9 12,576 Region Tigray 57.7 56.5 55.9 68.1 41.0 85.7 969 Afar 72.6 71.1 70.6 75.0 70.5 85.7 178 Amhara 66.6 67.9 59.4 65.1 51.6 88.4 3,820 Oromiya 64.7 62.2 55.5 61.5 54.0 84.1 5,937 Somali 48.8 61.5 63.9 65.2 52.0 80.6 175 Benishangul-Gumuz 70.4 61.4 59.0 68.7 51.1 85.0 160 SNNP 73.2 62.3 60.1 73.4 55.2 87.6 3,285 Gambela 57.1 54.4 52.7 56.5 36.1 83.4 40 Harari 28.7 25.8 17.7 28.6 19.9 49.8 41 Dais Ababa 23.3 20.8 22.9 40.0 11.8 54.4 684 DireDawa 42.2 46.7 44.0 55.3 37.8 66.9 79 Education No education 69.5 65.6 60.4 67.0 56.2 88.1 11,551 Primary 61.6 59.6 53.0 65.8 44.8 83.0 2,425 Secondary andhigher 27.4 28.3 27.0 41.8 17.1 56.9 1,391 Employment Not employed 63.7 59.5 54.8 62.5 49.6 84.0 5,630 Employed for cash 60.1 56.9 52.6 63.9 47.1 81.4 3,852 Employed not for cash 68.0 65.9 59.9 66.8 54.5 87.0 2,885 All women 64.5 61.3 56.2 64.5 50.9 84.5 15,367 Source: CSA and ORCMacro, 2001 42 2.17 Clearly, domestic violence is a deeply rooted cultural practice, symptomatic of women's overall empowerment position, with potentially detrimental effects on other development outcomes. Other forms of violence toward women also carry their broad support. For example 60 percent o f all women stated that they supported female circumcision, while 80 percent o f them had been circumcised (see Appendix 3, Table A.2.1). Moreover, the acceptance o f domestic violence, and women's disempowennent more generally, not only has a striking impact on the well-being o f the women themselves, it also appears highly correlated with other development outcomes such as their children's chances of survival. The under-five mortality rate for children o f women who do not accept any o f the given reasons as justification for abuse is 154 out o f 1,000 live births, while for those accepting at least one reason the rate i s over 192, Le. almost 40 children per 1,000 live births more.55 While commonly associated factors such as education and urbanization most likely partly drive these results, the difference i s nonetheless striking. In conclusion, as will be shown below, these practices and their wide acceptance by the women themselves are symptomatic o f their overall empowerment position. We now turn to a more systematic exploration o f the underlyingforces, Le. women's agency and their opportunity structure. 2.18 Women have extremely low educationalachievements and virtually no access to external information. As indicated earlier, women consistently have lower levels o f education than men, with over 75 percent having no education at all (compared to 50 percent of men, see Appendix 2, Figure A.2. l).56 despite encouraging improvements, the Moreover, gender gap in primary school enrolment remains ~ i g n i f i c a n t . ~Exposure to mass media ~ (radio, TV or newspaper) i s extremely low across Ethiopia, and it i s even more pronounced among women. While 73 percent o f men have no access to mass media, this i s the fate o f 86 percent of all women. Less than 11 percent o f all women listen at least once a week to the radio, compared to 24 percent o f the men.58 2.19 Women have a significantly lower employment rate than men, and have little representationin decision making positions. Nearly 43 percent o fwomen are unemployed, and over 36 percent o f them are chronically unemployed (see Appendix 2, Figure A.2.3). When employed or receiving cash eamings, about three quarters o f women decide themselves on how earnings are used. While there are large differences between the regions in who decides on the spending o f women's earnings (35 percent o f the women themselves in Benishangul-Gumuz versus 82 percent in SNNPR), there are no significant differences between urban and rural or educated and uneducated women. Despite such authority over spending decisions in the household, professionally women have little representation in decision-making positions. For example, although 40 percent o f government employees are women, 71 percent o f them are concentrated inthe lower levels.59 55CSA and ORC Macro, 2001. 56Bridging the Gap, 2003. '*Worldand 57 Bank, 2002. CSA ORC Macro, 2001. 59World Bank, 2004 (draft). 43 2.20 The GoE recognizes the disadvantagedposition of women, and has implemented a number of policies, laws, and initiativesto promote women's empowerment, including removal o f discriminatory laws from the Constitution. With the announcement o f the National Policy on Women in 1993 and promulgation o f the new Constitution in 1995, the FDRE highlighted its commitment to the equal development o f women. Article 25 o f the Constitution clearly guaranteesequality and makes any discrimination on the grounds o f race, color and sex illegal. 2.21 However, these policies and laws are often weakly enforced, and in many cases provide contradictory or incomplete coverage in their protection for women. For example, while violations such as female genital mutilation, wife battering, domestic violence, and sexual harassment are outlawed in the Constitution, the penal code contains no provisions for adjudicating them, and existing laws are often applied by judges in a manner that does not take account of women's rights.60 2.22 Most rural people in Ethiopia continue to apply customary laws to their economic and social relationships, and the example of Woineshet is not an isolated incident.6' Although these customary laws are not legitimized inthe Constitution, andArticle 9.1 o f the Constitution states that "any law, customary practice, or a decision o f an organ o f state or a public official which contravenes this Constitution shall be o f no effect," especially within the rural context customary practices have greater influence on gender relations than the formal system.62 Ellen Alem, a legal aid service coordinator with the Ethiopian Women Lawyers Association, said it i s almost impossible to bring a rape case to court in rural areas when the victim's virginity is questioned. She i s lobbying the government to specify in the law that non-virgins can also be victims o f rape and that their cases should be taken serious~y.~~ 2.23 While the legalintegration of customary conflict resolutionmechanisms with the civil courts was meant to enable citizens to retain their ethnic and religiousidentities, in practice this has reinforced damaging attitudes and customs toward women. Article 34(7) o f the Constitution reserves the option to adjudicate disputes related to personal matters inaccordance with religious or customary laws, rather than underthe civil code, ifthe parties to the disputes agree. In practice, personal disputes, particularly between men and women, are frequently directed to traditional adjudication mechanisms by the choice o f men, without the consent o f women. InMuslimareas, if a husband goes to the Sharia court first to institute divorce proceedings, then the wife often does not have recourse to the civil Focus group discussions among Orthodox Christians in Addis Ketama also note that if there i s a conflict between husband and wife, the case i s first handled by a traditional court. They note that even ifone goes directly to formal courts, the case would be passedto traditional courts.65 6oWorld Bank, 2004 (draft). 6'Similar accounts are regularly appearing inthe local newspapers, 62World Bank 2004, (draft). 63See Wax, 2004. 64World Bank, 2004 (draft). 65Legovini, 2004. 44 2.24 Testimony from a 32 year old, well educated, heado f the kebele women's association enough money for the house expense. ... He gets drunk every night and disturbs our peace. shows how damaging this situation can be for women: "My husband does not give me One day Ihad had enough and told him to leave the house, which Iown. Surprisingly, the community leaders said Ishould leave the house. ,.. At the end, Ihad no choice but to continue living with him".66 This was a repeated theme (even in many cases from men): traditional courts are the first recourse, and they generally favor men. (See Box 2.4 for a discussiono f gender inequalities inasset distribution indivorce.) Box 2.4: Asset distribution at divorce Recent studies of asset disposition during divorce show that halfof surveyed monogamoushouseholdmembers expect a household's land and house to go to the husbandupon divorce, while 40 percent expect it to be divided equally. If there i s a fault-based divorce (such as resulting from adultery, infertility, drunkenness, spouse beating, etc.), the distributional situation changes dramatically. The study found that if the husband i s at fault, the wife is slightly more likely to be granted land and livestock than if it i s a no-fault divorce. If the wife is at fault, however, no notablepercentage of those surveyedbelieved she would receive land at divorce, andjust 34 percentbelieved she would even receivethe livestock that she brought into the marriage. 2.25 One of the many consequences of the dominance of traditional beliefs and practices is continued prevalence of wife beating. As discussed above, the widespread acceptance o f violence towards women clearly indicates that it i s a deeply rooted cultural practice. As one man noted: "Women are obedient to their husbands because they have trained themselves to be obedient, so most of the time there i s no problem.... Sometimesmen can beat their wives in the community, but it i s taken as an expressionof love, so the women didn't take it as a big problem." Another manput this more bluntly, noting that if a woman does wrong she should be beaten. "Only a pot does not like whisking," he said.67(See Box 2.5 for additionalexampleso fviolence against women.) Box 2.5: Violence and abduction in marriage One of the more extreme forms of violence is by its very nature one that clearly goes against the wishes of women. This i s telefa, the practice of abducting a woman if she shows unwillingness to be part of a proposed marriage. In some cases telefa involves beatingsand rape. One study of rural householdsfound that 10 percent of all marriages were described as "kidnappings," and 2/3 of these (or six percent of all marriages) involved women being forced into marriage despite her wishes (the other 1/3 were more likely a couple marrying despite their families' disapproval).68 According to Ethiopian law, rape and abduction are not punishable if the victim "freely" contracts a valid marriagewith the abductor. One study respondent describedhow the parentsof his bride-to-be refusedto allow her to marry on the marriage day, arguing that some agreed gifts had not beenmade. The respondent's relatives kidnapped a teenaged girl on their way back from the bride's village because they did not want food prepared for the wedding to go to waste. The girl broke her arm fighting her abductors,but was still married that day. Source: Fafchamps and Quisumbing, 2002 66Legovini, 2004. 6'Legovini, 2004. Fafchampsand Quisumbing, 2002. 45 2.26 In sum, women's narrow asset base, not in the least their limited access to information and education, and the strength of traditional norms and behavioral systems reinforce each other in limiting women's ability to make effective choices. A continued focus on closing the gender gap in education, increasing access to information and enhancing women's awareness o f their rights via mass media i s needed to improve women's empowerment positioning. Inaddition, bolstering the GoE's capacity to enforce the laws that are inplace to protect women, and to close existing loopholes, gaps, and contradictions in the justice system that allow it to be manipulated to the disadvantage o f women will be necessary steps to start addressing the pronounced gender inequalities in Ethiopia. 2.2.2 The role of informalinstitutionsamongpastoralistsand other social groups 2.27 There appear marked limits to pastoralists' individual agency. According to a 2001 World Health Organization report, only about five percent o f the mainly pastoralist Afar population had access to proper health care, with two hospitals in the entire region.69 The Afar and largely pastoralist Somali regions rank behind only Amhara in proportion o f population with no access to mass media, and have the lowest levels of savings and per capita daily calorie intake among all regions. Somali has the lowest proportion o f population currently employed, and the highest proportion o f chronically unemployed, as well as the lowest net primary school enrollment rate. O f more concern i s that the situation appears to be deteriorating. For example, both regions have seen significant increases in the proportion o f income going toward food purchases. 2.28 Recent policy developments, ministerial restructuring, and statements have indicated that the GoE fully recognizes the extent of pastoralists' exclusion from development opportunities. The SDPRP notes that in pastoralist areas "there is poor understanding o f the holistic nature o f the problems. Interventions were following a piecemeal development approach and development failed to be s~stainable".~~The government has recently made structural changes in the bodies responsible for pastoralist policy. This includes the appointment o f a Minister o f Rural Development, whose ministry has been established to address the governance and administrative needs o f "emerging regions" which comprise the pastoral areas o f Afar and Somali Regions. In addition, a Pastoral Unit has been established within the Ministry o f Federal Affairs. 2.29 The GoE's rethinking of its approach to development and governance in pastoralist areas also acknowledges pastoralists' reliance on, and in some cases the resurgence of, traditional patterns of organization and behavior. As with women, there are indications that this persistence o f informal institutions can stifle citizens' capacity to make effective choices in support o f their own development. In an environment where traditional organizational patterns demonstrate both positive and negative consequences from an empowerment perspective, both the government and citizens have shown remarkable adaptability. 69Cited inUNOCHA, 2002. 'OFDRE,2002 (p.70). 46 2.30 There has been some interaction between traditional and state structures of representation in pastoralist areas. In many instances this i s made possible by different positions being heldby the same pastoralist representative. For example, a number o f federal level MPs are also high-ranking pastoral elders. In Borona, there are estimates that about 20 percent o f the kebele council (about 100 people) are also elders, and this has had a positive impact on citizen perceptions o f kebele councils, where the councilors are considered "one o f thepeople." 2.31 In some areas, specific mechanisms/structures have been established to allow traditional "representatives"to engage with formal politicalsystems, yet these have had mixed results and sometimes serve to reinforce traditional patterns of exclusion. In Somali Region there i s a Constitutional provision for an Assembly o f Elders and Ethnic Leaders. This has not yet been formally established but i s nevertheless currently operating and i s intended to bring traditional conflict management mechanisms into government. In addition, the government has appointed a number o f elders, known as "amakari" to advise them at regional, zonal and woreda levels on matters relating to customary and community issues. There exist conflicting views regarding the extent to which this structure i s in fact an effective means o f mediating interests, stemming primarily from the fact that the amakari were under the pay of the government and had therefore become divorced from the communities they represented. According to Lister (2003), "there i s also widespread agreement that the amakari exercise considerable control over the voting habits o f the communities from which they come.'' More importantly, while traditional pastoralist systems emphasize equal participation by all resident households, their decisions are based on the consensus reached by adult males, with women and outcast groups (such as disfavored clans/subclans) excluded from these proce~ses.~' 2.32 Other social groups have similar experiences, with both positive and negative consequences. For example, where Gurage traditions o f collective action persist, these areas have seen empowering impacts at the community level (see Box 2.6). However, there are also cases where the continued emphasis on traditional practices appear to result in arrangements that are problematic for individuals to freely express their agency, and hence stifle development. For example, in a recent study o f the political and historical traditions o f the Sidama people, Aadland (2002) notes a widespread skepticism with which the Sidama view the state and its administrative system, and a resulting negative impact this is having on development and social relations (see Box 2.7). "Lister,2003;PolhemusandYohannes, 2003. CitedinWorldBank,2004(draft). 47 Box 2.6: Gurage traditional law Zewdie and Pausewang (2002) note that while government courts enjoyed legal backing, customary courts such as the Sabat Bet Gurage exercised moral and ritual sanction. Although the existence o f a centralized court and police system has somewhat diminished the power and applicability o f traditional systems o f arbitration like the Sabat Bet Gurage, it is indeed a measure o f the vitality o f those traditional institutions, that, far from disappearing completely, they continue to operate extensively. This continued reliance on traditional organizational practices is often quite a positive mechanism for community-level development. Gurage customary law has helped to foster and sustain the Gurage's sense o f identity, and has played a significant role inmobilizing Gurage populations to develop their region. Remarkable progress made in road construction, for example, would have been inconceivable without the network put in place by those customary institutions, not to mention the importance that traditional Gurage society attaches to transport networks. There have also been attempts, spear-headed by both urban and rural Gurage populations, to make such institutions adaptable to contemporary reality by trying to tackle a range o f issues from rural development to substance abuse and AIDS. Source: World Bank, 2004 (draf) 2.33 The GoE is therefore faced with the dual problemof improvingthe effectiveness of policies to enhance equity and inclusion, and balancing a respect for traditional cultures. The current approach, which blends a mixture of increased autonomy within the ethnic regions with continued exercise o f federal authority over broad policy directions, appears to result in somewhat contradictory outcomes. On the one hand, ethnic tensions are to some degree diminishing within the country; on the other hand, citizens in ethnic regions are experiencing a pull toward traditional ways o f organizing daily life, with mixed results in terms o f empowerment. The GoE therefore faces a difficult decision regarding whether it i s time to retract central control in ways that professionalize government and the state-society interface and give rise to new forms o f organization at the local level. Pursuing this will require a system where citizens trust and work with the government instead o f using their traditional, sometimes unequal, institutions. This requires a demonstrated capacity by govemment to effectively represent, be accountable and respond to citizens. Box 2.7: The effect of traditional Sidama institutions With the coming to power o f the new government and the reestablishment of control from above, the Sidama's previous experience o f subjugation and domination appears to have been revived, and this has in turn revitalized old customs, traditions, and institutions without analysis o f their current relevance or democratic potential. Two traditional cultural administrative structures are noteworthy components o f this revival that have significant impacts on power relations within Sidama communities. First, the Sidama are split into those who do and don't possess anga, or "noble heritage." Those who are not o f `noble heritage' are potentially victims o f social discrimination, since in historical times they were not given land, and were forced to make a living from professional work as tanners or potters, weavers or blacksmiths. Further, the stratification layers in the Sidama hierarchical structure consist o f four strata: high-ranking lineages, low-ranking lineages, caste-groups and slaves. Along with this wave o f reaffirming traditional Sidama organizational structures and norms, a strong trend o f resistance against innovation is emerging. Recent research shows that such indigenous structures become "nostalgic and rather mythological, out of touch with the reality o f present day challenges, as a result o f being neglected". Thus, an uncritical attempt at the revival o f such institutions i s likely to exclude large segments o f the Sidama population and render equality and general participation difficult. Source: World Bank, 2004 (draf) 48 2.2.3 State-society relations-the role of informationand traditionalinstitutions 2.34 People's lack of material, financial and human assets is compounded by their sheer disconnect from the rest of the world, suggesting that they are likely to be substantially limited in their capability to make choices and transform these into action to improve their own lives, irrespective o f their opportunity structure. The emerging picture from the cursory review o f the people's asset base in the previous sections, indicating that people have little material, financial and human resources will be documented in much more detail and confirmed in Part I1of the study when we explore the determinants of monetary well-being. Compounding the limited agency resultingfrom this poor material resource base i s the sheer disconnect o f the Iarge majority o f Ethiopian citizens from the rest o f the country and the world. Inparticular, people's immobility, their limited access to roads and transport services, and above all, their lack o f access to information combine to completely isolate Ethiopian people in their traditional habitats, defacto limiting their capability to aspire and expand their choice set. Figure 2.4: Incidenceof radioand TV ownership among rural householdsin selected Sub-Saharan African countries, 1995-2001 > 110 , Note: Incidence o f radio ownership is added to incidence o f TV ownership. As a result, total incidence can reach up to 200 % maximum. Source: Own calculations from 1995-2001 DHS surveys 2.35 There is virtually no migration or exposure to information in the country. According to the 1999 Labor Force Survey,7285 percent o f the rural population has been continuously residing in the same area (woreda) as they were born, without ever having lived elsewhere. Only 2.6 percent of the total rural populations were recent migrants, i.e. people l2 CSA, 2000. 49 who have moved into the area over the past five years.73 Given that 85 percent o f the total population lives in rural areas, exposure to new ideas or outside influences i s clearly very limited. Moreover, while access to mass media could make a difference in such circumstances, according to the 2000 Demographic and Health Survey (DHS) only 13 percent o f the rural households owns a radio (and virtually nobody owns a TV), and 87 percent of the population i s not exposed to any form o f mass media (radio, newspaper, TV) on a regular basis.74 These figures are striking even within the Sub Saharan context, where Ethiopia ranks lowest in terms o f radio ownership among all countries (Figure 2.4). Sheer remoteness and isolation epitomizes life in rural areas, rendering it extremely difficult to reach the rural population, limiting cross-fertilization, and perpetuating ongoing knowledge patterns, traditional practices and customs. 2.36 Turning to people's opportunity structure, the continuing focus of the central governmenton ensuringstability has been effectivein managingthe period of transition from the Derg regime, the conflict with Eritrea, and ethnic reconciliation within the country. The country i s seen as increasingly stable, with one index o f law and order showing improvement from 33 out o f 100 in 1990 (where 100 represents highperformance), to 83 out o f 100by 2002.75 2.37 However, the country ranks in the bottom 25 percent globally in indicators of voice and accountability,and its position deteriorated from a 30.9 percentile rank in 1996, to a 14.6 percentile rank in 2002. Government performance suffers in this environment, where Ethiopia ranks in the bottom 25 percent globally, and has fallen over the years from a 34.6 percentile rank in 1996 to a 16 percentile rank in2002 (Figures 2.5 and 2.6). Press freedom in the country is limited (an index score o f 0, the lowest possible score), and bureaucratic performance i s recorded as highly ineffective, with an index score o f 25 where 100 i s considered low risk.76 73People moving inand out of the area were consideredmigrants if they spent less than 6 continuous months in the area where they were surveyedprior to the survey, or if they intend to leave the area again within the next 6 months. 74There are currently three governmentowned radio stations. l5World Bank, 2004f. 76World Bank, 2004f. 50 Figure 2.5: Governance indicatorsinEthiopia Government Effectiveness I 0 1996 I i 0 1 9 9 8 ~ Voice and 29.8 1 I H2000 Accountability I ~ 0 10 20 30 40 50 60 70 Source: Kaufmann, Kraay and Mastruzzi 2003 Figure 2.6: Governance inEthiopia compared with similar categories(2002) Govemment ! Effectiveness 28 9 I I I I 10Ethiopia 1 Voice and Accountability 31 I 0 5 10 15 20 25 30 35 Source: Kaufmann, Kraay and Mastruzzi 2003 2.38 Through the SDPRP and other initiatives, the GoE has recently shown a strong commitment to professionalizing its governance apparatus and empowering citizens through political, fiscal and administrative de~entralization.~~ However, these changes are relatively recent and, as in other countries, processes o f change of such magnitude are slow. Great strides have already been made in terms of decentralizing to certain woredas and in urban areas, but even here government representatives and citizens report that progress toward making the state apparatus more responsive and accountable i s limited as of yet. With 77The recently initiated Public Sector Capacity Building Program marks the seriousness o f GOESefforts to decentralize and professionalize services, as does its commitment to the Civil Society Capacity Building Project, currently under preparation. 51 formal institutions o f govemment in a state o f transition, as the following discussion shows, there i s evidence that many citizens rely on their own informal forms o f organization and norms o f behavior to manage everyday life rather than those o f government. In addition to not achieving the SDPRP's objectives in terms o f empowering its citizens, one danger inthis i s that informal practices at times exclude marginalized individuals and groups, thereby reducing their capacity to make effective choices about their own development. As local government i s the main point o f interface between the state and its citizens, we first explore the current state o f empowerment o f local governments and then the relationship between local governments and citizens.78 Empowerment of local governments 2.39 As the GoE operationalizes its strategy for decentralization, it faces three critical issues inempowering sub-national levels o f govemment. 2.40 First, there remain important capacity constraints within political and bureaucratic bodies at all levels (see Table 2.4) (Appendix 3, Table A.2.2 provides a description o f Ethiopia's administrative structure and responsibilities). Local governments find it difficult to recruit staff equipped with the skills required for management of the processes o f decentralized, democratic governance. The educational qualifications of local government staff have improved from a situation in the mid-90's where the mode was 6th or 7th grade to one where most staff now have at least a diploma.79 However, the qualifications achieved rarely relate to the substance o f govemment, which means staff have a steep-and often unaided-learning curve. Community Capacity to: understand roles and responsibilities; be informed on development; identify Level development needs; demand accountability; participate indevelopment programs; pay rates etc.; engage with support agents. Woreda Level Capacity to: establish systems and structures to: manage finances, prepare and implement budgets; plan implement and monitor development programs; promote civic participation; ensure service delivery; manage capital and human resources; promote economic development and sustainable land use. Capacity to: support and regulate decentralization and local govt. i.e. to: provide clear Regional legislation, policy and technical guidelines; establish and manage woreda financing systems; Level provide training; facilitate organizational development; co-ordinate external support initiatives; facilitate information sharing; establish monitoring systems. Federal Level Capacity to: establish policy; set minimumservice standards; outline C B designs; outline staffing structures; mobilize external resources; co-ordinate, monitor and evaluate PSCAP and LG performance; ensure training facilities. 78 An objective o f the SDPRP is to "Establish institutional arrangements to ensure empowerment, democratization and efficient administration at woreda and kebele levels." 79 Vaughn and Tronvoll, 2003. There remains room for caution, however, as these capacity improvements have come largely thanks to government-initiated programs, thus raising the possibility o f enhancing upward loyalty rather than downward accountability. 52 2.41 A second constraint is the inadequacy of budgets. While the switch to woreda block grants has the potential to enable more locally responsive planning and budgeting (see Box 2.8), substantial challenges in implementation remain. Even with the intention o f supporting local level discretionary expenditures, the paucity o f funds available via block grants results inmost funds being allocated to recurrent, rather than capital, expenditures. For example, salary obligations consume between 85 and 90 percent o f the overall budget. At the kebele level, the fact that kebeles receive little o f the capital finding requested through the planning exercise creates few incentives for councilors to engage with citizens. A recent assessment o f local governments noted that "kebele councils needto control a budget, which will deliver tangible demand-led projects, so [that] the council is seen to be responding to local needs".80 Box 2.8: Recent reform in fiscal resource transfers InJuly 2002, the GoEbeganapilot reform o fthe resource transfer systeminaneffort to ensure less discretion in total transfers to woredas. Under the previous system, transfers from the federal level to the regions were made on a formula weighing population (55 percent), a composite development index (20 percent), an index o f revenue effort (15 percent) and a poverty index (10 percent). Regions distributed these to zones sometimes mimicking the federal formula (e.g., Amhara, SNNP), and sometimes using alternative formulas. Transfers from zones to woredas, however, were always discretionary, although there was evidence that food-insecure woredas did receive transfers greater than their share o f zonal population, while the reverse was true in food-secure woredas. Under the revised system, currently being piloted in Oromiya, Tigray, SNNP, and Amhara and intended to spread nationally in coming years, transfers are now made directly from the Regions to the woredas, and based essentially on the same formula as from the Federal level to the Regions. Source: World Bank Country OBce in Ethiopia, 2002; World Bank, 2003a 2.42 Finally, the transition from a centrally managed system to a mode of devolved fiscal responsibilityrequireschanges in expectationsand behavior. The previous system o f transfers through zones was critiqued for capture o f finds at the regional and zonal levels and imposing limits on the ability o f woreda officials to adjust spending on the basis o f local realities.8' Now, block grants to woredas allow for a more transparent and egalitarian system o f resource transfer, but it i s apparent that some regional officials continue to limit "the discretion o f individual woredas in expenditure assignments through providing budget guidelines on the functional and economic split o f expenses". 82 In addition, federal guidelines with respect to wage levels and firing staff limit woreda councils' capacity to re-allocate financial resources.83 To some extent, capacity problems at both the woreda and kebele levels explain a continued reliance on higher levels o f government. In2002 most woreda staff did not think that zonal interventions eroded their authority, but rather that their own administration lacked the skills necessary to formulate budgets. 2.43 Historically, woreda and kebele officials looked upward in terms o f loyalty, concern, and accountability, rather than downward toward their constituencies. For example, while INTRAC, 2004a. "See,forexample, Girishankar, AlemayehuandAhmad, 2001. 82World Bank, 2003a; World Bank, 2003b. 83World Bank, 2003b. 53 they regularly transmitted financial reports to the zone, these were not disseminated to kebeles, and there was no evidence o f kebele-level gatherings to discuss the woreda budgetaS4 As the 2004 studies quoted in this section indicate, this behavior has changed in a small number o f locations, but it will take time, experience, capacity building o f both government representatives and citizens, and the development o f appropriate incentive systems before the traditional tendency for upward rather than downward accountability i s fully addressed. Local government and citizens 2.44 In practice, citizen voice and influence are partial in Ethiopia. Citizens are largely disengaged from local govemance and decision-making processes. The Woreda Studies vividly illustrate the disconnect between what communities identified as priority needs and what i s ultimately communicated to woredas, noting that "communities had very low expectation o f gaining from the official planning While it is unrealistic to expect that citizen "wish lists" will always be translated into official programs and policies, the remarkably low expectations o f citizens are evidence that more i s needed from local level officials to meet citizen demands, to manage expectations where they cannot be met, and to explore synergies betweenlocal and national level priorities. Box 2.9: Citizen's relationshipswith kebeleand woreda governments The predominant reasons for organizations/ institutions o f civil society to associate with kebele and woreda admtnistrationsi offices were: Kebele cabinets commonly communicate information and make requests for community assistance (e.g. labour), particularly through agricultural work groups and iddir. Organizations/ institutions at kebele level use kebele cabinets in the enforcement o f disciplinary actions, when penalties by group leaders have proved ineffective. Woreda sector offices have regulation and financial control responsibilities with respect to cooperatives and credit groups. Technical support and service delivery. These relationships are characterizedby: Extensive overlap in membership: members o f kebele cabinets are also members o f iddir and other centrally important organizations/ institutions inpeople's lives. A wariness on the part o f members/ associates o f organizations/ institutions o f civil society about the possible nature o f government support (when perceived as `interference' rather than, for example, technical and financial management support). A submissive respect for hierarchal authority, which considers challenging public figures to be inappropriate. As Box 2.9 shows, citizens interact with government in very limitedways. As yet processes o f decentralization have not given many the capacity to influence resource allocation or management, and there i s little evidence of government-either elected representatives or line ministrystaff-ever being held accountable by citizens.86 84World Bank Country Office inEthiopia, 2002. 85World Bank Country Office inEthiopia, 2002, p. 30. 86JNTRAC, 2004a; Vaughn and Tronvill, 2003. 54 2.45 Citizens rely on their own informal organizations, rather than counting on formal governmentbodies to deliver basic needs. Research in 1999 indicated that citizens rely more on informal local institutions than on formal governmental entities or NGOs. This was particularly true inrural areas, although urban residents inmany cases also ranked formal institutions low in terms o f their importance in everyday life (see Appendix 3, Tables A.2.3 and A.2.4). Where kebeles were mentioned as important, this was largely "because (i)[they link] the community residents with the government, and (ii)that is where community residents go to receive ID cards or any other kind o f official d~cument".~'A 2004 study in nine woredas across six Regions found that households associate with between one and 24 organizations at the local Membership-based community organizations were found to be particularly important inpeople's lives. The most commonlyjoined groups included iddir (ranging from 100 percent o f kebele households to just under 50 percent), and agricultural mutual assistance groups (ranging from almost 100 percent o f kebele households to just over 38 percent). Also prominent, but less so, were rotating savings and credit groups (iqqub), social religious groups, businesdtrade groups, women's organizations and clan, sub-clan and age groups (particularly in pastoralist areas).89 Government organizations were rarely mentioned. 2.46 Kebele councils, as well as representatives from other levels of government, recognize this, and often use community organizations as a communicationand access point to the community. Citizens are however concerned about the influence o f local governments, and some people are reluctant to join iddir because o f concerns that they might be used as political instruments. Inone kebele, while the iddir was perceived to currently be free from government interference, "the possibility o f this was felt to always be present, especially if they were to obtain significant funds for development - from an NGO, for example.... Projects initiated by government or NGOs are unlikely to be `owned' by the people - government and NGOs simply tell you what they are going to do and go about it in their own way".9o 2.47 Citizens appear to continue to have a general distrust of government that limits the extent to which they rely on formal institutions. As two studies completed in the last few months indicate, "there is still a strong residual suspicion o f the state"." A survey on corruption conducted in 2001 by the Ethics Subprogram o f Ethiopia's Civil Service Reform Program found that citizens rank corruption second inpriority in a list o f 18 socio-economic problems, Forty nine percent of a sample o f over 2,500 people reported observing a corrupt act by a public official in the past two years. Thirty-five percent o f 477 households involved in lawsuits said there had been an indication of bribery during the proceedings. Finally, payments to public officials.9 households reported spendin8 an average o f 4.5 percent o f household income on unofficial _____ ''INTRAC, "RahmatoandKidanu, 1999. ''INTRAC, 2004a. The median was nine. 2004a. ''INTRAC,2004b, 2004a. INTRAC, referencing INTRAC, 2004a. 92Cited inPolhemusand Yohannes, 2002. 55 2.48 Citizen reliance on their own resources and local informal organizations has limits. Researchers in Ethiopia have identified numerous strengths associated with these organizations, but with these also come significant weaknesses (Table 2.5). In particular, exclusion along the traditional fault lines o f society and the limited power o f local organizations to influence government indicate that this dependency may not only be exacerbating existing problems but that it also does little to enhance citizens capability to interact with government. Table 2.5: Strengths and weaknessesof informalorganizations - Strengths Weaknesses Spontaneous formation Short-lived; liable to dissolve Voluntary membership Exclusive membership Voluntary contributions Social compulsion Self-reliance and autonomy Weak and isolated Membership control Leadership control Mutual benefit o f members Benefits limited group Sharing o f risks, costs, benefits Disadvantages poorest Group cohesion Excludes others Trusted and respected Limited to the group Social security and insurance Assistance limited Informal and flexible Hidden and invisible Accountable leadership Weak control o f leaders Democratic structure Represent hierarchies Efficient pooling o f resources Wasteful consumption Promote democratic behavior Influence and role limited Role inresource management Limited influence and authority Role inresolving conflicts Lack o f enforcement mechanisms Local counterbalance to state authority Limited scope to act; no higher groupings Promotes participatory development Limited to small isolated grouD - > Promotes sustainability o f projects Weak ability to manage projects Source: Pankhurst, 2003 2.49 Lack of citizen participation in local governance and resource management is reducing returns to government spending. Where local authorities have engaged with citizens and collaborated in local development initiatives results have been remarkable.93 Evidence suggests that citizens are eager to participate, but only where they know it will make a difference. For example, the Woreda Studies note that "The increased enthusiasm displayed inrelation to the planning of activities outside the formal woreda plan suggests that apparent problems in engaging participation relate more to a lack o f motivation than to capacity constraints." When communities do participate, their contributions often tum out to be major sources o f development finance for infrastructure construction and maintenance. In Butajira 1.4 million Birr (approximately US$165,000) was contributed in cash by local people to the construction o f a new hospital building. Additional contributions o f labor and materials reduced the initial cost estimate from 13 million Birr to seven million Birr. The local government provided some technical advice and has now assumed complete responsibility for hospital management, staffing and financing.94 In the Amhara Region, the value o f community contributions appeared to exceed official capital transfers to the woredas there. 93INTRAC, 2004b. 94Report o f field visit to Butajira World Bank project preparation team June 2004. 56 Awabel Woreda, for example, received a subsidy o f ETB 0.74 million in 1998/99, while communities contributed labor and materials valued at ETB 0.82 million for ago-forestry, soil and water conservation, roads, water schemes, and housingg5 95World BankCountry Office inEthiopia, 2002. 57 CHAPTER 3. PEOPLE'S WELL-BEING -CONCLUDING REMARKS 3.1 Despite a decade and a half of relative peace and stability, broad economic reforms, and far-reaching political decentralization, progress inthe well-being o f Ethiopia's people has remained below expectations. To examine how Ethiopians fared over the past decade and a halfwe took both a utilitarian and capability approach and explored progress on monetary and non-monetary indicators of well-being. We looked in particular at people's poverty status, their expectations with respect to poverty in the future, i.e. their vulnerability, and the evolution o f inequality in Ethiopian society. Here, we summarize what we have learned and identify areas for further analysis and data collection to enhance our diagnostic o f people's well-being in Ethiopia. Follow up actions in terms o f strategies, policies and investments to improve people's well-being will be discussed in the concluding chapters o f Parts I1and I11 following an analysis o f the determinants o f monetary and non-monetary indicators o f people's well being respectively. 3.2 First, the micro and macro evidence paints a picture of limited to no decline in consumption poverty incidence since 1992. While there i s a growing consensus that poverty incidence in urban areas i s deteriorating, rural poverty incidence remained largely constant with signs o f a one to two percentage point decline. Overall, consumption inequality inEthiopia remains low, though inequality inurban areas is on the rise. The reasons behind these broad trends are largely found in the disappointing performance o f the agricultural sector which barely kept up with rural population growth. Whatever poverty reduction occurred in rural areas probably resulted from improved access to services and infrastructure. Inurban areas, growth inthe service sector (estimated at about sevenpercent per year), which fuelled overall economic growth, was substantially eroded by urban population growth (estimated at 4.7 percent) following rural-urban migration. Moreover, growth in the service sector was mainly driven by government expansion, with a strong increase in military expenditures between 1995 and 1999 financing the border war with Eritrea, followed by a shift out o f defense into poverty sectors since 2000. While not much (urban) poverty reduction i s to be expected from the increase in expenditures on defense, it was probably too early to already feel the positive effects o f the more recent investments in the poverty sectors (e.g. doubling o f expenditures on education). Also, some o f these investments had a deliberate rural bias, which i s consistent with the small decline inrural poverty inthe absence o f agricultural per capita growth. Finally, only a small percentage o f the urban population appears to have benefited from the growth in the service sector during this period, as exemplified by the observed increase inurban inequality. 3.3 Poverty is more severe among the uneducated, agriculturalists and coffee growers. While education is generally observed to be an important way out o f poverty in developing countries, the probability o f being poor is far higher for households with uneducated heads. In support of this finding based on bi-variate analysis, people in participatory surveys often mention education as the most important way out o f poverty for their children. This will be further investigated in a multi-variate setting in Chapter 4. Households employed in the service or industry sector appear better off than those in agriculture. The implications o f this observation will be addressed at length in Chapter 5 and 6, Poverty incidence among cash crop growers was estimated to be substantially lower than 59 poverty among agriculturalists. Households growing coffee appeared to be equally poor during the 1995-99 period as the rest of the nation. This is surprising, especially given the coffee price peaks in 1997 and 1998 and the general observation that coffee growers (and cash crop growers more generally) tend to be richer. This observation deserves further investigation given that about 30 percent o f the rural population grows coffee and in light o f the recent collapse ininternational coffee prices. 3.4 Risks permeate daily life in Ethiopia, and drought shocks severely affect current and future consumption. The slow pace o f poverty reduction does not mean that everyone everywhere endured the same fate. Indeed, the averages hide a substantial amount o f churning. Many people move in and out of poverty, often in tandem with annual rainfall patterns. This underscores the immediate impact o f rainfall on households' current consumption. Moreover, there are clear signs that the negative effect o f severe rainfall shocks often persists over time. For example, households who reported to have suffered substantially more during the 1984I5 famine experienced on average 16 percentage points lower growth in the 1990s, and 10 percent lower rainfall was found to reduce consumption growth rates by one percent four to five years later. Harvest failure proves to be especially harmful for child growth and female enrolment, thereby permanently damaging the earning prospects o f the next generation. This underscores the need for appropriate risk management programs. 3.5 In addition to droughts, HIVIAIDS, and also malaria, continue to present substantial risks to the Ethiopianpopulation. The prevalence of HIVIAIDSis estimated at 4.4 percent o f all adults infected (2.6 percent inrural and 12.6 percent inurban areas) in2003. The latest estimates further suggest that the rate at which the HIVIAIDS epidemic i s progressing, has been declining over the past years, especially inurban areas, which i s in line with observed changes in behavior. Nonetheless, with about 1.5 million people currently infected and an estimated 539,000 AIDS orphans, HIVIAIDS continues to threaten future development and poverty reduction in Ethiopia. While largely neglected, malaria i s also a major contributor to the disease burdenin Ethiopia. Highmalaria endemicity further presents an important obstacle to unlocking the agricultural potential o f the lowlands. 3.6 Second, in contrast to lackluster progress in terms of monetary measures of well- being, human assets have improved. Starting from a low base, Ethiopia's enrolment expansion at all levels o f education has been impressive, with the number o f students in the first level o f primary school (grades 1-4) almost tripling since 1994. Yet substantial gender and regional differences remain. While progress was also recorded interms o f the prevalence o f pre-school stunting, malnutrition rates remain unacceptably high, imposing a substantial drag on the development o f the next generation and the future productivity o f the economy. Further progress in the health sector will continue to be slow if advances in water and sanitation are not rapidly made. Finally, note that while seemingly inconsistent, the difference inprogress between people's human assets and their well-being inmonetary terms follows from the time discrepancy, Le. the observed progress inhuman capabilities i s not only quite recent, it mainly concerns Ethiopia's children (primaryenrollment rates, child mortality, child malnutrition) and its effects on monetarypoverty will only be felt inthe future. 3.7 Third, women and pastoralists (40 million people in total) appear particularly disempowered-"The husband's beatingstick is like butter." The widespread acceptance of violence against women by women themselves epitomizes the deeply rooted existence o f 60 pronounced gender inequalities. Results from a nationally representative household survey conducted in by the Central Statistical Authority o f Ethiopia indicates that 85 percent o f women believe that a husband i s justified inbeating his wife for at least one o f the following reasons: burning the food, arguing, going out without telling, neglecting the children, and refusing sexual relations. The same survey reveals that 60 percent o f all women support female circumcision. 3.8 More broadly, the full effects of government reform and action to empower citizens have yet to be fully felt by citizens and sub-regional governments. The GoE has committed itself, through the Constitution and the SDPRP amongst other initiatives, to the empowerment o f citizens through decentralization. In addition, specific laws, policies and initiatives have been launched to address the position o f women and pastoralists. However, transitional processes have moved slowly, meaning that the potential effectiveness o f the GoE's efforts i s not yet fully realized. Partly as a result o f this, the strength o f informal, traditional practices persist at the community level, where resistance to change represents a tremendous impediment to opening opportunities particularly for traditionally disempowered people such as women and pastoralists. The shift from an historically deep-rooted political culture favoring a strong role for the federal government to devolving real power to local government and people i s difficult to bring about. As a result, opportunities for meaningful participation o f Ethiopian citizens in governance and political life to shape their lives are currently few. This i s further exacerbated by the few resources Ethiopian citizens can claim access to, raising serious concerns regardingthe extent to which they are able to take action to improve their own lives. 3.9 Immobility and isolation further limit people's capability to aspire and expand their choice set. Not only do Ethiopian citizens command few (material and human) resources, their sheer disconnect from the rest o f the world, helps perpetuate traditional practices and customs and substantially limits their opportunities to broaden their horizons. Eighty-five percent o f the rural population has been continuously residing in the same area (woreda) as they were born, i.e. without ever having lived elsewhere. Only 2.6 percent o f the total rural population were recent migrants, i.e. people who have moved into the area over the past 5 years. Further compounding their physical isolation i s their limited access to mass media, with only 13 percent o f the rural households owning a radio (virtually nobody has a TV), the lowest among 29 Sub Saharan African countries. Given that 85 percent o f the total population lives in rural areas, exposure to new ideas or outside influences i s clearly very limitedamong the Ethiopianpopulation. 3.10 Key areas for further analysis and data collection to improve the evaluation of people's well-beinginclude: 0 New methodologies to track poverty in a timely fashion. In light of the increasing focus on results and the Millennium Development Goals, it will be critical to monitor progress towards achieving these goals on a regular basis. Poverty numbers in Ethiopia prove highly sensitive to rainfall patterns and fluctuate significantly over time. Mere comparison o f a couple of points over time i s likely to strongly bias conclusions about poverty trends. It will thus be necessary to develop methodologies which are on the one hand sufficiently accurate to properly capture poverty incidence across the nation and the 61 major geographical areas, and on the other hand not too data intensive or technically demanding so as to ensure timely generation o f the results. Consumption prediction techniques developed under the poverty mapping projects combined with a systematized compilation o f regularly collected administrative and other secondary data as well as the welfare monitoring and/or Core Welfare Indicator Questionnaire surveys holdpromise. 0 A targeted, but comparable, household survey on coffee growers. Given the share o f households growing coffee, its importance for the economy as a whole both in terms o f export earnings and as a source o f growth through back and forward linkages96, and the recent collapse in coffee prices, it i s imperative that we better understand the dynamics surrounding the high poverty rate among coffee growers. More broadly, existing household surveys could be enhanced if they would allow for a better identification o f food and cash crop producers (preferably by type o f cash crop). 0 Systematic data collection and analysis of pastoralists' livelihoods. The dearth on systematic and comparable information on pastoralists renders it difficult to assess how disadvantaged pastoralists are compared to the rest o f the population. 0 In addition, poverty monitoringsystems need to broadentheir range of indicatorsto includeempowerment indicators. Empowerment is one o f the eight major thrusts o f the government's poverty reduction strategy, yet there i s an almost complete lack o f direct indicators o f empowerment. Inaddition, existing data on intermediary indicators of empowerment, i.e. assets and institutions, would benefit from analysis that sheds light on empowerment and its association with poverty reduction. One simple improvement in data could include disaggregation o f the population by social, ethnic or economic attributes. 0 The need for better documentation, dissemination and integration of existing data bases. While there i s a wealth of information available in Ethiopia both nationally representative as well as purposively sampled, the data i s often fragmented across different agencies and organizations and not systematically documented. Efforts to better document, disseminate and integrate these different data systems would be enormously useful to further refine the knowledge base o f Ethiopia's status o f poverty and well-being and its evolution over time. 3.11 Steps have already been taken to address some o f these data and knowledge gaps as discussed in Appendix 1, which describes Ethiopia's Action Plan to Strengthen SDPRS Monitoringand Evaluation. 96World Bank, 2004c. 62 Part11:Determinantsof MonetaryWell-Being and Poverty Having described the evolution o f well being in Ethiopia over the past fifteen years in Part I, the following chapters now seek to better understand the forces determining people's monetary well being. They further explore the potential and effectiveness o f different strategies to reduce people's vulnerability and lift them out o f (consumption) poverty. In doing so, the report addresses a specific request by the GoE related to this study, i.e. to strengthen the current empirical knowledge base, drawing on the Ethiopian experience, to help inform the ongoing revisions of the current strategic framework for poverty reduction (SDPRP). Part I1will also gauge the potential o f reaching the poverty MDGby considering different strategic scenarios. The conceptual framework used to address these questions is derived from the standard economic household According to this framework, reducing people's consumption poverty and vulnerability will require a combination o f (1) increasing their endowments, (2) enhancing the productivity or returns to their endowments, and (3) reducing the volatility o f those returns, or increasing people's ability to cope with it, and (4) increasing the security o f the endowments themselves. The HICES surveys and the agricultural sample surveys, augmented with secondary data, are used to explore how households in Ethiopia fare along these different dimensions. We beginby empirically describing the endowment base o f the Ethiopian population and the risk factors it faces, across time and space, followed by an analysis o f the returns to these endowments (Chapter 4). The importance o f geography for poverty reduction and the relation between risk management and poverty reduction in Ethiopia are also explored in detail. Given the importance o f agriculture for most people's livelihoods, Chapter 5 explores the performance o f the agricultural sector in Ethiopia. In particular, the determinants of agricultural productivity are examined and the potential for increasing agricultural incomes in different geographical areas i s explored. Using the empirical insights on the potential to increase agricultural income and the relative importance o f the different determinants o f monetary well-being more broadly from Chapters 4 and 5, Chapter 6 discusses the role o f agriculture, intersectoral linkages, and aid for further poverty reduction in a broader macro perspective, taking into account the structural nature o f poverty. It also uses micro simulations to explore the effectiveness o f different interventions (e.g. education, infrastructure, access to information, soil degradation) inreaching the poverty MDG. 97 In this approach, people are considered to be endowed with a wide array o f assets including physical capital (the amount and quality o f belongings such as land, trees, tools, machinery, etc.), human capital (the amount and quality of labor), financial capital (savings), and social capital. They live in environments with distinct agro- ecological (e.g. weather and soil characteristics) and socio-economic (e.g. infrastructure availability and population density) features and risk factors, which affect both the returns to and the security o f their endowments. Within these environments, and subject to these risks, people allocate their endowments across different activities which constitute their livelihoods (Ellis, 2000a), in order to maximize their income. The resulting income thus depends on the portfolio and the size o f an individual's endowments, as well as the security o f and retums to those endowments. Retums to people's endowments usually fluctuate due to natural (e.g. rainfall, diseases), economic (e.g. prices) or political (e.g. civil unrest, policies) factors. Yet, there is typically not a one-to-one mapping from income to consumption as people try to smooth their consumption over time either through depletion (and accumulation) o f savings or by engaging in formal or informal insurance schemes (e.g. iddir, iqub). 63 CHAPTER 4. ENDOWMENTS, RETURNS AND RISKS 4.1 This chapter begins with a detailed review o f the level o f people's private and public endowment bases, their distribution across space, and their evolution over time as well as the different risks people face. It subsequentlyestimates the returns to these private and public endowments, i.e. the relative importance o f each o f these factors in determining people's welfare, and explores how this differs across time and space. Particular attentionwill be paid to the effects of shocks, and relatedly households' ability to cope with them, on people's welfare. 4.1 People's Endowmentsand their Distributionin Ethiopia 4.2 Endowment levels seem equally distributed at strikingly low levels. To characterize people's resource base in Ethiopia, Table 4.1 and Table 4.2 present key statistics on the levels and distribution o f people's assets across time and space in Ethiopia. From looking across the tables it i s most striking that while assets appear to be equally distributed, people in Ethiopia have to earn their living from a very poor asset base, with households in urban areas only slightly better endowed than those in rural areas. Little progress has been seen in the private endowment base across the survey years 1995 and 1999, with the exception o f some improvements in access to public services and infrastructure. As was regularly expressed inthe "Ethiopia Consultations with the Poor" report o f 1999, inequality i s not considered an issue, as virtually everybody i s extremely poor.98 However, this commonly heard statement deserves to benuanced further, as will be illustratedbelow. 4.3 Looking across years, household size dropped from 6.01 to 5.90 persons per household, accompanied by an increase in the dependency ratio in rural areas (from 1.43 to 1.49) and a decline inurban areas (1.17 to 1.12). Further inspection shows that the number of children aged 5 years or less increased substantially inrural areas, while it dropped inurban areas, reflecting a decline in fertility99and possibly the effects o f HIV/AIDS. HIV/AIDS i s much more prevalent in urban areas and highest in absolute terms for both males and females in the 20-29 year old group'oo. The number of household members in the 16-25 year age group declined by 0.06 in both rural and urban areas, boding illfor the future labor supply. The number o f household members between 6 and 15 years old also declined substantially. There was only a marginal change in the number o f household members older than 25. Average age o f the household head remained around 45 years. Almost one-fifth o f households were headed by a woman, and female headship of the household i s about twice as common in urbanareas than inrural areas (33 percent o f all households versus 17percent). 98Rahmato and Kidanu, 1999. 99CSA and ORC Macro, 2001 looWorld Bank, 2004b. 64 Table 4.1: Endowment base and riskfactors of Ethiopian householdsin 1995 and 1999."'*' 1995 1999 25" 75" 25" 75" Mean Percentile Percentile Mean Percentile Percentile Privateendowments Humancapital/demographics Size o f household 6.01 4.00 7.00 5.90 4.00 7.00 Dependency ratio 1.39 0.67 2.00 1.44 0.75 2.00 Ratio o f females inhouse 0.50 0.38 0.60 0.51 0.40 0.63 Female head o f house 0.18 0.00 0.00 0.19 0.00 0.00 Age o f household head 44.9 35.0 55.00 44.5 34.0 53.0 # o f household members I5 yrs old 1.10 0.00 2.00 1.19 0.00 2.00 # of household members bw 6-15 yrs old 2.02 1.oo 3.OO 1.94 1.oo 3.00 # o f household members bw 16-25 2.00 yrs old 1.06 0.00 2.00 0.99 0.00 # o f household members bw 26-35 yrs old 0.63 0.00 1.oo 0.61 0.00 1.00 # o f household members bw 36-45 yrs old 0.54 0.00 1.oo 0.53 0.00 1.oo # o f household members greater than 0.00 45 yrs old 0.65 0.00 1.oo 0.63 1.oo Grade obtained by adult males 1.80 0.00 2.33 1.80 0.00 3.00 Grade obtained by adult females 0.82 0.00 0.00 0.88 0.00 0.00 N o o f adults completed post 0.00 secondary 0.01 0.00 0.00 0.03 0.00 Physical capital Land ownership (ha) 1.03 0.34 1.31 0.97 0.3 1 1.24 Own farm animal 0.54 0.00 1.oo 0.80 1.oo 1.oo Own transport animal 0.24 0.00 0.00 0.34 0.00 1.oo Ownplough 0.62 0.00 1.oo 0.63 0.00 1.oo Own bicycle 0.01 0.00 0.00 0.01 0.00 0.00 Own TVlRadio 0.16 0.00 0.00 0.21 0.00 0.00 N o toilet inhousehold 0.86 1.oo 1.oo 0.82 1.oo 1.oo Livelihoods Obtain some income from coffee (1=yes) 0.29 0.00 1.oo 0.25 0.00 1.oo Obtain some income from chat 0.00 0.00 (1=yes) 0.09 0.00 0.00 0.09 Share o f income from agriculture 0.74 0.64 0.98 0.65 0.52 0.89 Share o f income from wages 0.08 0.00 0.00 0.07 0.00 0.00 Share o f income from self 0.04 employment 0.13 0.00 0.09 0.08 0.00 Share o f income from other sources 0.05 0.01 0.06 0.19 0.08 0.24 Numbero f livelihood strategies 2.00 engaged in 1.19 1.oo 1.oo 1.42 1.oo Public endowments Electricity as source o f household 0.00 energy 0.10 0.00 0.00 0.10 0.00 Distance to food market (km) 6.41 2.00 9.00 5.35 2.00 7.00 Distance to water (km) 2.42 0.00 1.oo 0.77 0.00 1.oo Distance to health facility (km) 8.84 3.00 12.00 7.38 3.00 11.00 Distance to transport services (km) 17.18 2.00 20.00 16.23 2.00 19.00 Distance to dry weather road (km) 9.30 0.00 12.00 65 1995 1999 Mean 25" 75" 25" 75" Percentile Percentile Mean Percentile Percentile RiskfactorsIShocks Householdhead was sick inlast 2 months (l=ves) 0.26 0.00 1.oo 0.36 0.00 1.oo 2 year lagged rain shock3) 0.06 -0.05 0.16 0.02 -0.09 0.17 1year lagged rain shock3) (0.00) -0.14 0.13 0.10 0.00 0.20 Contemporary rain shock3) (-0.03) -0.14 0.07 (-0.03) -0.17 0.12 Percentage cultivated land area damaged4) 0.16 0.06 0.24 0.17 0.08 0.24 Real expenditure per adult equivalent inAddis 1995Drices 1439 865 1702 1432 884 1685 ~~~ ~ ~ ~~ ~~ ~ ~ Unless otherwisespecified, all results in parenthesesare not significant up to the 90 percent level. All 1995 and 1999means are otherwise significantly different beyond the 99 percent level. 2, All householdlevel variables are population weighted, while all community level variables are household weighted. ') All rain relatedvariables are Woveda level means. Shock is defined as percentage deviation Eromthe long run mean. 4, Crop damage concems only the rural population. Observationsare averages at the community or enumeration area level. 66 o o N o ~ o o o o o o o o o 0 0 0 0 0 0 9 ? 7 0 m 9 3 9 9 9 9 9 9 9 9 9 9 9 9 9 b O O O m O 0 0 0 0 0 0 0 0 0 0 0 0 0 o o r - m i - m 0 0 0 0 0 0 ? 9 9 \ 4 N O O N P - - 0 0 9 r - q - o o o o o c o m m m m v i -0 90 90 90 Y n ~ 9 9" "9 t - N C 0 0 o o o o o c O O O O N U 9 9 9 9 9 - - 0 0 0 0 c o m w o o o o o o o o o o o 9 ? " 9 b 9 9 9 9 9 9 9 9 9 0 0 0 0 0 0 9 9 9 9 9 9 0 0 9 0 0 c * ~ Q ~ O C b 0 0 0 m 0 - 0 0 0 0 0 0 0 o - o o o - o o o~ o o c m - * m m e N - r - 9 9 - o o o o o c 0 0 0 0 0 0 9 9 9 0 9 9 0 0 w O N W ~ 3 - 1 0 0 1 - 0 0 0 0 c~ " ~ 9 ? " 0 m 9 9 9 0 9 9 9 9 0 o r - w o o o o o o o o o o o ~ o o o m o - o o o o o o o - " - a m ? ' 4 N 9 9 9 0 0 0 0 0 0 4.4 Education levels are extremely low across the country, but slightly improving over the four year period. As indicated inPart I, educational achievement among Ethiopian adults i s disastrously low. With an average completed grade o f 1.8 among male and only 0.8 among female adults, Ethiopia i s basically left with an uneducated work force. While the results are better inurban areas and improving across the country especially for females, there remains a long way to go. Even in urban areas the average completed grade for male adults was only 5.5 years in 1999. Encouraging trends are also observed in post secondary education. While still virtually nobody completed post-secondary education inrural areas, the number of adults who completed some form o f post secondary education more than doubled inthe urban settings (from 2.7 percent o fall adults in 1995 to 6.1 percent in 1999). 4.5 Regardingland size, one of the most important assets a rural householdhas, on average each holder has about one ha of land under temporary and permanent crops, which translates to 0.21 ha per rural person.Io2 This has declined from 0.5 ha per person in the 1 9 6 0 ~ .At~ ~0.11 ha the available land per person seems especially low in SNNPR and ' Gambela (Table 4.3). Landholdings per person are also lower in the food insecure zones in line with common perceptions o fhigher land scarcity inthese areas. . 4.6 A land poor class is emerging. Following the radical land refodredistribution o f 1975 under the Derg regime, Ethiopian agriculture has basically become a smallholder farm sector, with only a limited number o f large scale commercial/state farms. Indeed less than one percent o f land holders had more than 4.35 ha per holder, and the largest holding measured was only 85 ha. As land sales are forbidden,Io4 equal access to land i s commonly assumed to characterize Ethiopian agriculture. However, as illustrated in Table 4.3, land appears to be much more unequally distributed than commonly assumed, even though inequality does not arise from large land ownership. The Gini coefficient among those with land i s estimated at about 0.45.Io5 Further disaggregation o f the data shows that 20 percent o f the rural population has less than 0.06 ha per person, or 0.08 ha per person when SNNPR i s excluded-holders inSNNPRhave smaller plots and tend to grow mainly coffee and enset.Io6 To fix ideas, 0.08 ha corresponds to a plot o f 20 m by 40 m, which with current cereal production technologies would yield on average a daily equivalent o f about 779 calories per I O 2Figures obtained from the CSA Annual Agricultural Sample Surveys. The figure excludes landless households, though including landless households would not affect the results much, since about 98 percent o f rural households reported to "own" land in the 1999 WMS. Strictly speaking land in Ethiopia belongs to the state and farmers only have user rights. Land tenure security and private land ownership are longstanding and hotly debated issues in Ethiopia. The effect o f lack o f private land ownership on agricultural productivity and environmental degradation will be extensively discussed inthe ESW on Poverty and the Environment. Among a selected sample o f Sub Saharan African countries, Jayne et al., 2003 calculate, based on FA0 statistics, that the cultivated landperson ratio (=land cultivated under permanent and annual crops/population in agriculture) declined by 50 percent between the 1960s and the 1990s. For Ethiopia, they find the cultivated landperson ratio to decline from 0.508 to 0.252, which closely resembles our estimate for the late 1990s derived from household survey data. I O 4The state owns the land and farmers only hold usufruct rights. Land transfers through short term rent or contract are however possible, though relatively limited. This is very similar to the Gini coefficient o f land size per holder among those with land, which is estimated at 0.462 in 1995 and 0.477 in 1999. Enset, or false banana, i s a perennial tuber crop. 69 person, or slightly more than half the person's daily cereal caloric needs.lo7 Further analysis suggests that households inthe bottom uartile o f landper person still derive a substantial part o f their income from crop production.'" A "landless" class is rapidly emerging inEthiopia. Table 4.3: Land holdings per person and land inequality Total Area Cultivated per Capita per Holder') Average Gini coefficient*) 1995 1999 1995 1999 Administrative region Tigray 0.20 0.23 0.45 0.43 Amhara 0.26 0.26 0.42 0.46 Oromiya 0.24 0.23 0.43 0.45 Benishangul-Gumuz 0.27 0.34 0.40 0.49 SNNPR 0.11 0.11 0.43 0.47 Gambela 0.16 0.17 0.50 0.47 Food security food secure zones 0.21 0.22 0.47 0.47 Mediumfood secure zones 0.23 0.22 0.43 0.47 Food insecure zones 0.16 0.18 0.45 0.48 Total 0.21 0.21 0.45 0.47 Only land under temporary and permanent crops i s considered. Permanentcrops make up about 7 percent of the total land under temporary andpermanentcrops. Temporary and permanent crops make up about 85 percent of total cultivated land. The other 15 percentare usedfor grazing, fallow, wood land and others. 2, Gini i s among those with land, yet 97 percent and 98 percent of the households inrural Ethiopia reported to own land in 1995 and 1999 respectively. Source: Own calculationsfrom 1995 and 1999Agricultural Sample Surveys, CSA 4.7 Several factors explain inequality in land access, including geography, life cycle effects, gender of household head, and household composition. Analysis o f variance suggests that inequality in land access cannot be attributed to large differences across provinces or zones. Inparticular, only three and eight percent o f the variation in land access across the country could be explained by between-region or between-zonal differences respectively. There appear to be larger differences in land size across woredas which explain about 21 percent o f the variation in land size. Other important factors include life cycle lo' see To this, note that one kg of cereal flour yields between3,500 and 3,800 calories dependingon the cereal, with teff at the lower endof the range and maize and sorghumtypically at the upper end. Given 15 to 20 percent loss during the milling or pounding process, we assumed that one kg of cereals yielded about 3,000 calories. Given average cereal yields of about 1.2 ton per ha, 0.08 ha per person would generate 779 calories per person per day [0.08(ha/person)* 1,20O(kg/ha)*3,000 (calikg)]/365=779]. Fromthe 1999 HICES survey we know that the pooresthalf of the Ethiopianpopulation obtains about two thirds of calories from cereals. Givena minimum daily requirement of 2,200 calories per adult, this implies that daily 1,466 calories per adult are needed from cereals, which correspondsto about 0.5 kg of (unmilled)cereals. `Os Jayne et al. (2003) estimated that crop income share for households in the bottom quartile in land per capita was still 86 percent, with only 14 percent of income derived from non-farm income. Their estimates are based on a sub sample of the 1996 household budget survey for which they collected more information. Similarly, they observed crop income shares ranging from 57 percent in Zambia to 79 percent in Mozambique among households inthe bottom quartile inlandper capita. Only inKenya were those with little land mainly engaged inother remunerativeoff-farm activities. 70 effects, gender of the household head, and the number o fmale and female adults. Households with heads that are 50 years of age have roughly 0.025 hectares more land per household member than those with heads 30 years o f age. Female headed households have on average 0.07 hectares less land per household member than male headed households. This translates into one third less land per person when compared to the national average of 0.21 hectares per person, While households with more male and female adults tend to have more land per household, they also have less land per household member.'09 The average plot size, the amount o f calories it can generate, the degree o f land inequality and the correlates o f land access are important stylized facts to gauge the merits o f the Agricultural Development-Led Industrialization (ADLI) strategy. We retum to the implications o f these facts in Chapter 6, where we discuss the potential for ADLI interms of poverty reduction inmore depth. 4.8 Average livestock ownership, another key household asset, has largely remained constant among rural (agricultural) households across the two survey years (Table 4.4). Rural households engaged in temporary and permanent crop production have on average slightly more than one ox, almost three unitsof cattle, and two to three sheep and goats. Oxen are critically important in the ox-plough farming system which characterizes Ethiopian agriculture, and they must be used inpairs. However, only 29 percent o f all households with oxen possess two or more oxen (not reported in the table). Moreover, the proportion with at least one ox has declined from 54 percent in 1995 to 42 percent in 1999. Similarly, the proportion o f households possessing any large or small ruminants declined. Slightly more than one quarter of the population owns a transport animal. This corresponds to the 1995 figure reported in Table 4.4, which is derived from the WMS."' Regarding farm equipment, about 70 percent o frural households own their own plough (Table 4.2). Table 4.4: Livestock ownership in Ethiopia between 1995 and 1999.'' 1995 1999 Number per holder Average Percentage o f Average Percentage o f holders holders cattle for agricultural purposes, oxen 1.15 54.00 1.08 42.2 cattle for other purposes 2.85 66.10 2.88 59.5 sheep and goats 2.85 49.20 2.36 38.9 horses, mules, asses, camels 0.55 28.60 0.58 26.8 Note: Figures do not include pastoralist households that do not cultivate land. Source: Own calculationsfrom 1995and 1999Agricultural Sample Surveys, CSA 4.9 Ownership of consumer durables such as bicycles and radios/TVs has increased, both in rural and urban areas (Tables 4.1 and 4.2). Nonetheless, still only 14 percent o fthe population in rural Ethiopia has a radio (or TV), exemplifying the sheer isolation o f the Ethiopian countryside, not only in terms of market access but especially in terms o f information. Sanitation conditions improved inboth urban and rural areas, though again from very low levels. The proportion o f households using buckets or their surroundings as their toilet declined from 86 percent in 1995 to 82 percent in 1999. Once again the difference 109J a p e et al., 2003. Note that the difference inownership o f farm and transport animals reported inTables 4.1 and 4.4 is related to definitional issues. 71 between urban and rural areas i s stark, even though mere population density also encourages increased use o f toilets. 4.10 Households in rural areas derived on average about 85 percent of their income from agriculture in 1995, which dropped to about 75 percent in 1999. These figures highlight once again the complete dependence o f the Ethiopian countryside on agriculture. About 29 percent were coffee growers (which fell slightly to 25 percent in 1999) and about ninepercent ofhouseholds obtained some income from growing chat, a mildnarcotic. About one third o f households in urban areas were wage earners, while another 35 to 45 percent were self employed. The remainder earned their living from agriculture and other sources. The number o f livelihood strategies' '' increased overall from 1.19 to 1.42 in rural and urban areas alike, most likely inresponse to deteriorating living conditions. 4.11 Access to public endowments, includingelectricity and transportation networks, i s extremely limited in rural areas, and distances to markets and public services are increasingfor urban residents. The number of households that usedelectricity increasedby six percentage points inurbanareas from 65 to 71 percent, but at one percent use o f electricity remains virtually non-existent inrural areas. All distance measures indicate increased access to services in rural areas. This is critical as it reduces transaction costs and increases households' connectivity to the overall economy, through which they can reap the benefits from overall economic growth as well as the benefits that accrue from agglomeration and information externalities. Nonetheless, extreme remoteness continues to characterize the countryside, with food markets on average six kilometers away, a dry weather road almost 10 kilometers away, and public transport on average 18 kilometers away. The average distance to drinking water, however, dropped from 2.83 to 0.85 kilometers. Incontrast, inurban areas distances to food markets, water sources, health centers and transportation services all increased by at least 60 percent between the survey years. As indicated before, this could be the consequence o f increasing rural to urban migration, with the new settlements being built further and further way from the main hubs o f activity. This is consistent with the observed increase inurban poverty. 4.12 Rainfall and illness are critical risk factors. Rainfall forms a critical, but highly variable, input for agricultural production and thus rural income generation. Inthe regression analysis in the following section, we capture rainfall shocks using the percentage deviation from the long run average in a given year for a woreda. For each survey year, we have information for rain shocks for that year (1995, 1999), which we call contemporary rain shocks, a one year lagged rain shock (1994, 1998), and a two year lagged rain shock (1993, 1997). However, no major nation-wide drought occurred during any o f these years (Table 4.1 above), and shocks were mostly positive. Rainfall was especially plentiful in 1998, with 10 percent more than the long term average. Nonetheless, the nation-wide averages also mask quite large differences across the country, with some woredas experiencing serious shocks even in years with moderately good average rainfall, as illustrated by the 25th percentile rainfall shock o f 14 to 17 percent in 1995 and 1999 respectively. We also use the percentage o f cultivated land area damaged as reported by the farmers themselves, which captures both ''IA household i s considered to engage ina particular livelhood strategy ifit obtains more than 20 percent of its income in that activity. 72 idiosyncratic (insectdpests) as well as covariant (droughts and floods) shocks. The total area damaged i s fairly steady at around 16 percent in both years. Finally, we consider health shocks proxied by self reported illness o f the household head over the past two months. Self reported sickness incidence increased substantially between 1995 and 1999 from 26 to 36 percent . 4.13 Woreda level measures of soil degradation"', agro-ecology, population density and malaria risk are used to characterize geography inthe regression analysis discussed in the following section. Rural populationper arable landi s on average 232, ranging from 90 to 281 and over two-fifths o f the population are exposed to high malaria risk (Table 4.5). We include two measures o f soil degradation. First, using information on the slope and characteristics o f the soil, the Biomass project estimated average annual topsoil loss per woreda. The depth o f the topsoil i s an important measure o f the moisture retention capacity o f the soil. At less than 1.2 m o f topsoil, moisture retention capacity rapidly declines. Average annual topsoil loss per year amounts to 14 tons, which corresponds to about 1.12 mm of topsoil per year. One quarter o f woredas lose 1.6 mm o f topsoil annually. 4.14 Nutrient loss due to dung collection appears to be significant. To further proxy nutrient loss, we include an estimate o f the total phosphorus and nitrogen loss due to dung (and crop residue) collection from the fields expressed in DAP and Urea equivalent tons per ~ 0 r e d a . l ' ~On average, 185 ton DAP equivalent and 624 ton Urea equivalent i s removed from the fields per woreda due to dung (and crop residue) collection. In one quarter o f the woredas the amount o f DAP and Urea equivalent removedper year amounts to more than 300 and 1,000 tons respectively. Overall, the total nutrient loss due to dung collection in the country would result in a total DAP and Urea equivalent nutrient loss o f 364,050 tons."4 This i s astronomical, especially when compared with the total amount o f DAP and Urea annually applied to the fields by all farmers across the country, which in 2000 amounted to 341,000 tons. Yet, it must be emphasized that there i s not at all a one-to-one mapping o f the phosphorus and nitrogen content o f dung (and crop residue) lost due to their removal from the fields and actual soil nutrient Nonetheless, the magnitudes are sufficiently large to suggest significant detrimental effects to agricultural productivity and thus income and poverty. Further investigation o f this matter inthe Ethiopian context i s called for. With only slightly more than one quarter of all cultivated land fertilized with commercial fertilizer, fertilizer use in Ethiopia remains limited. The use o f natural fertilizer appears even less (only 10 percent o f total cultivated land). ~~ ''' DataA 'I2 have been collected by the Biomass project. D P i s diammoniumphosphate and urea is carbonyl diamide, which provide phosphor and nitrogen respectively, the two key soil macro-nutrients. 'I4 450 woredas*( 184 ton DAP/woreda+624Urea lodworeda) yields 364,050 ton DAP and Urea equivalent nutrient loss. 'I5 The extent to which dung (and crop residue) can enrich the nutrient base o f soil critically depends on its composting, the timing o f the application and the cultivation practice used to turn these materials in the soil. Roose and Barthbs (2001) indicate that "the commonly used dry manure presents poor qualities, has lost most o f its nitrogen and potassium, and carries pests, germs and weed seeds as faeces are not heated up sufficiently to kill these contaminants. Good quality manure is rare in Africa, but its positive influence on yields, its slow release o f nutrients and positive effects on pH and other soil properties are well documented (FAO, 1975, Shaxson, 1999). However, 40-60 percent o f the carbon and 30-50 percent of the nutrients from the grazed biomass do not returnto the soil (Roose, 1996)". 73 Table 4.5: Parameters of soil degradation, agro-ecology, population density and malaria incidence in Ethiopia" Mean 25" 75" Percentile Percentile Soil degradation D A P equiv. nutrient loss (tonsiyrlworeda) due to dung and crop residue collection 185.07 6.36 305.01 Urea equiv. nutrient loss (tonsiyrlworeda) due to dungand crop residue collection 624.06 13.14 1016.26 Average soil loss (tonslhaiyr) 14.36 7.39 19.90 % o f cultivated land fertilized naturally in 1999 0.10 0.00 0.14 % o f cultivated land fertilized by chemicals in 1999 0.27 0.00 0.50 Agro-ecological characteristics Mean Altitude (m) 1935.84 1664.00 2200.00 Long RunAverage Rainfall (mm) 1085.69 804.33 1282.75 Long RunCoefficient o f Variation o f Rain 0.24 0.16 0.29 Rural Population per arable area (people per km') 232.41 89.81 280.79 Malaria risk (proportion) HighMalaria Risk(l=yes) 0.45 0.00 1.oo MediumMalaria Risk (1=yes) 0.11 0.00 0.00 L o w Malaria Risk (l=yes) 0.44 0.00 1.oo All variables are at the woreda level. 4.15 Overall, most households use firewood, though about 16 percent of the households use dung cakes, crop residues and saw dust as their primary source of cookingfuel. Only five percent use charcoal, kerosene, gas and electricity andvirtually all of these reside in urban areas. Looking across the regions, about twenty-five percent o f all households in Amhara use dung cakes (and crop residues) which, combined with high population density, results in large annual DAP equivalent nutrient losses (Table 4.6). While the proportion o f households using dung cakes and crop residues is even higher in Tigray, total DAP equivalent loss per woreda i s not so high, as Tigray i s much less densely populated. On the other hand, in S W R , where more than 90 percent o f households rely on firewood as opposed to dung cakes, high levels o f population density still result in above average rates o f DAP nutrient loss per woreda. Use o f dung cakes is also higher in the food insecure areas (Table 4.6), resulting in larger nutrient loss, further lowering labor productivity per capita in the face o f less productive landholdings."6 'I6 Pender et al. (2001) observe a robust association o f decreased use o f manure as fertilizer with population growth in Ethiopia, and attribute this finding to the increased demand for manure as fuel. 74 Table 4.6: Biomass use for cooking fuel')*) 1995 1999 Firewood Charcoal, Dung cake, Firewood Charcoal, Dungcake, Proportion of (collected Kerosene crop (collected Kerosene crop households using and Gas, and Residues, and and Gas, and Residues, and purchased) Electricity Saw Dust purchased) Electricity Saw Dust Tigray 55.65 2.25 34.13 63.86 1.78 31.11 Amhara 66.33 1.20 25.31 71.16 0.92 26.01 Oromiya 77.88 1.83 16.58 77.10 2.26 16.01 SNNPR 90.57 0.36 4.44 93.36 1.21 2.52 Rural 75.88 0.23 18.65 78.92 0.25 17.31 Urban 59.92 29.15 7.10 58.48 33.39 6.62 Food Secure zones 83.42 1.06 9.92 85.54 1.19 8.93 Between 68.91 1.52 24.24 72.40 2.23 22.43 Food Insecure zones 73.84 3.71 17.40 75.70 3.68 18.13 Total 73.50 4.54 16.93 76.13 4.77 15.85 ') The rows don't add up to 100, the difference being "other" sources than the ones the reported. 2, Zones have been divided interciles based on a food security index which captures the average proportion o f people estimated inneed o f food aid in 1994 and 1995. 4.16 We explorethe differencesin endowments across administrativeregionsinTables 4.7 and 4.8. We focus on the four most populous regions for which we have sufficient data: Tigray, Amhara, Oromiya and SNNPR. While there i s a large variation inmean expenditure across the regions, systematic variations in endowments are limited. Some o f the more striking differences include the wide range among mean adult male education, with adult males in SNNPR having almost twice as many grades completed as their counterparts in Amhara (2.04 versus 1.08). Compared to the other regions, households in SNNPR have on average the lowest ownership o f almost all assets (including land: see Table 4.7) but greater access to services. Across the regions, households derive about 70 percent of their income from agriculture. Coffee production i s especially important in SNNPR (59 percent o f all households obtain income from coffee), but also in Oromiya where about a quarter o f all households are engaged in coffee production. Among the four regions, chat production i s most prevalent in Oromiya. During our survey years, the crop damage rate was about twice as highinSNNPR compared to Tigray. 75 x m 0 0 8 9 0 0 s 0 0 N 0 9 e e 8 8 - 8 2 8 In 0 x 0 08 s 0 0 0 - 0 0 u. 9 0 0 9 m 8 8 9 8 0 2 8 0 E x x 0 - 2 G. 0 v i z N W v ) i D 6 " Q1 Y r: 2 0 r: m E 2 8 8d 8 d 0 0 0 0 0 0 0 0 x 0 0 0 8 2 0 8 0 8 0 0 0 0 0 0 9 9 9 9 9 9 0 0 0 0 10 9 - - - 0 0 - e 8 2 8 8 v) m 2 80 8 0 0 0 0 m 0 0 2 8 8 s - 9 0 8 3 8 8 sfi 0 2 Do r- G. b N W - 0 W ? r : ' ? 9 ? b m - m r: W 2 0 2 8 0 0 0 0 0 x 2 0 0 8 0 x m 0 0 8 s 0 8 0 8 0 9 0 0 W 8 8 2 0 9 0 8 vi m - '? 0 0 0 0 0 9 0 0 r: 0 0 0 8 z m 8 8 8 8 0 8 8 - r: M 2 '0 Y W m e N v) v) G. 0 vi Y 0: 8 8 8 d 0 8 0 0 0 2r- 0 0 9 - e 0 8 0 8 s - 0 0 d 9 8 2 8 8 0 0 10 0 0 0 0 7 0 0 0 9 0 0 r? Y 0 0 0 8 2m 8 8 c 8 8 0 8 8 /1 N P- - 2 2 ? N m 2 8 2 vi N r- m d 8 8 3 8r- 8 E E e E 9x c! - \o x m m 8 N vi md c. 09 m 0 m 2 8 C ? d 9 r. 2 vi m ? N 0 01 m zc? 2 C 9 N z e C 8 0 8 0 8 W r? P- Lo m P- N 0r. m x -2 - 4? C 9 0 d '0 m R;; E: 4 C m r. 8 x C -E: r. m M N N 2 -2 N 9 vi 0: C N d C -c vi r! ? f0; C vi N 8 m - '4 c d m -sr, 2 5 4.2 Returnsto EndowmentsAcross Time andSpace 4.17 We now investigate the magnitude and precision o f the retums to the various endowments discussed above, with the intention o f shedding light on the relative importance o f these endowments for consumption and poverty reduction. We further examine how the retums differ across time and space. The dependent variable usedinall the estimated results presented is the natural logarithm o f per adult equivalent expenditures. The reported coefficients thus represent a percentage change in expenditures given a marginal unit change inthe independent variable of interest. All expenditures have beendeflated across space and time and are expressed inAddis Ababa 1995 prices. 4.2.1 Returnsacross time 4.18 We first estimate the effect o f private and public endowments on consumption. Table 4.9 presents the data, which are described in greater detail in this section. Inthe first column o f Table 4.9, we present the results for our pooled sample, where we combine observations from both our 1995 and 1999 data. Inthe second and third column we decompose the pooled regression across years to get a sense o f the change in retums across time. The fourth regression includes additional information on access to dry weather roads and enumeration area (EA) level measures on land availability and cultivation practices only available in 1999. We control for all other location-specific characteristics (e.g. differential prices, development interventions, government effectiveness, ethnicity, etc.) by including woreda dummies. Inthe next section we will hrther unbundle the effect of geography on consumption and poverty through explicit inclusion o f measures o f agro-ecological potential, population density, soil degradation and spatially correlated risk factors. Table 4.9: Estimated effects of household and public endowments on ons sump ti on.')'^''^' Logarithm o f per adult Pooled 19951992' 1995 1999 Extended 99" equivalent consumption Coef- Coef- ficient. t-~tat.~' Coef- ficient . t-stat. ficient. Coef- t-stat. ficient. t-stat Private endowments Human CapitallDemographics Size o f household -0.0969 -61.34 -0.0938 -40.97 -0.103 1 -50.16 -0.1001 -32 Dependency ratio -0.0182 -5.87 -0.0122 -2.64 -0.0232 -5.92 -0.0209 -3.71 Ratio o f Females inhouse 0.1673 12.37 0.1801 8.8 0.1497 8.94 0.1495 6.12 Female head of house -0.0082 -1.12 -0.0253 -2.28 -0.0051 -0.56 0.0166 1.23 Age -0.0078 -7.43 -0.0123 -7.77 -0.0016 -1.25 -0.0002 -0.13 Age squared 0.0001 5.63 0.0001 6.49 -3.57E-06 -0.27 -1.49E-05 -0.77 Mean education of adult males 0.0158 14.62 0.0147 8.81 0.0162 12.27 0.0159 7.34 Meaneducation of adult females 0.0198 13.93 0.0250 11.36 0.0163 9.37 0.0149 4.53 No o f adults completed post secondary 0.1910 12.26 0.2095 7.04 0.2125 12.39 0.2589 4.06 Physical capital Community average land size (ha) I holder 1.20E-05 5.81 No toilet inhousehold -0.1232 -12.41 -0.1143 -7.02 -0.1360 -11.35 -0.1033 -5.42 79 Logarithm o f per adult Pooled 19951992) 1995 1999 Extended 9g3' equivalent consumption Coef- ficient. Coef- t-~tat.~) Coef- ficient. t-stat. ficient. t-stat. ficient. Coef- t-stat Own plough 0.0963 12.76 0.0842 6.8 0.0820 8.95 0.0689 5.4 Own farm animal 0.0605 8.13 0.0932 8.05 0.0507 5.06 0.0476 3.21 Own transport animal 0.1227 17.63 0.0781 6.81 0.1390 16.41 0.1334 11.49 Ownbicycle 0.1247 4.34 0.0602 1.36 0.2042 5.79 0.2292 3.1 Own TVlRadio 0.2082 24.56 0.2007 14.67 0.2122 21.08 0.2027 13.16 Livelihoods Obtain some income from coffee (l=yes)) 0.0253 3.15 0.0170 1.45 0.0299 2.7 0.0421 2.68 Obtain some income from chat (l=yes) 0.0945 7.16 0.1376 6.8 0.0811 4.86 0.0899 3.79 Share o f income from agriculture 0.0790 5.02 0.0467 2.15 0.1487 6.62 0.0315 0.75 Share o f income from wages -0.0292 -1.82 -0.0316 -1.38 -0.0506 -2.34 0.0373 0.66 Share o f income from other sources -0.3020 -14.78 -0.3193 -8.27 -0.2619 -10.86 -0.5780 -11.35 Number o f livelihood strategies engaged in -0.0777 -13.15 -0.0482 -4.66 -0.0851 -12.51 -0.0860 -8.25 Public endowments Electricity as source o f household energy 0.0607 3.77 0.0682 2.64 0.0668 3.32 0.0052 0.1 Distance to food market (km) -0.0017 -2.43 -0.0006 -0.51 -0.0034 -3.33 -0.0027 -1.78 Distance to water (km) -0.0028 -5.63 -0.0008 -1.06 0.0022 1.37 0.0053 2.14 Distance to health facility (W 0.0017 3.59 0.0015 2.01 0.0029 4.35 0.0032 3.38 Distance to transport services (km) -0.0018 -8.17 -0.0021 -6.19 -0.0002 -0.44 0.0002 0.45 Distance to dry weather road (km) -0.0016 -3 Cultivationpractices % o f cultivated land fertilized naturally 0.4692 3.23 % o f cultivated land fertilized naturally) squared -0.6149 -2.9 % o f cultivated land fertilized chemically 0.3935 3.26 (% o f cultivated land fertilized chemically) squared -0.4144 -3.17 Urban 0.1403 8.14 0.1834 6.84 0.1242 5.51 Year (=1 if year==1999) 0.0233 3.41 Constant 7.7939 253.34 7.8398 167.38 7.6816 195.87 7.5660 109.55 Number o f observations 27,081 11,273 15,808 6,422 R-squared 0.4508 0.5243 0.4935 0.5313 `) Woreda Fixed Effect Model. Coefficients on woreda dummies not included. 2, Inthis model, we poolthe observations for boththe 1995 and 1999 samples. 3, As we only have information on mean land size and fertilizer variables inthe rural areas in 1999, this model i s limited to only rural areas in 1999. Land size and fertilizer use are community level averages. 4, As a rule o f thumb, the coefficient i s statistically different from zero if the absolute value o f the t-stat exceeds 1.96. 80 4.19 As is commonly observed, larger households and households with more dependents tend to be poorer. This relationship is robust across the different specifications. The relation between population growth and poverty at both the macro and micro level is explored inmore depth in ongoing World Bank ESW on poverty and population. Somewhat more surprising i s the large and significant relationship between the ratio o f females in the house and consumption. However, this result i s largely drivenby the fact that the dependent variable i s in terms of per adult equivalents, which by design results inwomen placing less o f a burden on household expenditure as men o f equal age."' We do not find the regular lifecycle effects whereby households with older heads tend to be better off. This may be related to the fact that we already control for land holdings and that part o f the lifecycle effects work through increasing landholdings as discussed in previous section. Note furthermore, that while the lifecycle effects are found to be negative in 1995, they disappear altogether in 1999. 4.20 Urban (but maybe also rural) female-headed households appear worse off. While female-headed households tend to be poorer according to the 1995 results, this i s not confirmed by the 1999 results. When controlling for land size, female households in rural areas may even be slightly better off. Given that we already control for many important determinants o f consumption, which are also correlated with the headship o f the household (household size, dependency ratio, age, education o f household members, ownership o f land and other assets), it may not come as a surprise to not find a clear-cut effect o f the gender of the headship per se on consumption.l1 Analysis o f the endowment base o fmale- and female- * headed households suggests that educational endowments are substantially less among female headed ho~seholds."~Given the high returns to education, as indicated below, the lack o f educational endowment implies a significant handicap for female-headed households. Female-headed households have less farm equipment, are less engaged in cash crop production and earn more o f their income from off-farm activities, which appear less remunerative, as illustrated below. Further disaggregation by space shows that female-headed households are especially worse off in urban areas, where their consumption tends to be nine percent lower, though not in rural areas. This difference may be a result o f the breakdown o f traditional social support systems in urban settings.'*' We also observe some differences across the regions, with female-headed households seemingly slightly better o f f in SNNPR and potentially slightly worse off in Oromiya. 'I7 Running a similar test on per capita expenditures results in a loss o f significance in 1995 and a reversal o f signin 1999. Quisumbing,Haddad, andPena, 2001. ' I 9While the female adults in female-headed households are on average better educated than the female adults in male-headed households (1.23 versus 0.77 grades), female adults attained o n average 0.6 grades less than the male adults in the male-headed households. Also, those few male adults present in the female-headed households are also less educated than their male counterparts in male-headed households. '*'Among rural households in the Kilimanjaro Region o f Tanzania, female-headed households were found to be especially likely to receive transfers from family members in times o f need; in urban settings these households, while still needy, may receive less such support.(Christiaensen, Hoffmann, and Sarris, 2004). 81 4.21 Marginal returns to education are positive and high, both for male and female adults.12' They also tend to be higher for female than for male adults, especially in urban areas. In particular, increasing average male adult education in the household by one grade would enhance per adult equivalent consumption by 1.6 percent (Table 4.9, col. l), while an additional year o f education among female adults in the household would increase per adult equivalent consumption by two percent. The magnitude o f this result i s simply astounding. While these findings would be very misleading if they were mainly driven by substantially higher returns to secondary schooling, further decomposition shows that this is not the case. When we include primary and secondary schooling separately in the regression analysis (results not reported in table), we still find that one additional grade o f primary schooling for male adults would increase consumption by 1.4 percent, while one additional grade o f secondary schooling would increase consumption by 1.8 percent. When increasing primary and secondary schooling among female adult women by one grade, average consumption would go up by 1.7 and 2.1 percent respectively. 4.22 To better appreciate the magnitudes o f these effects, note that over the past 12 years GDP in Ethiopia grew on average at 1.7 percent per capita. Given that, it follows that increasing average female adult primary education by one grade from its currently low level o f grade completion o f 0.8 grades would yield the same effect on consumption as one year o f economic growth at its current pace. Post-secondary education also has a great effect on expenditure, though the retums are much higher in urban areas. In 1999, for each extra household adult having completed a post-secondary education, expected household expenditure increased by an average o f about 20 percent. This result i s robust across all specifications. In sum, at average adult grade completion rates o f 0.8 for female adults and 1.8 for male adults, promoting primary school enrollment as well as adult literacy will clearly have to be at the center o f any poverty reduction policy. Chapter 10 will explore further which are the morepromising interventions to enhance primary school enrollments. 4.23 In linewith our expectations,the larger a household's landholdings,the higher its welfare. Controlling for all other individual, household, and community characteristics as well as geographic conditions, the estimated effect of landholdings on consumption suggests that households which are at the 75th percentile interms o f their landholdings (corresponding to 1.30 hdholder) are on average 12 percent richer than those at the 25th percentile (who have 0.32 hdholder).'22 This raises a series o f important policy questions. On the one hand, these results would lend some support to a focus on fostering labor mobility, either to other unexploited areas (agricultural extensification--the government's current resettlement policy only beingone variant o f that), or out o f agriculture. Yet, the evidence could also be taken to support the need for concerted efforts to raise labor productivity in agriculture (agricultural `'I It i s noted that there i s no issue o f reverse causality (Le. consumption levels affecting the observed education levels) because it concerns the effect o f education o f the adults in the household, and not the effect o f the education o f the children. While it cannot be excluded that the education variables may pick up unobserved individual characteristics-the more entrepreneurial and savvy individuals are also those with education-it i s unlikely to play an important role given the importance o f supply side factors (distance to schools) in determining chddren's enrollment (see Chapter 9). It could be safely assumed that the correlation between the '''lacementexistence o f schools and people's innate characteristics is very tenuous indeed. The o f potential non-linearities in the consumption-land holding relationship suggested by the bi- variate evidence inJayne et al. (2003) needs to be further explored. 82 intensification). This in turn would increase real incomes and generate demand for locally produced goods and services which in turn would generate off-farm employment opportunities for the emerging landless class. 4.24 The estimated results regardingthe use of chemicalfertilizer suggest that there is still some scope in terms of economic growth and poverty reduction from increased fertilizer use, also in food insecure areas. Chemical fertilizers are currently only usedon 13 percent o f the total cultivated area in food insecure areas (see Table 5.1). Raising the total cultivated areas fertilized with commercial fertilizer in these areas to the level in the food secure areas (32 percent) would increase consumption per adult equivalent by four percent. While the results do only represent an average effect identified from a single cross-section, the estimated coefficients on commercial fertilizer use appear suggestive after taking into account several consideration^.'^^ We control for differences in soil quality and rainfall which may also affect the effectiveness o f fertilizer use by using woreda fixed effects. It is possible that the effects o f fertilizer use are slightly overestimated as they may also pick up the positive effects o f improved seed use, for which we do not control. Yet, the bias is likely to be small given that the use o f improved seeds i s still extremely low and largely limited to two cereals (maize and wheat). We retum to the critical importance o f using combined improved seed-fertilizer packages to maximize the effect o f technology adoption in Chapter 5. Also, some crops are more responsive to fertilizer, which i s consistent with the estimated declining marginal returns on commercial fertilizer use beyond 47 percent o f the total cultivated area at which expenditures peak (see Figure 4.1). Current application rates are usually less than the recommended amounts, so it i s not improbable that the estimated effects represent lower bounds. Potential contamination from individual unobserved effects such as individual managerial capabilities and entrepreneurship is avoided through the use o f average fertilizer use per holder per community. We also control for the education level o f the adults inthe household. 24 4.25 Insum, the estimated returns to increased fertilizer use suggest that there is still scope for agricultural intensification through increased use o f modern inputs, also in food insecure areas. It i s thus imperative to better understand the current constraints to further technology adoption including the farmers' incentive structures to adopt fertilizer and the effectiveness o f the current delivery mechanism^.'^' Anecdotal evidence also suggests that the demand for fertilizer by poorer households may be limited because o f the downside risks involved in case '23The effect o f fertilizer on agricultural income and consumption depends on a series o f other factors, such as soil characteristics, adequate and timely rainfall, or food prices. Furthermore, commercial fertilizers tend to be more effective when used in combination with improved seeds, and certain crops are much more responsive to fertilizer use than others. Fertilizer users also tend to be more productive and entrepreneurial farmers. The reported estimates on the effect o f fertilizer may be affected by these different factors and thus not picking up the effect o f fertilizer alone, but rather the combined effect o f fertilizer and these other factors (agro-ecological environment, use o f other inputs, managerial capacity o f farmer, etc.). '24 Finally, the simulations only represent a partial equilibrium analysis and it might be argued that increased fertilizer use would depress cereal prices, and thus also the retum to fertilizer use. This requires a broader discussion including an analysis o f the net market position o f farmers (net cereal buyeriseller) as well as the price and income elasticities o f demand and supply o f cereals, to which we will retum inChapter 6. '25This will addressed at length inthe ongoing ESW undertakenby the Rural Development Sector of the Africa Region. 83 o f rainfall failure.'26 At the same time, other measures (soil conservation) and other agricultural (livestock, bee keeping) and remunerative non-agricultural activities will have to be promoted as well. The role and potential for agricultural intensification in further poverty reduction inEthiopia inrelation to other strategies such as market development and migration out o f low potential areas are discussed inmore depth inChapter 5 and 6. Figure 4.1: Effect of fertilizer use on per adult equivalent expenditures 2050 2000 1950 1900 1850 1800 1750 1700 1650 1600 1550 1500 I % of land I-Yoof natural fertilizer used -YOof comercialfertilizers used I I 4.26 Productive assets, consumer durables and sanitation are all correlated with higher consumption. Consumption appears highly correlated with ownership o f animals and farm equipment. Possession o f bicycles i s a clear sign o f wealth, with consumption per adult equivalent in households with a bicycle up to 12 percent higher than those without. Those lacking access to proper sanitation, proxied here by use o f buckets or the surrounding area for toilet needs, have an average o f 12 percent lower expenditures across all specifications, underscoringthe critical importance o f proper sanitation. 4.27 Diversifying as a coping strategy. While the move out o f agriculture portrayed inthe descriptive statistics may suggest a diversification into better livelihoods, the regression results suggest otherwise. Those who engage in more livelihood strategies tend to be substantially poorer. The results hold both inrural and urban areas. While surprising at first sight, this is very much in line with the empirical findings o f the livelihood diversification literature. Several authors'27 have observed a U-shaped relationship between the proportion o f income earned from off-farm activities and total income in Sub-Saharan Africa, with the poorer and the richer getting a larger share o f their income out o f off-farm activities, but with quite different returns to these activities. Those inthe middle are found to have a larger share inagriculture as they mayhave landholdings sufficient to earn a living and not bepushedinto low-return off-farm activities, but not enough capital to engage in the more remunerative off- farm activities open to richer households. '26The role o f households' risk coping capacity in adopting fertilizer i s addressed in depth in an ongoing study by Christiaensen and Dercon, as part of a multi-country study on the role o f agriculture inreducing poverty in Sub-Saharan Africa. 127Collier and Gunning, 1999; Barret, et al., 2000; Reardon et al., 2000; Toulmin et al, 2000. 84 4.28 The diversification literature often makes the distinction between "push" and "pull" factors promoting diversification out o f agriculture into non-farm activities.12* Adoption o f non-farm activities may occur because households are "pushed" to do so when returns to agriculture are inadequate (either because o f a low asset base or low productivity, or in response to shocks such as droughts). They may also engage in non-farm activities despite higher average returns inagriculturebecauseof seasonalshortages o fcash for consumption or agricultural inputs or due to a need to diversify their risks. This often occurs inthe absence o f functioning formal or informal credit or insurance markets as in Ethiopia. However, households may also be pulled into off-farm activities because they offer higher returns. Yet, this often requires access to skills, capital and (urban) land. The latter i s continuously identified as a key constraint to set up non-farm businesses in Ethi~pia.'~'Evidence from the ERHS shows that poorer households diversifiedinto activities with low access constraints but also low returns, such as charcoal, firewood collection or weaving, while the richer were more able to engage in off-farm activities requiring human and physical capital.130The increasing diversification out o f agriculture and the reported negative effect o f this livelihood diversification suggest that diversification inEthiopia has so far mainly been driven by "push" as opposed to "pull" factors. These findings further underscore the need for a complementary approach to agricultural led development, i.e. the need to reduce the barriers to entry to more remunerativenon-farm activities. We will elaborate on this inChapter 6. 4.29 Coffee and chat producers tend to be better off. Contrary to the bi-variate descriptives presented in Table 1.11 (Part I,Chapter l), which show lower consumption and higher rates o fpoverty among coffee growers compared with other cash crop producers, being a coffee grower generated, ceteris paribus, a premium o f 1.7 percent in overall consumption in1995, which evenincreased to 2.9 percent in1999. While several households movedout of coffee, these results suggest that the wealthier stayed on. This i s also consistent with the small drop inpoverty incidence observed among coffee growers. A similar pattern has lately been observed in the Kilimanjaro Region o f Tanzania, where the poorer coffee producers dropped coffee production first.131 Chat producers on the other hand fetched an even higher consumption premium o f 13.7 percent in 1995, which declined to eight percent in 1999, consistent with the observed increase in poverty incidence among chat producers. Without further detailed information on the income portfolios of the coffee growers it is impossible to conjecture what has happened to their incomes over the past few years when coffee prices collapsed. 4.30 Access to infrastructure and market connectivity are key. Households using electricity as their main source o f energy have on average six percent higher consumption. Distance to food markets, water and transportation services are also highly correlated with consumption. These variables together pick up the benefits o f lower shadow prices, as well as 12* Reardon, 1998; Ellis, 2000b; Kydd et al, 2001. 12' Difficulties inobtaining land was identified as one o f the three key bottlenecks for private sector development inthe EDRI-WorldBank Investment Climate Study. Similarly, Sharp andDevereux (2003) report that access to land in rural towns was often identified as a key constraint to set up a small business intheir destitution study o f the NortheastemHighlands (Amhara Region). I 3 ODercon and Krishnan, 1996. 13' Christiaensen, Hoffmann and Sarris, 2004. 85 the role o f information spillovers and network externalities resulting from connectivity.132 Distance to health centers, on the other hand, i s positively related to higher expenditures. While proximity to health services would supposedly boost expenditures indirectly by promoting health and thus productivity, this surprising result may be picking up placement effects whereby clinics are targeted to areas o f higher illness incidence where they can be more effective. 4.31 There appear to be important returns to access to information. The estimated results suggest that households with a radio/TV are significantly better off than those without, with the difference estimated at 20 percent. Further exploration o f this general result confirms the critical importance o f access to information (see Tables A.4.1 and A.4.2, Appendix). When we split up radio and TV ownership, include average radio ownership in the community to capture externality effects, control for household wealth, and explore the effects inrural and urban areas separately, we find that inrural areas households with a radio are 17.5 percent richer than those without, and that an increase in the proportion o f households owning a radio in a community ownership by 10 percentage points increases a household's consumption in that community by 3.9 percent. As expected, givenmuch wider dispersion o f individual radio ownership, the externality effect i s less in urban communities. Giventhat only 14 percent o f rural households own a radio, there is clearly tremendous scope for improving people's well being by enhancing their access to information, especially given their current isolation from the outside world. Increasing radio ownership emerges as an important and cost effective interventionto do so, which deserves much more attention. 4.2.2 Returns across space 4.32 We now broaden our analysis to provide a spatial perspective o f poverty and consumption in Ethiopia. We examine inparticular how returns to endowments differ across different spatial classifications: rural versus urban as well as by region (Tigray, Amhara, Oromiya, and SNNPR). The estimated results (Table 4.10) are based on a woreda fixed effects and are described in more detail below. Understanding the differences in returns across space can help target investments to optimize their effect on growth and poverty reduction. 4.33 Education matters a great deal across the regions. Returns to female education tend to be higher inurbanareas than inrural areas and inurban areas they are about twice as high relative to returns to male education. Returns to primary and secondary schooling are especially high in SNNPR, though surprisingly there seems to be no pay-off from completing post secondary school. This justifies extra efforts to raise primary school attendance and completion in SNNPR, especially for girls. The effect o f household amenities and asset ownership appears relatively robust across space. Returns to electricity, however, are especially highinTigray. 132We refer to the World Bank, 2004c, for a more detailed discussion of the role of public infrastructure in promoting economic growth. 86 Table 4.10: Variationsin returns to endowments across space Rural Urban Tigray Amhara Oromiya SNNPR Coef. t Coef. t Coef. t Coef. t Coef. t Coef. t Private endowments Human CapitaVDemographics Size of household -0,0982 -46.95 -0,1006 -42.02 -0.0967 -15.92 -0.0910 -23.37 -0.0969 .32.52 -0.0982 -26.19 Dependencyratio -0.0108 -2.76 -0.0580 -10.05 -0,0238 -2.25 -0.0358 -4.83 -0.0122 -2.1 -0.0100 -1.37 Ratio of Females in house 0.1529 8.83 0.2474 10.69 0.2773 5.98 0.1368 4.77 0.1496 5.49 0.1592 4.86 Femaleheadof house 0.0037 0.39 -0.0879 -7.57 -0.0247 -0.97 0.0064 0.38 -0.0224 -1.58 0.0291 1.66 Age -0.0053 -4.01 -0.0198 -10.26 -0.0061 -1.75 -0.0076 -3.18 -0.0069 -3.46 -0.0059 -2.28 Age squared 4.00E- 05 2.97 0.0002 8.04 0.0001 1.92 0.0001 2.68 4.84E-05 2.41 0.0000 1.49 Grade obtained by adult males 0.0186 11.82 0.0101 7.44 0.0083 1.9 0.01I 5 4.17 0.0186 8.95 0.0191 7.66 Grade obtained by adult females 0.0156 6.62 0.0213 14.18 0.0010 0.25 0.0228 6.65 0.0191 6.49 0.0316 7.89 No of adults completed 0.3 post secondary 0.1927 3.76 0.2050 16.94 0.3414 4.39 0.1282 2.56 0.1830 4.47 0,0188 Physical capital Own plough 0.0899 9.83 0.1301 5.78 0.0442 1.46 0.1053 6.08 0.0961 6.83 0.0602 3.56 Own farm animal 0.0588 6.41 0.0559 3.39 0.0801 2.88 0.0870 5.13 0.0610 4.33 0.0663 3.7 Own transport animal 0.1202 14.44 0,0055 0.23 0.0906 3.75 0.1104 7.63 0.1061 8.31 0.1200 6.21 Own bicycle 0.0981 1.99 0.1691 5.94 0.2066 2.46 -0.0037 -0.04 0.0859 159 0.2064 3.21 Own TViRadio 0.1902 16.49 0.2348 19.9 0.I847 6.31 0.2425 10.29 0.1751 11.74 0.2192 10.58 No toilet in household -0.0903 -6.37 -0,1573 -12.36 -0.0446 -I -0.1942 -6.58 -0,1259 -6.78 -0.0999 -5.12 Livelihoods Obtain some income 0.27 0.0196 1.27 0.0251 I.57 from coffee (I=yes) 0.0337 3.4 -0.0460 -2.44 0.0697 1.61 0.0053 Obtain some income 2.54 0.1055 5.24 -0.0095 -0.28 from chat (l=yes) 0.0849 5.23 0.0969 2.39 -0.0786 -0.98 0,1221 Share in income from 0.0775 0.0200 0.55 agriculture 0.0387 1.6 0.0517 1.8 0,1500 2.68 -0.0765 -2.37 2.65 Share of income from -0.1279 -0.53 wages 0.0587 1.73 -0.0455 -3.21 0.0689 1.08 0.1822 4.67 -3.79 -0.0242 Share of income from -7.37 -0.3409 -0.6742 -1 1.68 other sources -0.5491 -16.08 -0.0005 -0.02 -0.0046 -0.07 -0.3492 -8.33 Numberof livelihood strategies engaged in -0.0820 -10.17 -0.0558 -6.23 -0.0881 -4.33 -0.1337 -9.68 -0.0655 -5.68 -0.0537 -3.93 Public endowments Electricity as source of 2.89 0,1028 0.1624 4.29 householdenergy 0.0807 2.16 0.1003 6.14 0,1943 4.04 0.1229 3.8 Distanceto food market -0.0009 -1.13 0.0124 4.51 -0.0020 -0.71 -0.0014 -1.09 -0.0052 -3.45 -0.0008 -0.44 (km) Distance to water (km) -0.0024 -4.11 -0.0105 -3.16 -0.0007 -0.2 -0.0038 -4.83 -0.0010 -0.89 0.0065 3.05 Distance to Health(km) 0.0016 2.76 -0.0026 -1.33 -0.0042 -1.82 0.0009 0.86 0.0048 4.86 0.0005 0.56 Distanceto transport -4.54 -0.0025 -2.28 services (km) -0.0017 -6.29 0.0007 1.1 -0.0017 -2.1 -0.0020 -5.3 -0.0011 Year (=1 if year=1999) 7.7387 181.96 8.2540 168.87 -0.0617 -1.92 0.1402 9.91 -0.0393 -2.82 0.0277 1.53 Constant 7.7387 181.96 8.2540 168.87 7.5078 66.52 7.9625 118.22 7.9312 136.19 7.6880 110.1 Number of 14984 12097 1787 5724 6930 4275 observations R-squared 0.4570 0.4098 0.4955 0.4281 0.4161 0.4953 4.34 Returns to market connectivity vary substantially, as do returns to livelihood diversification. The results on several o f the distance variables appear unstable across the regions, which may be related to the use o f the woreda Fixed Effects Model in addition to the region-specific regressions. If access to services and markets are largely woreda-specific, their effects would be largely captured by the woreda dummies. However, proximity to transportation services, a measure o f market connectivity, emerges as an important correlate 87 o f consumption in all regions, though the size o f the effect differs. The results on diversification are also robust across all regions: increasing the number of sources from which a household receives their income i s negatively cowelated with expenditures. Such apparent distress diversification seems especially prevalent inthe Amhara Region. 4.3 Geography and Poverty 4.35 Having discussed the spatial variations to returns, we now analyze the specific effects of geographic characteristics on poverty. We look inparticular at the effect o f agro- ecological characteristics, population density, measures o f soil degradation (or proxies thereof), malaria risk, and rainfall shocks. To do so, we replace the woreda-level dummies with proxies of these geographic characteristics. Most of our geographic measures are at the woreda level. We continue to use pooled 1995 and 1999 data. Table 4.11 presents our results. The first column is our base householdworeda fixed effects model for c~mparison.'~~ Inthe second column, we introduce all relevant and available geographic variables. Notethat for these models, we drop all observations that do not having matching data for these secondary community-level variables. As a result, we lose about 10,000 of the 27,000 observations o f the base (pooled) m0de1.l~~ The estimated effects on household and public endowments remain robust to the replacement o f the woreda dummies by the actual geographical characteristics. In the third column we introduce crop damage, dropping all urban observations, for which no crop damage data are available. Table 4.11:Effects of agro-ecology, population density, soil degradation and shocks on consumption Base') Model 1') Model 2" Coef. t Coef. t Coef. t Private endowments Human CapitaVDemographics Size of household -0.0962 -48.09 -0.0929 -42.99 -0.0939 -35.9 Dependency ratio -0.0176 -4.61 -0.0225 -5.41 -0.0161 -3.25 Ratio of Females in house 0.1612 9.63 0.1716 9.33 0.1656 7.52 Female head of house 0.0013 0.15 -0.0063 -0.63 -0.0080 -0.67 Age -0.0075 -5.8 -0.0085 -6.04 -0.0059 -3.52 Age squared 0.0001 4.44 0.0001 4.62 0.0000 2.5 Grade obtained by adult males 0.0161 11.62 0.0139 9.29 0.0157 8.1 Grade obtained by adult females 0.0203 10.63 0.0169 8.13 0.0122 4.15 No of adults completedpost secondary 0.1433 4.85 0.1289 3.94 0.2014 3.11 Physical capital Own plough 0.0967 10.63 0.0493 5.3 0.0418 3.87 Own farm animal 0.0607 6.71 0.0621 6.43 0.0655 5.76 Own transport animal 0.1221 14.54 0.1382 15.64 0.1338 13.18 Own bicycle 0.1043 2.76 0.0806 1.95 0.0541 0.87 Own TViRadio 0.2020 18.91 0.2215 19.15 0.2175 15 ' 3 3Note that for the base model we continue to include woreda-level dummies. This permits us to gauge the robustness o f the effects o f the household and community characteristics when replacing the woreda dummies with their actual geographical characterizations. '34 As a result o f dropping all observations without information on the relevant secondary woreda variables, we lose all observations on Somali, Afar, as well as observations in some other regions. Tigray, Amhara, Oromiya, SNNPR and Benishangul all remain inthe reduced sample. 88 Base') Model 12) Model Z3) Coef. t Coef. t Coef. t No toilet inhousehold -0,1148 -9.15 -0.1044 -8.09 -0.0958 -5.75 Livelihoods Obtain some income from coffee (l=yes) 0.0284 2.94 0.0255 2.94 0.0259 2.57 Obtain some income from chat (l=yes) 0.0889 5.45 0.1341 9.89 0.1219 7.79 Share in income from agriculture 0.0707 3.58 0.0591 2.84 0.0347 1.19 Share of income from wages 0.0181 0.79 0.0086 0.35 0.0509 1.22 Share of income from other sources -0.3779 -13.87 -0.4496 -15.6 -0.6109 -14.89 Number of livelihood strategies engaged in -0.0834 -I1.32 -0.1055 -13.22 -0.1083 -10.78 Public endowment Electricity as source of householdenergy 0.0622 3.02 0.0523 2.59 0.1576 3.48 Distance to food market (km) -0.0009 -1.13 -0.0005 -0.77 0.0000 0.06 Distance to water (km) -0.0019 -3.05 -0.0008 -1.37 0.0002 0.3 Distance to transport services (km) -0.0016 -6.21 -0.001 1 -6.27 -0.0010 -5.1 Riskfactors/Shocks Household head was sick in last 2 months (1=yes) 0.0007 0.09 0.0109 1.22 High Malaria Risk (l=yes) 0.0470 4.33 0.0392 2.96 High Malaria Risk*Dist. To health -0.0022 -3.09 -0.0018 -2.15 1 year lagged rain shock -0.1491 -8.34 -0.2049 -9.24 1 year lagged rain shock squared -0.0992 -2.28 -0.1406 -2.57 Contemporary rain shock 0.0775 4.48 0.0956 4.69 Contemporary rain shock squared -0.0078 -0.15 0.0039 0.07 Percentagecultivated land area damaged4 -0.1588 -4.96 Agro-ecological characteristics Mean Altitude (m) 0.0005 8.52 0.0005 7.41 Mean Altitude (m) squared 0.0000 -8.54 0.0000 -7.52 Long Run Average Rainfall (mm) -0.0001 -4.49 -0.0001 -4.62 Long Run Coefficient of Variation of Rain -0.1541 -3.21 -0.18 14 -3.13 Soil degradation DAP equiv. nutrient loss (tonsiyriworeda) due to dung and crop residuecollection -0.0002 -8.05 -0.0002 -7.34 Average soil loss (tonshdyear) -0.0002 -0.49 -0.0004 -0.9 Rural Population per arable area (people per km') 0.0000 -5.63 -0.0001 -5.29 Urban 0.1294 6.09 0.1508 7.91 Year (1 if year = 1999) 0.04 14 4.81 0.0682 7.87 0.1162 10.95 Constant 7.7931 203.85 7.4902 102.21 7.4819 83.98 Number of observations 17,168 17,166 11,685 R-squared 0.4484 0.2922 0.2653 ') The Base is presented for comparison and continues to incorporate woreda-level dummies, which we do not report. We use the same observations as Model 1 to facilitate observations 2, Model 1 i s run over the pooled observations, adding all geographic variables for which data were available, and constrained by the addition o f secondary variables for which data was not available across all observations. 3, Model 2 adds the variable o n cultivated land damaged to Model 1, thereby losing all urban observations for which no information on crop damage was collected. 89 4.36 Living at higher altitudes is generally associated with higher levels of consumption. Figure 4.2 captures the relationship between altitude and expenditures which, given the significance o f the altitude and altitude squared coefficients, are estimated with extreme precision. While expenditures initially increase with altitude, the negative coefficient on altitude squared signifies a concave relationship, and meanexpenditures eventually reach a maximum around 2,200m after which the squared term dominates and higher altitudes are correlated with lower expenditures. In sum, there seems to be a premium to living in the 1,800-2,400 m altitude range, consistent with the historical and political focus on the highlands. Figure 4.2: Correlation between altitude and expenditures o ! , , , I , , , , , , ~ I : , Altitude (m) 4.37 Stable rainfalls correlate with higher consumption. For populations whose livelihoods depend largely on rain-fed agriculture, the amount and especially the variability o f rainfall would appear critical. Fluctuations in rainfall are indeed inversely related to consumption. The larger the coefficient o f variation in rainfall totals, the lower i s consumption. Moreover, households in areas at the 75th percentile in terms o f rainfall fluctuation consume on average two percent less than those in areas at the 25th percentile, which corresponds to a difference in the coefficient o f variation o f 0.13. This underscores the importance o f appropriate risk management interventions such as water management and ex- post insurance schemes to reduce the detrimental effects o f rainfall fluctuations on welfare. From Figure 1.6 we further recall that larger fluctuations in rainfall go hand in hand with lower average rainfall, suggesting that those households would be penalized twice. Yet, contrary to intuition, we don't find a positive effect o f long-run rainfall averages on consumption. Our estimated results indicate even a negative effect. 4.38 Welfare losses from using dung collection as energy source appear to be substantial. Our estimated results suggest that the welfare reducing effects o f using dung as an energy source are sizeable. For example, reduction o f annual DAP equivalent nutrient loss 90 due to dung collection in Amhara from the current estimated average o f 325 tons to the national average of 185 tons per woreda could increase average per capita consumption by 2.8 percent, equivalent to more than one and a half years o f economic growth at its historical pace o f 1.7 percent per capita. Dungremoval could affect welfare indifferent ways. An important channel will undoubtedly be soil depletion, though several environmental economists have emphasized that the extent of nutrient loss critically depends on the nature of composting, the timing o f the application and the cultivation techniques through which dung is worked into the soil. Furthermore, welfare loss from dung collection may also follow from the strenuous demands it places on household labor, especially girls and women. While negative, the effect o f annual topsoil loss, another proxy for soil degradation, i s much less precisely estimated. This may partly follow from the fact that its effects largely depend on the deptho f the topsoil, which determines the moisture retention capacity. Annual topsoil runoff may be too imperfect a proxy to capture this. As further illustrated inFigure4.3, these results suggest the urgent need to promote the use o f alternative energy sources to dung cakes and the use o f more efficient cooking stoves. Increased fertilizer use would further help counteract soil depletion. Figure 4.3: Effect of DAP equivalent nutrient loss due to dung collection on expenditures 1820 , DAP loss (tondyear) 4.39 The larger the rural population density over arable land, the lower is average consumption. Population density may affect household consumption in several ways. On the one hand, larger population density would be conducive to attract o f f - f m employment due to network externalities and a reduction in transaction costs. Yet, in a rain-fed rural economy, and without controlling for the possession o f land, it may also capture increasing land pressure. The latter effect clearly dominates in our results, with woredas with a rural population density o f 100 more people per square kilometer o f arable land being on average one percent poorer. 4.4 Risk and Poverty 4.40 As discussed in Chapter 1, risks permeate daily life in Ethiopia. To explore the contours o f the effect o f risks on welfare, we empirically estimate the effect o f risk as well as 91 the occurrence o f actual shocks on observed consumption. We look in particular at harvest failure (due to such natural occurrences as droughts, floods, pests, and frost), which has been identified as the most frequently occurring shock, and at health indicator^.'^' We first present the results from the national surveys (Table 4.11) and complement the findings with those from the large emergingbodyofrigorous empirical work on risk, vulnerability and poverty in Ethiopia. 4.41 The effects of rainfall shocks appear inconclusive. We examine the effect o f both current and lagged rainfall shocks.'36 This i s partly motivated by the fact that our expenditure data are out o f phase with the agricultural year. By introducing squared terms, we also try to capture non-linearities in the effect o f rainfall shocks on consumption-larger negative shocks are likely to have larger negative effects. Figure 4.4 graphs the estimated relationship between expenditures and the lagged and contemporary shocks as well as their combined effect. While the effect o f the contemporary shocks i s consistent with our intuition, the lagged effects suggest that those who experience larger negative shocks tend to be better o f f on average. When combing the effects o f both shocks, the latter (perverse) effect dominates. Conforming with our regression results, descriptive investigation indeed shows positive correlations between the contemporary shocks and expenditures, though much higher expenditures among those who experienced the largest shocks in 1994 and 1998, the years lagging the survey years. These ambiguous results may be related to unobserved effects o f food aid in response to shocks. Food aid distributions were especially high in 1994 despite average rainfalls, and these may have offset the negative effects o f the shocks. Moreover, there are still large differences in rainfall within a woreda which our rainfall shock measure does not capture. Figure 4.4: Effect of rain shocks on expenditures 2000 1900 1800 3 1700 2 , --Contemporary Shock Effects 5 I i1600 -Combined Shock ~ Effects i 1500 -0.75 -0.55 -0.35 -0.15 0.05 0.25 0.45 0.65 % Variation of shock from long run rainfall 4.42 Yet more precise measures of harvest failure point to strong negative effects. When we take crop area damaged reportedby the farmer as our measure o f harvest failure as opposed to deviations from long runaverage rainfall measured at the woreda level, we do find substantial negative effects o f crop damage on consumption. More precisely, a 10 percent 135Dercon and Krishnan, 2000 136Shocks are defined as the percentage deviation o f current rainfall amount fromthe long runaverage, 92 increase in damaged land results in a 1.5 percent decrease in mean expenditure (see Model 2, Table 4.1 1 above). It could be conjectured that overall about 2.7 percent o f consumption was lost due to crop damage.I3' Other micro evidence supports these findings. Using data from the EMS panel between 1994 and 1997, it was shown that a 10 percent increase in rainfall resulted in approximately a 10 percent increase in agricultural Further analysis o f the same panel indicates that a 10 percent increase in rainfall increased total income (including from agricultural and non-agricultural sources) by about five percent between 1989 and 1995 and consumption expenditure by about four percent.I3' It was also found that 10 percent more crop damage relative to the mean reduced consumption by about 0.4 percent, while 10 percent more livestock disease would have resulted in a 1.5 percent reduction in con~umption.'~~ 4.43 Although there is little evidence that short term illness affects incomes and consumption,malaria and serious illnessepisodes may have important (negative) effects on consumption. Equally puzzling are our findings on the effect of malaria risk in Table 4.11 (above), which suggest higher consumption among those residing in highly infested areas. While consistent with our bi-variate findings on poverty and malaria incidence, it also suggests that our malaria incidence variables may be picking up some other unobserved effects. Yet case study evidence from Tigray estimates the value o f preventingmalaria with vaccines at about US$ 36 per household or about 15 percent o f the imputedannual household income.I4l We do not find a significant effect o f reported illness on consumption. Other evidence from the ERHS does not indicate a strong effect o f illness on consumption outcomes either-a 10 percent increase in illness was found to reduce consumption by 0.1 percent.I4* This could be caused by a variety o f factors, including that communities and families are able to insure each other against this type o f misfortune, and there i s evidence that this indeed happens in many instance^.'^^ Furthermore, the relevant shocks are likely to be confined to serious illness episodes, not short periods of illness that temporarily affect individuals and households. A serious illness episode o f one o f the adults in the household in the last few years reduced consumption by about seven percent.144 Other estimates suggested that illness of the household head reduced annual income o f the householdby 11.2 percent.145 4.44 Food aid provides some protection against the fluctuations in living standards. Food aid could be expected to help households to cope with these fluctuations. Both from nationally representative cross-section data and panel data there i s some evidence o f income targeting, with the poor more likely to receive either food aid or food-for-work, although targeting i s typically relatively weak.146 There i s less evidence o f a sensitivity o f these forms 13' Given that 17 percent of all cultivated land was reported damaged, total consumption loss amounts to 0.17*0.16=2.68 percent. 13' Van den Broeck, 2004. 13' Dercon, 2002. I 4 ODercon andKrishnan, 2000a 141 Cropper, et al., 2004. 142 Dercon and Krishnan, 2000a. 143 Dercon and Krishnan, 2000b. 144 Dercon, 2004. '45 Asfaw, et al., 2004. 146 Jape et al., 2002; Dercon and Krishnan, 2004. 93 o f support to local level common and idiosyncratic shock. One o f the reasons may well be that support is often not timely, at times exacerbating fluctuations rather than dampening it.'47 Still, despite relatively poor targeting, the impact of food aid may be more positive than this would suggest, as food aid may be shared to some extent in communities, compensating for some o f the targeting fai1~res.I~~ 4.45 But large fluctuations in consumption and nutrition continue to occur. Participatory work as a background to the World Development Report 2000/01 highlighted the large impact of shocks on households. Quantitative data confirm this, with poverty levels in relatively bad harvest years, i.e. years with substantial rainfall failure and crop damage, about a quarter higher than those ina reasonably good year, i.e. years without serious drought or crop damage.'49 Seasonal movements are also substantial, suggesting that mechanisms to keep consumption stable are even deficient in coping with predictable fluctuations. These fluctuations in consumption translate in some o f the highest recurring fluctuations in adult nutrition observed in the world.150 Furthermore, these fluctuations in consumption related to risk and seasonality seem to have important gender dimensions, with men typically experiencing less fluctuation than women.'51 4.46 The impact of uninsuredshocks is not confinedto fluctuations,but its effects are persistentand contributeto continuingpoverty subsequently. Qualitative data has shown the ratcheting effects of adverse shocks in rural Ethiopia, with people's livelihoods permanently affected by the loss o f assets and the needfor further coping ~trategies.'~~ There i s also quantitative panel data evidence o f the long term implications for poverty persistence o f shocks. Past rainfall shocks continue to affect consumption subsequently, not just rainfall inthe immediatepast. For example, a 10percent reduction inrainfall one or two years ago, relative to the long run mean, reduced consumption by about 3.5 percent, while a similar shock between three and five years ago reduced consumption by about 1.5 percent. This implies that a shock in one year will also further reduce consumption in subsequent years. The persistent effects o f a serious crisis were noticeable more than 10 years later. Evidence on the impact o f the famine in the mid-1980s was linked to household consumption in the 1990s, using an index o f severity o f impact based on the type o f coping strategies households had to resort to, such as cutting back meals, eating wild foods or moving to feeding camps. The results indicate that households more seriously affected in 1984-85 experienced much lower growth in the period 1989 to 1997: comparing the 25thand 75thpercentiles o f famine impact meant that the less affected group experienced at least three percent higher per capita growthinconsumption than those more affected.'53 147 Barrett et al., 2004. 14' Dercon and Krishnan, 2004. 14' Dercon andKrishnan, 2000a. Dercon and Krishnan (2004) find that the average lowest versus highest level of the body mass index was about 90 percentinthe E M S sample. ''I Derconand Krishnan, 2000b. Rahmatoand Kidanu, 1999. Dercon, 2004. 94 4.47 The options available to households to manage risk are limited, and there would be substantial scope for broader interventions beyond food-for-work and other safety nets. The evidence on the large impact o f shocks, and especially their ratcheting, persistent effects highlight the high benefits o f containing any crisis and the need to find ways o f supporting those affected by a crisis well beyond the initial crisis period. Food-for-work and other safety nets clearly have a role to play but by their nature targeting i s often difficult while timing issues affect their effectiveness to handle local level crises, not least those caused by idiosyncratic or localized effects. Existing mutual support systems, part o f many communities' social capital, provide some additional assistance but they are not suitable to handle large, covariate shocks, for example related to climatic conditions. 4.48 Reducing household vulnerability to crisis will also have to imply reducing the dependence on a small number o f agricultural-based livelihood strategies. Finding ways to get people to diversify their livelihoods towards higher return and sustainable agricultural and non-agricultural activities i s clearly essential. However, the issue highlighted above i s relevant here as well: access to profitable diversification i s seriously restricted. Evidence from Dercon and Krishnan (1996) suggests that despite the benefits from diversification as part o f a risk management strategy, many poor households cannot engage sufficiently into it, since profitable diversification into highreturn agricultural activities (including intensification or high value livestock) or non-agricultural activities (such as business or wage employment) i s restrictedto those households with sufficient access to physical and human capital. 4.49 Pastoralists employ a variety of risk coping strategies, with mixed success. As indicated in Chapter 1, pastoralists are especially exposed to risks. Mobility o f herds i s the traditional strategy to reduce exposure to risk among pastoralists, but it i s proving increasingly difficult due to a combination o f regulation, urbanization, insecurity in areas away from towns, and growth in agriculture. In the absence o f insurance markets, investing in greater herd size is a costly, but effective way of ensuring that the animal stock does not fall below threshold levels which would threaten households' livelihoods. Income diversification can be a risk-reducing strategy, though it happens to be differently motivated at the high and at the low end of the distribution as discussed above. There i s evidence that income i s increasingly diversified for pastoralist households, though there seems to be a lot o f heterogeneity in the opportunities as well as the motivation for doing so. Depending on the climate and on the characteristics o f the towns whose markets they can access, households might engage in agriculture (to secure food supplies or to sell their crops on the market), gather natural resources (e.g. wood, out o f which coal is often made), or send household members to work for wages. 4.50 This complex picture suggests that three policy areas should be addressed to better cater to the needs of pastoralists. The first is collecting data to document the monetary and non-monetary dimensions o f the poverty experienced by pastoralists and compare their most urgent needs with those o f the settled population. The second i s to take into account the heterogeneity o f pastoralists and the differential risks and opportunities to which they are exposed when targeting assistance, favoring for example self-targeting to community based modalities.154 Lentz and Barrett, 2004. 95 4.51 Policies to strengthen households' asset base should be supplemented with promoting a broad range of ex-ante risk management strategies,.including broadening micro-finance activities to include savings and insurance to cope with shocks, further insurance initiatives such as related to rainfall insurance, and activities to strengthen existing community based institutions such as the iddirs (originally funeral societies, but increasingly offering broader support). Promotion of better water management (either through water harvesting, micro dams or irrigation) i s also important to reduce dependence on rainfall. To help pastoralists better manage their risks, policies should aim to facilitate rather than to hinder pastoralists' own mechanisms o f risk-reduction and coping, including increased herd mobility and size. With respect to the latter, it has been pointed out that in resource- constrained contexts, this means redirecting newly stockless or near-stockless pastoralists out o f pastoralism through managed purchases and skills training which offer the means to more sustainable li~elihoods.'~~ '55McPeak and Barrett, 2001. 96 CHAPTER 5. PERFORMANCE AND POTENTIAL OF THE AGRICULTURAL SECTOR 5.1 As discussed in Chapter 1, failure of the agricultural sector to exceed population growth lies at the heart o f the observed stagnation or only limiteddecline inpoverty reduction over the past decade and a half. This underscores two key strategic points. First, agriculture will continue to have to play an important role inEthiopia's development, which i s still inthe early stages o f its structural transformation. This i s not because agriculture has an inherent superior growth rate compared to the non-agricultural sectors, but rather because o f its size in the economy and its importance for the livelihoods of virtually all rural pe0p1e.l~~Second, in- depth analysis of the causes for the lackluster performance of the agricultural sector, despite concerted efforts by the government to boost agricultural productivity over the past decade and a half (see Box 5.1), i s called for. In doing so, it will be important to account for the agro-ecological diversity o f Ethiopia. Box 5.1: Agriculturalpoliciesin Ethiopia since 1992 Agricultural Development-Led Industrialization (ADLI) has been the cornerstone o f Ethiopia's poverty reduction strategy since the EPRDF assumed power. Ina first step to reinvigorate the agricultural sector and the overall economy more broadly, the government introduced a series o f policy reforms to stabilize its economy and to liberalize output and input markets. It removed quantitative restrictions o n private grain trade and abolished the compulsory delivery o f grain quotas at prices well below the market price. These measures, together with the elimination o f export taxes (except for the reduced one on coffee) and the devaluation o f the Ethiopian Birr, improved price incentives and fostered the integration o f food markets'". Empirical evidence suggests that these early (food market) reforms paid off both in terms o f economic growth and poverty reduction. In a purposively selected sample o f six villages across Ethiopia, Dercon (2002) observed an average reduction in poverty by 29 percentage points between 1989 and 1995, o f which 18 percentage points could be attributed to better producer prices and 23 percentage points to increased retums to road infrastructure and proximity to urban centers. Both changes were largely fuelled by the reforms, and to a lesser extent, also helped by peace. While the studied sample i s admittedly small (and not nationally representative), the results are nonetheless striking and illustrative. In the mid 1990s the focus shifted from policy reforms designed to "get the prices right" to public investmentin agricultural extensionaimed at boosting productivity through the use o f improved technologies. Through the Participatory Demonstration and Training Extension System (PADETES) the government delivered off-the-shelf packages o f fertilizer, improved seed and credit, as well as information on input use and better agricultural practices. The promotion o f the credit-fertilizer packages was accompanied by a further liberalization o f the fertilizer market. By 1997, fertilizer subsidies were completely removed and retail prices were fully liberalized, which also resulted in higher fertilizer prices. The use o f fertilizer increased, though diffusion and adoption rates remained disappointin despite-some even argue because of-strong-handed promotion o f the credit-fertilizer packages at times.'" On average, agricultural output continued to fall behind population growth. Acknowledging the limited success of PADETES, the government revisited the program and formulated an integrated rural and agriculture development strategy which was launched in 2002. Agricultural development through agricultural extension combined with credit-fertilizer-seed packages remained one o f the comer stones o f the strategy. However, a much wider array of packages adapted to particular agro-ecological circumstances is now being promoted in recognition o f the geographic diversity o f Ethiopia, Promotion o f water harvesting and 156 Mellor, 1995. Is' Dercon, 1995. Is* Rahmato and Kidanu, 1999. 97 market development form other key pillars. The need for additional measures to ensure land tenure security was recognized, though ultimately ownership o f land will remain with the government. The strategy was further complemented in 2003 by the Poor Area Program, developed by the government/donor Coalition for Food Security inresponse to the looming threat o f a massive famine in2003. Key components o f the program include the planned resettlement o f 2.2 million people by 2005, and the development of a safety net program for the chronically food insecure. 5.2 The potential for the agricultural sector to help reduce poverty in Ethiopia is indeed hotly debated, though the debate is often not well served by rigorous empirical evidence. Key questions dominating the policy debate in Ethiopia include the following: I s there a future for agricultural development given the increasingly small plots from which farmers must earn their living? How much growth and in particularly poverty reduction can we expect from the agricultural sector by imparting better technologies (modern inputs, improved cultivation techniques and reversal o f soil degradation), better institutions (markets and land tenure security), and better risk management tools (irrigation and water harvesting; safety net programs and rainfall or index-based insurance)? Have the current policies and extension programs, which have mainly focused on liberalizing input and output markets (early 1990s) and promoting the use o f modern inputs (late 1990s) paid off? When looking at country-wide averages, successes in high potential areas could for example have been obscured by the failures observed in the low potential areas, due among other factors to increasing soil depletion. More recently, agricultural policies have been broadened to also focus on promotion o f water harvesting and market development. 5.3 This chapter seeks to contribute to this debate and shed some light on these questions from a microhehavioral and poverty reduction perspective. In particular, it will explore the scope for increasing productivity and income in staple crop production by analyzing the relative importance o f the different determinants o f staple crop productivity. The focus i s on staple crop production because most rural farmers are engaged in staple crop production and staple crops make up 50 to 70 percent o f total expenditures among the rural poor and 40 to 50 percent among the urban poor.'59 For this analysis, the chapter draws on the nationally representative agricultural sample surveys which the Central Statistical Authorities have kindly made available for the first time, as well as the emerging empirical literature on the determinants of cereal production in Ethiopia. Nonetheless, given the scope and complexity o f the issue, our discussion i s inevitably partial. More specifics from a sectoral perspective (such as micro-finance, environmental degradation, labor markets, input and output markets) will be provided in a series o f other sectoral studies currently under way.'6o The focus in this report is on the strategic directions o f investments and policies across sectors from a poverty-reducingperspective. 15'Staple crops include cereals, pulses, oilseeds and root crops (especially enset). Cereals are the most important staple, making up between 35 and 50 percent o f the total expenditures among the rural poor and 30 and 35 percent o f total expenditures among the urban poor. I6OWorld Bank, 2004c. A detailed and comprehensive study on the challenges to improve the performance o f the agricultural sector is beingpreparedby the World Bank's Rural Development Sector o f the Africa Region; a report on rural finance is being prepared by the Finance Sector o f the Africa Region; a report on poverty and environmental degradation i s being prepared by the Environment Sector o f the Africa Region; a labor market study is being preparedby the Poverty Reduction and Economic Management sector o f the Africa Region. 98 5.4 The chapter begins by providing a broad review o f the sector's performance from a macro perspective. This i s followed by a micro analysis o f the determinants o f staple crop production as well as a discussion o f the potential for increasing smallholders' income from staple crop production across different geographical areas. A brief discussion o f the different options to foster broad-based agricultural development highlightingkey policy implications and identifying areas which merit further analysis concludes the chapter. Chapter 6 will then build on the insights obtained from this partial equilibrium analysis and comment further on the role o f agriculture and its different sub sectors (staple crops, livestock, traditional andnon- traditional export crops), as well as the role o f the non-agricultural sector inpoverty reduction from a more integrated macro and strategic perspective, also taking into account the geographical diversity. 5.1 StagnatingAgricultural Performance-A Macro Perspective 5.1.1 Stylizedfacts 5.5 Ethiopia still finds itself at the very beginning of its structural transformation, despite a decade and a halfo f policy and investment focus on agricultural development. Even today, agriculture i s still responsible for 85 percent o f employment, 45 percent o f national income and more than 90 percent o f exports.'61 With 96 percent o f the rural population employed in agriculture, rural households continue to rely almost exclusively on low input, low output, subsistence-oriented, rained agriculture and agriculture related activities. Some four to five million people are considered chronically food insecure and annually in need o f food aid. An additional six to seven million are transitorily food insecure and inneedo f food aid when the rains or harvests fail. 5.6 Labor productivity in agriculture remains extremely low, with Ethiopia ranking 78thin a comparison o f 84 countries.'62 O f the total area under temporary crops inthe 1990s, cereals, pulses, and oilseeds accounted for 88.7 percent, 8.7 percent and 2.7 percent respectively. Commercial fertilizer was applied to approximately 40 percent o f total farmland under cereals over the past years, with wheat the most fertilized crop, followed by teff and maize (about 60, 50 and 25 percent o f total cultivated area respectively). In 1999, improved seeds were applied on less than five percent and pesticides on less than 6.3 percent o f the total cultivated cereal area. Less than one percent o f the total cultivated area in Ethiopia i s irrigated, despite massive fluctuations in rainfall. 5.7 Agriculture is mainly concentrated in the highlands, which contain nearly 85 percent o f the population, 95 percent o f the cultivated land, and 80 percent o f the country's 35 million cattle, which form a critical part o f Ethiopia's ox-plow cultivation system.'63 Staple crop production (which i s largely dominated by cereal production, though enset i s an important staple in the southern parts) makes up 65 percent o f the total agricultural value- added, with livestock production ranking second at 26 percent and non-traditional export Exports include coffee (the country's main export until recently), chat, oilseeds, pulses, livestock products and increasingly horticulture products. The importance of coffee has declined over the past years following the collapse ininternational prices. 16'World Bank, 2005, 163Milas and ElAynaoui, 2004. 99 crops and coffee ranking third and fourth with 4.4 and 4.8 percent respectively. When compared to the total economy, staple crops make up one-third o f the total economy, the livestock sector 13.5 percent and non-traditional export crops and coffee 2.3 and 2.5 percent re~pectively.'~~ dominant farming system in the highlands o f Ethiopia i s mixed cereal- The livestock production with dairy production becoming more important in the central highlands in urban and peri-urban areas around Addis Ababa. While cereals (mainly barley, wheat, maize, teff and sorghum) are the dominant crops in the sub humid central and northwestem highlands (within Oromiya and the westem part o f Amhara), especially coffee and enset, but also cereals (maize) are g o w n in the humid high potential southem and western highlands (mostly within Oromiya). 65 Barley and sorghum are the more important crops inthe northem and northeastern highlands (Tigray and eastern Amhara), where livestock holding continues to be important. Most pastoralists live in Afar and Somalia as well as southem Ethiopia (Borena plateau). 5.8 Analysis of agricultural performance over the past 15 years paints a bleak picture (Figure 5.1). Despite agricultural policy reforms in the early 1990s and substantial investments in extension afterwards, agriculture has failed to keep up with population growth, resulting in zero growth in agricultural GDP per capita. The picture is further characterized by huge fluctuations from year to year, largely inresponse to erratic rainfall (as opposed to price fluctuations). This can be seen from Figure 5.2 which shows that agricultural GDP closely tracks overall food production. The correlation coefficient between agricultural GDP and rainfall i s estimated at 0.26. Figure 5.1: Agriculturalperformancein Ethiopia during 1990-2004.'' Real agricultural GDP growth per capita (%) from 1990-2004 ' 1 5 - 'IDatapoints for 2003 and2004arebasedonprojections. ___ ~~ 164Diao, et al., 2004. 16'Ehui and Pender, 2004. 100 Figure 5.2: Trends in the value of agricultural output and food production1990-2004." -+Realagricultural GDP growth per capita (%) +Food production growth per capita (YO) I ') Datapoints for 2003 and 2004 are basedonprojections. 5.9 Moreover, land expansion, rather than productivity gains, still contributed importantly to the observed increase in (cereal) output. Between 1994 and 2000, land under cereal cultivation increased by 5.6 percent while cereal yields increased by only 2.5 percent. It was thus estimated that between 1996 and 2002, about 90 percent o f the increase intotal crop production and 70 percent of the increase incereal production were due to area expansion.'66 Figure 5.3 further shows that higher yields are mainly driven by an increase in maize yields, largely due to increased fertilizer use. Overall, cereal yields have hovered around 1.1 to 1.2 tons per hectare over the past two decades, or still only one-fifth the yields observed inAsia since the Green R e v ~ l u t i o n . ' ~ ~ Figure 5.3: Trends in agricultural yields, 1980-2000 (quintaldhectare) -Total Grain "Cereals - - - Maize --W-Teff Source: Kuma, 2002 '66Diao, et al., 2004. 16'Gabremadhin,2004. 101 5.10 Use of commercial fertilizer per ha has gone up, but breaking this down by crop we find that maize and wheat accounted for all of the increase, with fertilizer use on barley and sorghum constant, and even decreasing on teff (Figure 5.4). This is in line with yield responses across crops, which tend to be highest for maize and wheat, and are much lower for teff. Figure 5.4: Commercial fertilizer use per ha per crop over time'' 0.80 7 3 0.70 ....... . ........~....... .. ... ... ...... .. e 2 0.60 .C -.-.-60.50 N -.-g 0.40 L 0.30 E ii E 0.20 I 0.10 t SORGHUM 4 II A A - I L 0.00 , T 1994 1995 1996 1997 1998 1999 2000 2001 Year Because o f apparent measurement error, data for 1998 and 1999 have been omitted. 5.1.2 Spatial differentiation in productivity response 5.11 The trends in national averages presentedso far should not necessarily be taken as proof that the agricultural extension policies have completely failed. Increased use o f fertilizer may for example have helped stem a further decline in agricultural productivity due to soil erosion and soil nutrient depletion. Furthermore, national averages may mask important differences in productivity response across different geographical areas and population groups. 5.12 While it has often been argued that the extension programs and the accompanying fertilizer-improved seed-credit packages paid off in terms of yield increases in the high potential but perhaps not in the low potential areas, finding yield increases in high potential areas has proven to be somewhat elusive. For example, when comparing the evolution o f yields in two high potential (Arsi and West Shewa) and two low potential (East and West Hararghe) zones, yields actually improved in the low-potential zones, while they remained stagnant or even dropped in the high potential zones. Fertilized area grew in all four zones.'68 High and low potential had been defined in this exercise in MacMillan, 2003. 102 terms o f yields per ha in 1994/5, before PADETES was la~nched.'~' (see Box 5.2 for alternative definitions of high and low potential areas). While these results are striking, firm conclusions cannot be drawn based on the comparison o f only four zones. Yet even though some geographical pattems o f yield evolutions emerge when the analysis i s extended to the national level and using other definitions of high and low potential, the picture remains largely mixed (see Box 5.3). Box 5.2: Alternativedefinitionsof high and low potentialareas The agricultural potential o f an area depends broadly on its agro-ecological endowments (soil fertility, climate, water availability), its connectivity to markets, its population density, and the risk factors governing the area, While these factors are usually fixed in the short run, most o f them can be changed in the long run through investments and policies. For example, lack o f water or rainfall variability could be overcome through irrigation schemes. Distances to markets could be reduced through road and telecommunication infrastructure. Population settlement pattems could be influenced through migration policies and urbanization strategies. Health risks (e.g. malaria and tse-tse fly contamination) could be eradicated. The cost-benefit ratios o f these investments will o f course depend on the original endowment structure o f the area, with investments presumably more cost effective inhighpotentialareas than lowpotential areas. Classification o f areas into low and high potential i s usually based on a combined assessment o f how the area scores on these different factors, i.e. innate agro-ecological potential, market connectivity, population density, health and climate risks, and cost effectiveness o f different investments. This also holds for Ethiopia. However, there i s currently no commonly agreed-upon classification, and different institutions and authors continue to use different classifications. For example, areas have been classified as high and low potential based on: (1) the proportion o f food insecure people in the woreda in the mid 1990s; (2) the existence o f a food deficit or food surplus at the zonal level (Diao et al., 2004); (3) the creation o f an index combining all the different factors mentioned above (World Bank, 2004~). Nonetheless, there is a core o f low and high potential areas which is shared by most classifications. In the absence o f a uniformly accepted classification, we will use the classifications used by the different studies we draw upon, carefully indicating the particular definitions used, with the implicit understanding that they share a common set o f areas with low and high agricultural potential across these different classifications, even though it i s acknowledged that there i s not a perfect match. 5.13 This hypothesis is only borne out when we follow the government's classification of high and low potential zones based on the number o f food insecure people in 1994/95 (see Figures 5.5 and 5.6).I7O As in the other classifications, we find cereal yields to be substantially higher in the food secure tercile than inthe food insecure tercile (1,272 kg/ha on average during the 1994-2001 period versus 994 kg/ha) (Figure 5.6). Average yields in the middletercile closely approximate those inthe food secure tercile. Yet contrary to the results from the other classifications, yields have trended upward in the food secure (and medium food secure) zones, while they stagnated inthe food insecure zones. Yields inthe food secure zones were estimated to increase annually on average by 26.6 kg per ha from 1,153 kg/ha in 1994 to 1,338 kg/ha in2001. The annual increase inthe medium tercile was estimated at 15.2 kg per ha per year, while yields increased only by seven kg per ha per year in the food insecure zones. The high potential areas averaged above two metric todha during the 1994-2000 period and the low potential ones 1.2 metric todha. "O Ineffect, the classification is based on the average number o f people estimated to be inneed o f food aid in each Woreda in 1994 and 1995. 103 Box 5.3: Trends in cereal yields in low, medium, and highpotentialzones Figures B5.3.1 - B5.3.3 below presenttrends incerealyields across administrative zones by terciles of length of growing period, by average distance from the nearest all weather road, and by highland and lowland, similar to the categories used by IFPRI in a forthcoming study to explore the potential for agricultural growth in Ethiopia."' When classifying the zones by terciles of length of growing period, we see that yields are indeed substantially lower inthe low potential group comparedto the medium and high potential ones, though changes inyields do not follow the expectedpattern(Figure B5.3.1). They havebeenupward trending between 1994 and 12001 for the medium potential group of zones, slightly upward trending for the low potential group, and stagnatinginthe highpotential group. ~ Similarly, cereal yields are closely related to market connectedness as capturedby the distance to an all-weather road, though they have been slightly upward trending irrespective of the distance to the road (Figure 5.3.2). Yields inhighlands and lowlands are very similar onaverage, as has beentheir evolutionover time. The upward sloping trend for the highlands disappears when the 2001 observationi s dropped. This illustrates the larger point that great caution i s warranted in interpreting these results given the relative short time period under consideration and substantial fluctuations in yields. Nonetheless, while the evidence points to higher yields in higher potential and more accessible areas, it does not confirm the hypothesis of upward trending yields inhigh potential areas. Figure B5.3.1: Trends inyields by lengthof growing period between 1994 and 2001 Trends in Cereal Yields by Length of Growing Period (1994-2001) -. 1 I y = 6 . 0 2 4 2 ~+ 13.765 15.00 .~ . . . R2=0.0023 - - c 3 14.00 .-cC2 9 13.00 2 In -m 12.00 11.00 - R -0.3944 ............ g 0 10.00 9 Q. 9.00 y = 0.1134~+ 9.733 R2 0.1751 - 8.00 I 1994 1995 1996 1997 1998 1999 2000 2001 Year Linear (High potential) - Linear (Medium Potential) --- Linear (LOWPotential) 1 "' Diao, 2004. 104 Figure B5.3.2: Trends in cereal yields by market access between 1994 and 2001 Trends in CerealYield by Distancefromthe nearest all Weather Road (1994-2001) y=0.135x+11.837 1994 1995 1996 1997 1998 1999 2000 2001 Year &first nearest -+-secondnearest A third nearest -Linear (first nearest) -Linear (second nearest) --." Linear (third nearest) Figure B5.3.3: Trends in cereal yields in high and lowlands between 1994 and 2001 Trends in Cereal Yield for Highland and Lowlands -+ (1994-2001) - 16.00 .-3 15.00 ----I ................................................................................................. 0- ......................... .................. ........................................ 1994 1995 1996 1997 1998 1999 2000 2001 Year +Highland -4-Lowland - - Linear (Highland) Linear (Lowland) I 105 Figure 5.5: Food security potentialin Ethiopia FOOD SECURITY POTENTIAL IN ETHIOPIA ATZONAL LEVEL I Figure 5.6: Trends in cereal yields by food security index between 1994 and 2001 16.00 1 _________? $ $,I5.O0 14.00 0 -m 0 13.00 89 12.00 g 10.00 a 9.00 e9 8.00 1 1994 1995 1996 1997 1998 1999 2000 2001 1 - +Food secure -4-Mediumfoodsecure I-. Food insecure Linear Food insecure) Linear (Medium food secure) -*I" fLinear [Food secure) 106 5.14 Starting in 1996, the roll-out o f the extension-fertilizer-credit packages has been the pillar o f the government's agricultural policy. Giventhe close relationship observed inFigure 5.6 above between change in commercial fertilizer use and change in yields, we would thus expect an upward trend incommercial fertilizer use inthe food secure zones and stagnation in the food insecure zones. This i s confirmed inFigure5.7. Figure 5.7: Trends in commercial fertilizer use by food security indexbetween 1994 and 2001 Trends in the Quantity of Commercial (DAP,UREA,mixed) FertilizerApplied by Food Security Index(1994-2001) 1 - .-.- 5 0.60 I LL .- 3 e $0 20.50 g 0.40 t 0.30 O E . E 2 0.20 g 0.10 y=00024x+02138 2 m R2= 0 0094 v 0.00 1994 1995 1996 1997 1998 1999 2000 2001 Year 5.15 ' These findings raise a number of importantquestions regardingthe diffusion and adoption of fertilizer and itspotentialfor povertyreduction. The distinctionbetween high and low potential appears to be a misnomer as it implicitly assumes that farmers would be Imore inclined to adopt commercial fertilizer in high potential areas since it would be most productive in these areas. Yet fertilizer adoption depends both on its diffusiodavailability as well as the determinants o f adoption such as profitability, credit availability, access to input and output markets, etc. It appears that there have been important program placement effects at work during the earlier phases o f PADETES, which may explain why we fail to observe a clear-cut relationship between indicators of inherent production potential (such as agro- ecological indicators) and trends in yields. On the contrary, when we follow the government's conceptualization o f high and low potential areas, which it may have used to channel its extension packages, a positive correlation emerges. 5.16 The more food secure areas may indeed have been more conducive to fertilizer diffusion and adoption than the food insecure areas (Table 5.1). Not only do the food secure areas have greater agro-ecological potential (higher levels o f rainfall, less variability and flatter arable lands), they also have better road infrastructure and higher population density. Market connectivity and agglomeration effects reduce transaction costs o f fertilizer distribution and provide producers with outlets for their products. Furthermore, even before PADETES, fertilizer use was already more widespread and its application rates more intense 107 inthe food secure (and medium food secure) zones (Figure 5.7, above). This would suggest that the farming population was already aware o f the usefulness of fertilizer, which provides a fertile ground for promoting further uptake. Table 5.1: Characteristics of food secure and food insecure (rural) zones in 2000'' Food Insecure MediumFood Food Secure Zones Insecure Zones Zones Average proportion (%) ofpeople inneedof food 21 08 02 aid per zone Household characteristics Total Value o f agricultural production (Birr in 867 1354 942 2000 nominal prices) Household size 5.1 5.4 5.3 Number of individuals working inthe agricultural 2.1 2.1 2.1 sector (imputed) Age of head 46 43 43 Head female (%) 19 19 19 Completedgrade 1-3 (%) 14 14 15 Completedgrade 4-6 (%) 06 07 09 Completedgrade 7 or above (%) 03 04 06 Landarea per holder (ha) 0.83 1.23 0.97 Agricultural practices % of landirrigated 1 1 0 % of landusing improved seeds 2 2 4 % of landusing commercial fertilizers 13 24 32 Quantity of commercial fertilizers used 14.5 29.0 45.4 (Kilograms per Hectare) % of land using natural fertilizers 12 6 11 % of landusing pesticides 00 8 12 Number of cattle for mainly agricultural use .84 1.14 .89 Shocks % of crop area damaged 22 16 14 % deviation from average rainfall in2000 2 -1 1 Agro-ecological characteristics Average rainfall (1967-2000, mm) 920 1060 1253 Coefficient of Variation of Rainfall .27 .24 .21 Slope 14.0 9.8 9.5 Altitude (meters) 1976 1875 1996 Market connectivity& agglomeration Distanceto Food Market (km) 7.4 7.3 6.0 Distanceto all weather road (km) 15.7 12.9 11.0 Rural Population density (population I km') 121 110 153 Soil degradation Nitrogen loss (ton I year I woreda) 355 291 246 Phosphorus loss ( (ton I year I woreda) 99 79 63 ')Zones are classified based on the weighted averageproportion of people per woreda assessedto be inneedof food aid by the Disaster Prevention and Preparedness Commission in 1994 and 1995. Following a ranking of the proportion of food aid-needy ineach zone, the zones were distributed inthree equal groups from food secure to food insecure. 108 5.17 When in pursuit of quick success in raising agricultural output in Ethiopia, channelingfertilizer packages to the areas which are morelikely to adopt and where it is less costly to roll-outthe package makes politicaland economic sense. Nonetheless, two importantobservationsremain. First, while cereal yields inthe food secure areas grew at a rate o f 2.1 per cent per year, at an average o f 1,338 kg/ha cereal yields are still quite low compared to intemational standards. Second, taking a more distributional perspective, many o f the (chronically) poor live in the food insecure zones,172and they also face greater land pressure (average land size per holder i s 0.83 versus 0.96 in the food secure areas) and more soil degradation due to dung and crop residue collection. While these conditions would seem conducive to the use of land-saving technologies such as improved seeds and commercial f e r t i l i ~ e r , 'application o f these technologies have so far been limited in these areas. These ~ ~ observations raise several important questions. I s limited application o f fertilizer and improved seeds in fact not profitable in these areas, or are there important constraints (e.g. associated risks and limited ability to cope with shocks ex post, limited access to credit, input supply constraints, input delivery mechanisms) which prevent farmers from using these modem inputs? What i s the scope for further yield increases both inthe food secure and food insecure areas? What i s the role o f non-price factors, such as modem input use, soil conservation, water and risk management, and market access, in this process? What i s the role o f price factors such as input and output prices? These are the issues taken up in the following section usingmultivariate regression techniques. 5.2 EnhancingStapleCrop Productivity-A Micro Perspective 5.18 T o explore the scope for enhancing staple crop productivity and thus reducing poverty by increasing agricultural income in food secure and food insecure areas, we investigate the relative importance of modern input use and other agro-ecological and geographicalcharacteristicsin determiningthe total value of staple crop production. In particular, we estimate a Cobb-Douglas production function linking the total (nominal) gross value of staple crop production per holder (cereals, pulses and oilseeds)'74 in 2000 to total land and labor input,'75household characteristics, input use and geographical endowments. We do so for each o f the three groups o f zones: food secure, the medium food secure, and the food insecure. The results are presentedinTable 5.2. 5.19 The first three columns o f Table 5.2 present the estimated effects controlling for unobserved community variables by inclusion o f community dummy variables. This provides assurance that the estimated effects o f the different inputs and household characteristics do not pick up any other unobserved(community or price) effects. To further explore the effects o f community variables we explicitly include community and woreda characteristics (long run average rainfall and rainfall fluctuations, rainfall deviation from the long runmean in 2000 to capture the effect o f covariant shocks, and the average slope in a woreda to proxy soil moisture retention capacity). Separate production functions were also estimated at the plot Sharp, Devereux and Amare, 2003. Ruttan and Hayami, 1985. We do not have information on the production o f enset or other roots and tubers. The actual labor variable used is the imputed number o f active adults working inthe agricultural sector using the household and community characteristics reported in the agricultural survey, and their relationship to the number o f adults employedin the agricultural sector as estimated from the Welfare Monitoring Survey. 109 instead of the holder level and for each of the cereal crops. The results were largely consistent with the ones presented here, and the additional insights will b e highlighted in the discussion. All the coefficients in the table represent either elasticities (if the variable is expressed in log^),"^ or a percentage change in the total value of output (if the variable is expressed as a dummy variable i.e. taking the value o f either one or zero).'77 Table 5.2: Estimated determinants of the value of cereal output in 2000 by food security potential o f the area') Dependentvariable = log (nominal) total value of agriculturalproductionin2000 Food insecure food secure Medium Food secure insecure Food food secure Medium Food secure Log cultivated area (m2) 1.053"' 1.081*** 1.079"* 1.030'"' 1.081*** 1.094*** Log adult labor in agriculture -0.013 -0.016 -0.006 -0.03 1 -0.064" -0.011 Logage (years) -0.004 -0.001 0.039' 0.084" 0.013 0.036 Education: 1-3 years (dummy) 0.016 -0.005 -0.004 0.047 0.035 -0.020 -0.007 -0.031 0.025 0.000 -0.089*** 0.039 Education: 7 + years (dummy) Education: 4-6 years (dummy) 0.017 0.042 0.009 -0.032 -0.053 -0.001 Holder i s female (dummy) -0.003 -0.017 -0.011 0.016 -0.022 -0.020 Log (number of oxen) 0.01I** 0.007 0.006 0.028"' 0.014 0.012 Log (%area on which natural fertilizer used) 0.004 0.002 -0.004 0.001 0.019*** -0.005 Log (%area on which commercial fertilizer used) 0.012*** 0.012**' 0.018"' 0.030"* 0.035"' 0.028*** Log (% crop not damaged) 0.090"' 0.238"' 0.144*** 0.236'" 0.210"*' 0.240"* Log (% area on which pesticides used) 0.012"' 0.007'* 0.009*' 0.003 Log (% area on which improved seeds used) 0.005 0.001 -0.009 -0.002 Log mean rainfall (1967-2000) 0.046 -0.305*'* -0.336"' Percentagedeviation from mean rainfall in 2000 0.824"' 0.432"' -0.308** Log altitude (m) 0.947'** 0.583"* 0.5 1O**' Log (average slope percentage in a woreda) -0.169*** -0.061' 0.067 _.-. Constant -2 805"' -2.949*'* -3.334*** -9.794"- -4.957'+* -4.985*** EA fixed effects? YES YES YES NO NO NO Observations 4490 4631 4898 4011 4370 4785 R-squared 0.95 0.93 0.95 0.87 0.87 0.91 * significant at 10%; ** significant at 5%; *** significant at 1% ') Zones classified based on the weighted average proportion o f people per woreda assessed to be inneed o f food aid by the Disaster Prevention and Preparedness Commission in 1994 and 1995. Following a ranking o f the proportion o f food aid-needy in each zone, the zones were distributedin three equal groups from food secure to food insecure. 5.20 In discussing the insights deriving from these regressions, which are based on nationally representative household data, w e further draw upon the emerging Ethiopia- specific empirical literature on the determinants o f agricultural production. This permits us to triangulate the robustness and the magnitudes of the estimated results with other evidence. It also allows us to explore the role o f other factors, inparticular, output and input price factors, 176 So, for example, according to the estimated coefficient reported in column 1, a 10 percent increase in the cultivated area in food insecure zones would result in a 10.53 percent increase in the total value o f staple crop production. 177 So, for example, according to the estimated coefficient reported incolumn 1, the total value of the staple crop production among households in food insecure zones whose head had obtained one - three years o f education is on average 1.6 percent higher than in those households whose head has no formal education. Note however, that the coefficient is not statistically significant, indicating that there is no statistical ground to reject the hypothesis that household education has no effect at all. 110 as well as access to markets, which were either not explicitly considered in this analysis due to data limitations or which turned out to be statistically insignificant. Marginal valuesof labor and land 5.21 There are strong signs of labor surplus and land scarcity in cereal production. Perhaps the most robust, but somewhat striking results across all specifications (including those by crop and by plot not reported here) are the estimated elasticities o f land and labor, The elasticity on labor is consistently estimated at zero implying that an additional unit of labor does not add any value to overall production. For each percentage increase in landholdings, on the other hand, the value o f total output rises by at least one percent. Iftrue, this would imply that across the nation, there is a labor surplus, with people working at zero (ifnot negative) marginal returns on their land, and that the size of current landholdings is suboptimal given the amount o f family labor available. 5.22 While we likely overestimate the amount o f labor actually applied to the production of cereal, pulses and oilseeds inthis analysis, which partly explains the extremely low estimated marginal retums to labor,'78 other empirical studies o f agricultural production in Ethiopia also report few gains from additional labor use in agricultural production, holding other inputs constant, and a strong correlation between the amount o f land and output, though the estimated elasticities are not quite zero and one re~pectively.'~'These econometric results mirror farmers' testimonies collected during participatory work pointing to increasing land pressure and the emergence o f a class living on hunger plots.'" Finally, low marginal productivity o f labor and farming on sub-optimal landholdings are consistent with the observed absence o f fbnctioning land and labor markets, which prevents an optimal allocation o f land and labor. 5.23 The low marginalvalue of labor in terms of additional agriculturalincome from cereal production, given current landholding size and the higher marginal values of expanding landholdings, point to four possible policy routes to expand households' income from staple crop productionand reduce poverty. First, the land frontier could be pushed further and new areas could be exploited (agricultural extenszjkation). As discussed before, this has so far been the major natural source o f growth in agricultural GDP, not the increase in yields. The ongoing resettlement program i s consistent with this philosophy, though not necessarily an optimal policy to foster migration to unexploited areas given the related humanitarian crises ifoften generates. Land redistribution does not seem a valid Labor input inthe production o f cereals, pulses and oil seeds is proxied by the predicted number o f household members employed in agriculture (see footnote 172). This is not only an inaccurate measure, but likely also an overestimate as households may only work part o f their time in agriculture, and even within agriculture, they are likely to devote some o f their time to other agricultural activities (such as livestock, cash crops, or cultivation of perennials). Technically, measurement error in the labor variable leads to a downward bias o f the estimated marginal effect of labor use. Nonetheless, labor and land elasticities o f zero and one respectively, have been reported elsewhere inthe literature as well (see Hoddinott, et al., 2003, for Zimbabwe). Croppenstedt and Muller, 2000; Demeke, Mekonnen, Admassie, and Aredo, eds., 2003; Amacher, et al., 2004; Pender and Gebremedhin, 2004. "O Rahmato and Kidanu, 1999. 111 option, given that the observed land inequality does not result from a small minority holding the majority o f land, but rather from the emergence o f a land poor class following continuous fragmentation into ever smaller holdingsof the current, already small, landholdings. 5.24 Second, the marginal productivity o f labor and land could be raised through agricultural intensijkation (e.g. modern input use and more effective water and land management). Increasing modern input use has been the focus o f PADETES, which has recently been further complemented with better water management. Third, labor productivity could be increased through diversijkation into non-cereal (tradable) agricultural production, including cash crops (coffee, chat, cotton), tree production (eucalyptus), livestock (dairy, meat), bee keeping, medicinal plants, and others. Fourth, landpressure could be reduced and labor productivity enhanced through diversiJication out of agriculture into highly remunerative non-agricultural activities. The optimal (combination of) routes will obviously differ across space depending on the regions' comparative advantage in terms o f agro- ecological potential, and market access as determined by population density and access to infrastructure.lg1 We focus here on the scope for agricultural intensification in food secure and food insecure areas ina partial equilibriumsetting. Chapter 6 will revisit these issues and the optimal combination o f routes from a broader economic (as opposed to sectoral and micro) perspective, and discuss the linkages between performance in the cereal, non-cereal and non-agricultural sectors, and relatedly the critical role o f food price levels and fluctuations. Other non-price factors 5.25 There is substantial scope for agricultural intensification, especially through combineduse of fertilizer and improvedseeds. Farmers' practical experiences indicate that cereal yields could be more than doubled when using fertilizer combined with improved seeds (Table 5.3). However, while there appears to be a high premium from using fertilizer- improved seed packages, these are currently only applied in four per cent o f the cultivated cereal area. On the other hand, yield gains from using fertilizer alone, a cultivation practice currently applied in about 30 percent o f the cultivated cereal area, would only increase cereal yields by 18.7 percent when compared with no modern input use. This positive, though somewhat limited, effect o f fertilizer use alone i s consistent with the relatively small estimated elasticities o f value o f staple crop production reported in Table 5.2 (holding all other inputs constant) which range from 0.012 to 0.035, as well as those found in other empirical studies from Ethiopia which report elasticities to fertilizer use inthe range o f five to eight percent.ls2 The effect o f jointly using fertilizer and improved seeds has not been estimated inTable 5.2. "' EhuiandPender, 2004. Croppenstedt and Muller, 2000; Demeke, Mekonnen, Admassie, and Aredo, eds., 2003; Deininger, Jin, Adenw, Gebre-Selassie,Demeke, 2003; Amacher, et al., 2004. 112 Table 5.3: Cereal area and yield by moderninput use, 1997/98-2001/02 Area Yield'' percent difference with (percent intotal cereal area) (metric todha) no input use No input use 60.6 1.15 Fertilizer only 30.1 1.36 18.7 Fertilizer & improved seeds 3.8 2.46 114.3 Irrigation only 1.7 1.87 62.8 Improved seed only 0.2 1.73 50.8 Total 100 1.28 ')Averages calculated from Agricultural Sample Surveys 1997198-2000/01 and Agricultural Census, 2001102, Central Statistical Authority. Source: Extractedfrom Diao et al., (2004) 5.26 Substantialyield gains from combinedfertilizer-improved seed packages are not limited to food secure/food surplus areas, though the gains are likely smaller in the food insecure/food deficit areas. Farmers' experience from the 2001/02 season suggests that yields were 50 to 80 percent larger in food insecure/food deficit areas and food secure/food surplus respectively when using both fertilizer and improved seeds, compared to the overall average (Table 5.4). Table 5.4: Cereal yield and input use in food deficit, food balanced and food surplus areas, 2001/021' Food Food Food deficit balance surplus Cereal yield (tonka) 1.08 1.19 1.44 Cereal yield usingfertilizer only 1.24 1.25 1.44 Cereal yield usingfertilizer &improved seed 1.65 2.20 2.63 Absolute difference between using fertilizer & improved seed 0.57 1.01 1.19 compared to average cereal yield (tonka) % difference between usingfertilizer & improved seed 53 85 84 compared to average cereal yield (tonha) Fertilizer use rate incereals (% area) 29.12 26.40 56.13 Fertilizer combined with seed rate (% area) 3.08 3.15 4.88 'I)Calculated fromAgricultural Census, 2001102(FederalDemocratic Republic of Ethiopia, Central Statistical Authority). Source: Diao et al., 2004 5.27 Given that improved seeds and fertilizer are applied to only four percent of the total area, there is clearlystill substantialscope for increasingagriculturalincomes from enhanced input adoption, even if the estimated yield gains may be smaller than suggested above. Given that results reported above are based on bi-variate analysis, it is possible that they overestimate the gains from the combined fertilizer-improved seed package, as other factors associated with the use o f this package (e.g. improved cultivation techniques, soil quality, etc.) may partly be driving the results. Also, not all cereals may be equally responsive to the package, and the observed gains may have been largely driven by yield Zones are classified into food deficit, food balanced and food surplus if the zonal level per rural household cereal equivalent output is 20 percent below the national average, between 80 and 120 percent o f the national average, and 20 percent above the national average respectively (Diao, X. et al., 2004). 113 gains among certain cereals. To the extent that these two issues are important, they would reduce the gains to be expected from only promoting the fertilizer-improved seed package. Further analysis of the current experience suggests that maize and sorghum are most responsive to the combined package, with yields for maize increasing from 1.57 to 2.70 todha and yields for sorghum increasing from 1.28 'to 2.07 todha. Together they account for 40 percent o f the cultivated cereal area. Yet currently only 10 percent o f the sorghum area and 39 percent o f maize area i s fertilized, and virtually no improved seeds are used in sorghum production while only 13 percent o f the maize area was cultivated with improved seeds. Sorghum i s usually grown in the drier areas and maize in the more humid areas, suggesting that yield gains could be obtained both in the food insecure and the food secure areas. Moreover, even though yield gains are less when fertilizer alone i s used, this does not mean that there are no gains from using fertilizer alone. Clearly, given that the use o f basic modern inputs i s currently so limited, there i s still substantial scope for increasing productivity from broadening adoption o f these inputs, especially when combined with adequate technical advice on optimal input application and use. For adoption o f these technologies to be sustainable, it must further be ensured that they are compatible with farmers' incentive structures. 5.28 Results from experimental and demonstration plots indicate that the observed improvementsrepresent only lower bounds compared to what could be achieved through improved management practices and inputs both in food secure and food insecure areas (Tables 5.5 and 5.6). Table 5.5: Potential for yield gains under improved management Yield with improved practices (todha) Crop Yield with traditional Farmers' SG 2000 EMTPs3) practices (todha) Experimental ') Demonstration plots 2, plots2' Average Range Maize 1.63 4.5-9.0 5.0-6.0 5.5 3.4-9.5 Teff 0.86 2.0-3.0 1.4-2.2 1.75 1.46-2.7 Wheat 1.28 2.0-5.0 2.0-4 .O 3.1 2.1-6.0 Sorghum 1.32 2.5-6.0 1.5-4.0 Barley 1.13 2.1-4.5 1.8-3.5 ')Source: CSA Agricultural Sample Surveys 1988/89, 89/90, 91/92 (average). 2,Proposed agricultural technology recommendations 3,SG - Sasakawa-Global: EMTPs: Extension Management Training Plots Source: World Bank, 2001 Table 5.6: Average yield data from SPA practice compared to traditional practice Zone Traditional No. o f Yield intodha (todha) demonstrations Average Range North Gondar 0.5 120 1.26 0.76-1.8 South Tigray 0.5 41 1.12 0.43-2.03 NorthWollo 0.55 129 0.78 0.29-2.01 Total 0.5 290 1.05 0.49-1.98 Source: Sodhi, Manna and Wadhawan, 1999 114 5.29 Topsoil runoff negatively affects agricultural income, especially in the food insecure areas (Table 5.2). Given high land pressure, farmers in food insecure areas are increasingly forced to cultivate plots on steep slopes, most often without appropriate land cultivation techniques or terracing. Average slopes inthe food insecure areas were estimated at 14 percent while those in the other zones were estimated at about 9.5 percent (Table 5.1). Clearly, farmers in the food insecure areas are at a natural disadvantage in terms o f farming conditions. Experiments in Tigray indicated that for stone bunds three to 20 years old, soil loss decreased on average by 68 percent. Cereal yield increases through use o f stone bunds were estimated at 50 kgper ha (an increase o f about five percent) on average across 150 plots . inDogu'a Tembien (Tigray) in 2002.184 Other evidence from the highlands in Tigray found the value o f production on plots with stone terraces to be 17 percent higher, and 41 percent higher when trees were planted. Food for work programs could be cost effectively used to buildtechnically sound stone bunds and terraces. Giventhe strongrelationship found between dung collection practices and consumption, it was somewhat surprising not to find a statistically significant effect between DAP equivalent loss and agricultural income. The relationship between dung collection practices and agricultural productivity deserves further investigation, preferably with more detailed micro-data.ls5 5.30 Agricultural income loss associated with crop damage and droughts can be substantial. The elasticity related to crop area damaged was estimated at about 0.15, indicating that interventions which would help a farmer protect an additional 10 percent o f his cultivated crop area would raise his agricultural income by 1.5 percent (Table 5.2). Reported crop damage reflect damage resulting from droughts, floods, insect or pest attacks, frosts or other reasons. While these numbers may not speak to the imagination, for those already well below the poverty line who see half their crop area damaged, a 7.5 percent loss in agricultural income can push them into starvation. This i s especially the case when crop damage i s widespread and results in a collapse o f the off-farm labor market. Public work programs for those who have experienced a shock may provide welcome relief inthis case. 5.31 Moreover, given that the estimated elasticity on crop damage reported above is derived from the community fixed effect model, it only reflects idiosyncratic risks. When we unbundle the community effects as reported in columns 4-6 (Table 5.2), we see that the elasticity on crop damage increases to 0.21-0.24. In addition, covariant rainfall shocks measured at the woreda level also have a strong negative effect on agricultural income, especially in the food insecure areas. A 10 percentage point deviation from the long run average rainfall i s estimated to reduce agricultural income for farmers residing in food insecure areas by eight percent. Other empirical studies confirm the critical importance of rainfall, especially inthe food insecure areas.'86 5.32 Output appears responsive to market accessibility. We also examined the effect o f remoteness by including the average distance to a road among members in a community, though no precise relationship was found between this variable and the value o f total output, Zala-Daget Project, 2004. The ongoing study on poverty and environmental degradation by the Environment Sector o f the Africa Region will explore this further. Abrar, Morrissey and Rayner, 2004. 115 and it was subsequently dropped (Table 5.2). This result may follow from the fact that the effect o f market access on cereal production works largely though fostering agricultural intensification (Le. the adoption o f modern inputs) which was already controlled for.lS7 Aggregation across cereals may also have confounded the results. Other micro evidence points to a strong effect o f market accessibility o f cereal output, especially of teff and maize inthe more food secure areas (Central and Southern Zone inTable 5.7). Output ofcash crops (coffee and chat) and enset i s also very responsive to market access. Table 5.7: The impactof market accessibility on output for different crops Output elasticity (kgs) with respectto market Northern2) Central2' accessibility') Zone Zone Southern2)Zone Teff -0.03 +1.23*** -0.03 Maize 0.00 +1.30*** +O. 25*** Wheat 0.00 +0.54*** +1.07*** Sorghum 0.43*** Coffee +1.07*** Chat +2.4 1*** Enset +1.56*** *significant at 10%; **significant at 5%; ***significant at 1% ') Market accessibility i s defined as the population size of the nearest town (or big market) divided by the road distance to this town or market. 2,Pooled 1994, 1995, 1997 EHRS data - 1,500 householdsin 18 villages. Source: Abvar, Morrissey and Rayner, 2004 5.33 Age, gender and education of the holder appear not to affect the value of agricultural output, while oxen ownership has a positive effect (Table 5.2). While the results regarding education could be somewhat surprising, the effects o f education are likely to operate more through the adoption and diffusion o f new technologies (e.g. fertilizer) rather than through better allocative efficiency. Empirical evidence for Ethiopia suggests that educated farmers tend to be early innovators, providing an example to others, and that they also tend to be better at copying those who innovate first, thereby spreading the innovations more rapidly within the location.'88 Oxen ownership positively affects total agricultural output. The reported elasticities are similar to those found in other studies.1s9 Follow-up analysis shows that the critical difference lies in having at least one ox versus having none. Having at least one ox can enhance total agricultural output by three to 10percent. Pricefactors 5.34 Overall, cereal productionis price inelastic,with maize and teff productionin the food secure areas tending to be more price responsive. Duringthe 1980s andthe 1990s it was generally perceived that the overall policy bias against agriculture depressed agricultural prices, preventing the necessary supply response from farmers. Getting the prices right was believed to be an important factor in increasing agricultural production. This also inspiredthe GoE to liberalize the cereal markets in the early 1990s. Table 5.8 presents output supply Is'Pender and Gebremedhin,2004. Is*Weir andKnight, 2000. Is'Abrar, Morrissey and Rayner, 2004. 116 elasticities to own prices for major cereal (and other crops) across different ago-ecological zones. Table 5.8: Output supply elasticitieswith respectto own price Output elasticity (kgs) with respect to own price Northem Northem') Central`) Zone Zone Southem" Zone Maize 0.08 0.62*** 0.57*** Teff 0.06 0.44*** 0.35** Wheat 0.20** 0.24** Sorghum 0.20** Coffee +0.35 *** Chat +1.08*** Enset +0.13* *significant at 10%; **significant at 5%; ***significant at 1% I)Pooled 1994, 1995, 1997 EHRS data - 1,500 households in 18 villages which have been classified in three agro-ecological zones (northern, central and southem) Source: Abrar, Morrissey and Rayner, 2004 5.35 Several observations emerge from this table. First, especially maize, but also teff emerge as the more price responsive crops from this study, with a 10 percent increase in (expected) prices leading to a six and four percent increase in output in the more food secure areas. Given that maize i s quite responsive to the use o f fertilizer and improved seeds, its production can be more easily adapted to changing price levels. In combination with price inelastic demand for maize (see Chapter 6), this implies that maize prices are likely to fluctuate more, as we as also observed during the 2000 and 2001 bumper harvests. Teff i s often grown in rural areas as a cash crop, explaining its higher supply elasticity. This i s consistent with the fact that 50 percent o f the teff area i s fertilized and the relatively high estimated elasticity o f fertilizer demand (in the Central zone) to teff prices (0.43). The own price supply elasticity for barley on the other handwas even estimated at zero. 5.36 Second, farmers appear much less responsive to prices in the northern zone, which is also more remote and characterized by subsistence farming. In other words, non-price factors will be much more important to increase staple crop output in the food insecure zones. Third, as expected, short runprice elasticities o f the production o f perennial crops such as coffee and enset are small, given that one cannot easily switch to another crop once coffee and enset "trees" have beenplanted. Nonetheless, own price supply elasticity o f coffee i s still estimated at 0.35, indicating that even though trees will not be uprooted immediately, through input management (fertilizer, pesticides, maintenance) coffee production can still be somewhat managed in response to expected coffee prices. Supply o f chat on the other hand i s price elastic. Summary of micro-evidence and related emergingpolicy issues 5.37 In conclusion, the micro-evidence suggests that: (1) the marginal productivity of labor in staple crop production is very low; (2) there is still ample scope for expanding staple crop production through agricultural intensification, including in food insecure areas; and (3) there is a pressing need for appropriate risk management strategies. Based on the available evidence reviewed above, rough and conservative estimates suggest 117 doubling yields inthe more food secure areas and increasing yields by 50 percent in the food insecure areas would lie well within the realm o f the possible. Much could be gained from expanding adoption o f the fertilizer-improved seeds packages and increasing market access, especially in the more food secure areas, while broader adoption o f the fertilizer-improved seed packages will have to be complemented with the promotion o f soil conservation and better water and risk management techniques as well as improved market access in the food insecure areas. 5.38 At the same time, these insights raise a series of policy questions surrounding agricultural technology adoption, institutionalarrangements, and risk management. In particular: What are the major factors constraining wider adoption and diffusion o f land saving technologies such as fertilizer, improved seeds, and pesticides, both inthe food secure and food insecure areas, despite their apparent benefits and concerted efforts by the government to promote their adoption and diffusion. Relatedly, what i s the role o f land tenure insecurity in the adoption o f environmentally sustainable cultivation practices to help stem the alarmingrate of soil degradation? How can current institutional arrangements governing the input (land, labor, capital) and output markets be made more effective in improving productivity in agriculture? For example, what are the factors constraining labor mobility, which could relieve some o f the land pressure, and what i s the role o f land ownership and tenure security in increasing agricultural productivity? What are appropriate risk management tools both in food secure and food insecure areas to help farmers protect their productive assets from shocks, and to help them adopt higher returdhigher risktechnologies and portfolios. 5.39 Addressing these critical questions regarding technology adoption, the road map to get the markets right, and the appropriate set o f risk management tools adequately falls beyond the scope of this study. Here, we conclude by referring to other ongoing studies which address these issues in more depth, frame some of the key issues and summarize key emerging insights. InChapter 6, we further reflect on the sustainability o f rapidly increasing productivity in staple crop production, its anticipated effect on food prices, the linkages with other agricultural and non-agricultural sectors and its ultimate effects on poverty. 5.40 For an in depth discussion of the role of land tenure insecurity in raising agricultural productivity and the behavioral determinants of technology adoption includingthe importance o f an input delivery mechanism compatible with farmers' incentive compatible and the role o f households' risk coping capacity we refer to ongoing World Bank Rural DevelopmentEconomic Sector Work as well as the ongoing World Bank Rural Finance Economic Sector Work. The impact o f land tenure insecurity on agricultural productivity and the environment is particularly explored at length in the ongoing World Bank Environment and Poverty ESW. Highlights include that empirical evidence confirms common wisdom that farmers perceive their land tenure as inse~ure''~and that there i s emerging evidence that 190Ethiopian Economic Association, 2002. 118 security o f land tenure encourages long term investments in land improvements (such as terracing, soil bunds). While the pay-offs to these investments can be high, the benefits are usually not immediate and it may take time (up to four years) before substantial yield increases are rea1i~ed.I~'Use o f seasonal inputs (e.g. fertilizer and improved seeds) appear not to be affected by tenure security. How current tenure arrangementsI9* affect labor mobility i s explored inthe upcoming World Bank Labor Market Study. 5.41 Addressingthe "three 1's"-incentives, infrastructureand institutions-emerges as key to get the markets right. A comprehensive discussion o f how the functioning o f the output markets could be substantially improved i s provided in"Can Agriculture Lead Growth inEthiopia? The Importanceof Linkages, Markets, and Tradability" (Gabre-Madhin, 2004), a background paper to the World Bank Country Economic Memorandum, 2004. Inparticular, the report underscores the need to (1) rectify the incentive structure for traders, i.e. lower income tax and less levies, introduction o f proper licensing and elimination o f food aid distortions, (2) improve the institutional framework, i.e. better contract enforcement, the establishment o f standards and grades, and the creation o f a market information systems; and (3) expand the communication (roads and telecom) and storage infrastructure networks. 5.42 Finally, supplemental irrigation, weather based insurance and productive safety nets are the three interventions particularly highlighted in the World Bank Vulnerability Assessment, 2004 which provides a detailed discussion o f the appropriate mix of risk management instruments. There are compelling reasons for Ethiopia to focus on irrigation in general, and small scale irrigation in particular, for poverty reduction. First, unreliable rainfall i s the leading cause o f harvest failure and hunger. Second, the availability of new irrigation technologies (low cost drip systems) make small scale irrigation possible, and open up new opportunities for water conservation. Finally, the country already has successful experience with such a strategy: over 66,000 Ethiopians are reported to enjoy higher crop yields due to small scale irrigation through the Ethiopian Social Reconstruction Development Fund. A recent evaluation o f the experience in Tigray suggests that household incomes could be substantially increased through investment in rainwater harvesting ponds provided that they are close to the homestead, that they are properly constructed, that households grow high value crops (such as vegetables) on the irrigated plots, and that households receive adequate extension support.'93 The government has recognized Ethiopia's irrigation potential and has identified small scale irrigation as key instrument for reducing vulnerability and poverty in the SDPRP. Yet, further evaluation o f the current water harvesting program and the constraints to wider adoption o f irrigation techniques i s called for, as poorly planned irrigation programs introduce their own risks (e.g. increased malaria incidence). 5.43 An innovative low-cost risk management tool which is much less prone to the usual moral hazard issues, is to insure farmers against drought risk through formal contracts with private insurance companies or public institutions. Lrrigation will not be possible for many farmers and rained agriculture will continue to be at the core o f their `'IGebremedhin et al, 1999; Gebremedhin and Swinton, 2002. Ifhouseholdleavesitsvillageformorethantwoseasons,itmayalsolosetheuserrightstoitslands, a Landell Mills,2004. 119 livelihoods for years to come. In this context rainfall based insurance contracts provide a useful alternative. Such contracts insure the contracting party against a specific and objectively verifiable rainfall outcome, e.g. drought, and may be entered into by farmers directly, by credit institutions, or by governments. Such weather-based insurance mechanisms are already available to poor farmers in India, Mexico, and South Africa. There are many ways to deliver this insurance. The contract could stipulate a cash payment to participants upon the realization o f the event, or it could forgive partial loan repayment on an input (e.g. fertilizer) ifrainfall failure o fa certainthresholdoccurs. 5.44 Existing safety nets have saved lives but have been largely unproductive and often not well targeted. Yet they can continue to serve their vital insurance function while being made more productive through a mix o f programs aimed at buildingproductive physical and human assets. Inparticular, guaranteed multi-annual transfers to households in return for participation in public works and targeted health and education programs can: (1) encourage risk-taking behavior among small-holder farmers by insuringagainst downside risk o f consumption loss; (2) build public infrastructure and maintain community assets, which provide complementary inputs to private inputs and improve the productivity o f individuals; and (3) promote market development by increasing demand in places that are otherwise too poor. 120 CHAPTER 6. GROWING OUT OF POVERTY: THE ROLEOF AGRICULTURE AND AID 6.1 We now turn to the key challenge o f the Poverty Assessment: shedding light on the feasibility o f achieving substantial poverty reduction, and providing guidance regarding the appropriate policy and instrument mix to do so. An important benchmark to guide us in this exercise i s the first Millennium Development Goal, by which the Government o f Ethiopia has committed itselfto halve the incidence o f income poverty by 2015 from its 1990 level. 6.2 To address both the prospect and policy question, we take a threefold and complementary approach. We begin with a macro-perspective and examine the overall economic growth needed to reach the income poverty MDG as well as the conditions under which such growth can be expected to deliver substantial poverty reduction. Building on insights from macro-economic modeling undertaken inthe World Bank's Country Economic Memorandum as well as the rapidly expanding literature on the relationship between aid and development, we also explore the role o f external aid inthis process. 6.3 We subsequently move toward the meso-level and discuss growth linkages between the agricultural and non-agricultural sector ina qualitative and quantitative manner in Section 6.2. We empirically characterize cereal demand behavior (income and price elasticities o f cereal demand) in Ethiopia as well as the net market position o f rural Ethiopian households (net buyer versus net seller). We further explore the poverty reducing effects o f different agricultural growth patterns (enhanced staple crop productivity, expansion of the livestock production and non-traditional agricultural export growth), drawing on the findings from the multi-market model developed for World Bank-commissioned research on rural development performed by the International Food Policy Research Institute, which will be discussed in more detail inthe World Bank Rural Development ESW.`94 We also reflect on the optimality of different agricultural growth and non-agricultural growth strategies for different geographical areas, depending on their comparative advantage. 6.4 Finally, in Section 6.3 we complement the insightsobtained from the macro and meso analyses with a micro perspective. That section explores how pro-poor growth can best be brought about by simulating the effect o f different investment programs aimed at enhancing people's private (education, health) and public (infrastructure, environmental quality, information) endowments. We also include simulation o f the effect o f increasing people's access to better cultivation practices and risk management tools.'95 In this micro approach, distributional considerations are explicitly taken into account. We further reflect on the importance o f policy measures to facilitate the necessary supply response by the off-farm sector. Diao et al., 2004. `95For a more detailed account o f how people's private and public endowments can be enhanced see Chapters 8- 10 o f this report, World Bank, 2004a and 2004b, as well as the different Sector Development Programs. Measures needed to reduce land degradation and to improve risk management are extensively discussed in the ongoing Economic Sector Work on the environment and the recently completed Risk and Vulnerability Assessment by the World Bank (World Bank, 2004e). 121 6.1 A Macro Perspective 6.5 Since the EPRDF's ascent to power in 1991, the Ethiopian economy has grown at a decent pace of 4.3 percent (or 1.7 percent per capita)'96. Despite this healthy growth, the estimated decline in poverty incidence seems to have been marginal (from 38.4 percent in 1990 to 36.2 percent in 2004) and characterized by substantial movement in and out o f poverty197(see Tables 1.14 Chapter 1). These results mirror the strong relationship between household consumption and income from agricult~re'~~,which has been flat and highly volatile over the past decade (Figure 1.1, Chapter 1). As a result, the number o f poor people inEthiopia is actually expected to have increased from 18.8 million in 1990 to 24.3 millionin 2004. To explore the prospects o f reaching the poverty MDG, we simulate the poverty reducing effects o f three broad sets o f growth scenarios which differ in the following ways: (1) their overall growth potential; (2) the sectoral composition o f their growth (agriculture, industry, and services) and the labor mobility assumed across sectors over time; and (3) the explicit consideration o f increased foreign aid to help finance the investments necessary to enhance economic growth. 6.6 The poverty reducing effect o f the different growth scenarios i s obtained by applying (sectoral) GDP growth rates to the income distribution observed in the 1995 HICES. Household income was approximated by total expenditure per adult equivalent. In each scenario, the period 1995-2015 i s taken as our projection horizon (2015 i s the target year o f the MDGs). Ineach scenario, actual observed sectoral growth rates199have beenapplied from 1995 to 2004 and the projected sectoral growth rates according to the different scenarios have been applied thereafter. Similarly, observed population growth rates have been applied through 2004 and the medium variant quinquennial PO ulation growth projections from the UnitedNations Population Division are used thereafter!" The strengths and weaknesses o f this approach as well as other assumptions commonly used in simulations of this nature are discussed in more detail in Box 1.2, Chapter 1. To benchmark our results, we begin with a "business as usual" scenario and examine the evolution o f the consumption poverty head count if the Ethiopian economy were to continue at the historical average sectoral growth rates observed between 1992 and2004 without sectoral mobility (Scenario 1, Table 6.1). 6.1.1 The role of agriculture in achieving growth with poverty reduction 6.7 "Business as usual," or no improvement in the current rate of annual agricultural growth, would translate into virtually no reduction in the poverty rate and a larger overall number of people living below the poverty line. Duringthe past 12 years, overall GDP grew at a healthy pace o f 4.3 percent, albeit with substantial differences across sectors. While the industrial and service sectors evolved at 5.4 percent and seven percent respectively, well above population growth (estimated at 2.6 percent), average agricultural I96Average growth rate between 1992 and 2004. 197Dercon and Krishnan, 2000a; Dercon, 2004. 19*Dercon, 2004. I99Estimated sectoral growth rates reported inthe African Living DataBase were used for 2003 and 2004. 2ooThese were 2.49 percent in2005, 2.40 percent between 2006 and 2010, and 2.36 percent from 2011 till 2015, The urbanization rates used are 4.36 percent between 2006 and 2010 and 4.58 percent between 2011 and 2015 compared to 2 and 1.85 percent rural population growth respectively. 122 growth was only 2.2 percent. Abstracting from any structural transformation in the economy (Le. not allowing for labor mobility across sectors) poverty incidence in 2015 under these assumptions is projected to be 36 percent, or similar to its 2004 level and only marginally less than in 1990. Yet, seven million more people would havejoined the ranks of the poor (up to a total o f 31.3 million) and inequality would have increased from 0.33 to 0.40 as measured by the Gini coefficient. While these simulations are admittedly crude, the results are consistent with the findings obtained from an advanced macro-economic model which captures the linkages between the level and composition o f public investment, foreign aid, growth and poverty reduction when applying the "base case" or "business as usual" scenario.201 This poses the critical questions o f whether, and how, such a disastrous outcome can be avoided. Table 6.1: Growth, poverty, and inequality under alternative sectoral growth scenarios.'' Growth rates (percent) Projected Poverty Head Projected Count (percent) Gini coefficient Agri- Industry Services GDP 2005 2010 2015 2015 culture Scenario 1:Historical growth rates (1990s) Average sectoral growth 1992-04 2.2 5.4 7.0 4.3 36.2 35.9 36 0.4 Scenario 2: Differential sectoral composition of growth and labor mobility Scenario 2a) Successful agriculture (1.5% per capita 4.1 5.4 7.0 5.4 35.1 28 5 23.2 0.37 growth as opposed to negative per capita growth) and reasonably successful services and industry(growing at their historical average rates) Scenario 2b) As scenario 2a, but withlabor shifting 4.1 5.4 7 0 5.4 34.9 27.4 21.1 0.35 out o f agriculture into industryand services induced by urbanization (structural transformation) Scenario 3: Permanent increase in foreign aid Real private consumption per 34.6 23 5 1.5.6 0.33 capita growth incase o f permanent 5% increase in current aid-GDP ratio2) ')All scenarios use the 1995 income distribution derived from the HICESIWMS as baseline; they use actual sectoral and population growth rates through 2004 (last two years for GDP are based on projections) and different growth scenarios thereafter. The national population growth projections as well as projected urbanization rates are obtained from the UNPopulation Division. Population growth in2005 i s projected at 2.49 percent; between 2005-2010 at 2.4 percent; and between 2010-2015 at 2.36 percent. 2,AgCnor, Bayraktar, and El Aynaoui, 2004. 6.8 Buoyant agricultural growth provides hope. The second set o f scenarios considers the importance o fthe sectoral composition o f economic growth, the effect o f increased overall economic growth, and the effect o f the structural transformation by which labor absorptionby the non-agricultural sectors increases as agriculture becomes more productive and people migrate to the cities. We begin by examining the effects o f a successful agricultural sector (Scenario 2a). In particular, we keep growth in the service and industrial sector at their reasonably successful historical pace o f seven per cent and 5.4 percent respectively, but allow the agricultural sector to grow at 4.1 percent, which corresponds to 1.5 percent per capita growth as opposed to its historical record o f -0.38 percent per capita. The overall economy would grow at 5.4 percent in this scenario. The results are striking. The poverty head count Agenor, Bayraktar, and El Aynaoui, 2004. 123 would decline by about one third to 23.2 percent, not so far from the poverty MDG o f 19.2 percent. The number o f poor people would decline to about 20 million from the current estimate o f 24.3 million, despite population growth continuously exceeding 2.36 percent. Inequality, as measured by the Gini coefficient, would slightly increase to 0.37 from its 2004 level o f 0.33. 6.9 Accounting for structural transformation. In scenario 2a, we have assumed the sectoral employment shares to be constant over time. However, as agricultural productivity increases, people will move out o f agriculture to take up off-farm employment (industry and services) in the urban centers. To account for this structural transformation o f the Ethiopian economy, we assume that labor moves across the different sectors in line with the urbanization rate projected by the UNPopulation Division202and that they adopt the existing employment patterns o f the urban areas as they move from the countryside to the cities.203 With a successful agricultural sector growing at 4.1 percent, non-agricultural sectors growing at their reasonably successful historical rates, and accounting for labor mobility, the poverty head count would decline to 21.1 percent, slightly above the poverty MDG o f 19.2 percent (scenario 2b). Moreover, under this scenario, inequality would largely be contained (Gini coefficient estimated at 0.35 compared to 0.33 in 2004 and 0.30 in 1995). Given buoyant agricultural growth o f 4.1 percent, slightly faster growth in the non-agricultural sectors and slightly faster urbanization would likely suffice to reach the poverty MDG. Alternatively, slightly faster agricultural growth (which may be more difficult to bring about) would also suffice (scenarios not presented).204We return to the feasibility o f 4.1 percent sustained agricultural growth below. 6.10 The case for agricultural development. Several key insights emerge from this analysis. First, the extent to which the Ethiopian economy will be able to deliver poverty reduction critically hinges on the composition o f its economic growth, much more so than on the overall growth itself. In the absence of robust agricultural growth, there is virtually no decline in poverty (scenario l), decent growth in services and the industrial sector. despite Indeed, as Mellor (1995) observed, agriculture plays an important role in a country's structural transformation in its early stages not because o f its growth rate, but rather because o f its size in the economy. As a matter o f fact, agriculture i s likely to have an innately slower growth rate due to its structural parameters,205underscoring the importance o f the nature o f growth as opposed to growth per se. While trickle down o f growth innon-agriculture to those currently employed in agriculture cannot be excluded, a strong relationship betweengrowth in agricultural incomes o f households and poverty reduction has also been observed inthe micro data206and the observed poverty reduction from growth in the non-agricultural sector in Ethiopia has so far beenlimited. 202According to the UNPopulation Divisionprojection rates, total population will amount to 92.1millionpeople in2015 of which 19.2 million(or 20.8percent) will be living inurbancenters, comparedto 11.9million(or 16.7 percent) now. *03 In 1995, 24 percent of the working population in urban areas was involved in agriculture, five percent in industry and 72 percentinservices, comparedto 96, zero and four percentrespectively inrural areas. 204For example, were the agricultural sector able to grow at 4.6 as opposed to 4.1 percent annually over the next decade, poverty incidence would decline to 18.2percent. 205 Typically, agriculture's share of employment falls more slowly than its share of income, with the result that labor productivity lags behindthat of the non-agricultural sector. 206Dercon, 2002; Dercon, 2004. 124 6.11 Second, while these findings provide support to a central role in the strategic focus o f the GoE on agricultural development, they point at the same time to the failure o f the government's current policies in bringing about sufficient, broad-based, sustainable agricultural growth. With agricultural GDP growth continuously falling behind population growth over the past decade, a revision o f the current agricultural and development policies and public investment strategies i s called for. As Ethiopia still finds itself at the very beginningo f the structural transformationprocess, with 85 percent o f its population employed in agriculture, broad based agricultural development will have to be at the center of any poverty reduction strategy for years to come, even though it will have to go handinhand with relaxing supply constraints in the off-farm sector to maximize the virtuous linkages from agricultural growth as discussed fbrther below. 6.12 I s a decade of 4.1 percent agricultural growth feasible? Given that Ethiopia's agricultural system i s currently operating well within its production frontier, with extremely limited use o f modern inputs and tools, a 1.9 percent increase from its current historical growth rate seems feasible. To begin with, the agricultural sector could already grow one percentage point faster ifit managed to better protect itself from droughts and thereby reduced its negative growth rates by half. Furthermore, the micro evidence from Ethiopia also suggests that a decade o f 4.1 percent sustained agricultural growth should be well within reach. Such agricultural growth rates are also consistent with international experience. China for example, sustained 10 decadal periods o f above 4.5 percent agricultural GDP average growth2'' over the past 25 years and India and Vietnam have experienced several decadal periods o f more than 4 percent agricultural growth. Irrespectively, concerted efforts by all development stakeholders will be neededto sustain an agricultural growth rate o f more than 4 percent over the next decade and an appropriate institutional framework as well as massive public investment (in agricultural research and development, extension, soil conservation, irrigation, rural infrastructure, rural education and health) will be needed. In section 6.2 we will explore further how balanced growth across the different agricultural sub-sectors (staple crop production, livestock, non-traditional agricultural exports) and across different locations can generate the overall agricultural and GDP growth necessary to substantially reduce poverty inEthiopia through its multiplier effects. 6.13 The false debate of agriculture versus non-agriculture. However, to maintain the current historical growth rates in the non-agricultural sectors, and to facilitate the movement of farmers out o f agriculture into off-farm employment, further institutional and policy changes as well as public investment, especially in infrastructure, will be needed to permit the development o f a dynamic private sector. While non-agricultural growth (as well as economic growth more broadly) over the past 15 years has been largely driven by growth in the public sector (see Chapter l), be sustainable, agricultural growth will have to become to increasingly more important in fostering non-agricultural growth. An elastic supply response of the off-farm sector will indeed be necessary to maximize the multiplier and linkage effects between agriculture and non-agriculture which provide the rationale for agriculture led development (see Section 6.2). 207These numbersare based on 11year moving averages. 125 6.14 Excessive tax rates and poor tax administration, limited access to urban land (reflected in excessive rent), and lack of access to credit in addition to poor telecommunication services and unreliable electricity supplies feature as the more important constraints hampering the development of private enterprise.208 Clearly, substantial improvement in labor productivity in non-agriculture can be obtained through policy changes which do not place a burdenon limitedpublic financial resources. Moreover, public investment in rural education and rural infrastructure will not only stimulate the development o f the agricultural sector, it will also help generate much needed off-farm employment opportunities. Fromthis perspective, the debate about the relative returns from public investment in agriculture versus non-agriculture in terms o f poverty reduction seems ill-conceived. We will elaborate further on the theoretical, empirical and historical rationale for agricultural led development and the critical need for a balanced growth path in Section 6.2. 6.1.2 Aid and aid absorption 6.15 Additional resources will be requiredto finance the public investmentsneeded to reach the poverty MDG. Using a one sector macro-economic model Agknor, Bayraktar and El Aynaoui (2004) calculate that a permanent increase o f five percentage points o f the aid/GDP ratio (all donated as non-food aid) over the next decade would reduce poverty sufficiently to reach the poverty MDG. Increased aid is modeled to foster growth in private consumption and poverty reduction through a substantial increase inpublic investment which in turn leads to more and better infrastructure and increased labor productivity through enhanced education and health. The authors do not find evidence o f "Dutch disease" effects, in which rapid and large increases in a country's income through foreign currency inflows have negative impacts on unexpected sectors o f the economy. The model assumes distribution-neutral growth and the same growth across all sectors. In scenario 3, we apply the resultingprivate consumption growth rates uniformly to the 1995 household consumption distribution from 2004 onwards. The poverty headcount drops to 15.6 percent, well below the MDGtarget o f 19.2 (Figure 6.1). 208 A more detailed account of the obstacles to private business development is given inthe World BankiEDRI's Investment Climate Assessment in Ethiopia, 2004. 126 Figure 6.1:Economic Growth Scenarios and Poverty Reduction -5 45 40 E 35 6 2 30 . ~. High agricultural growth = 3 25 20 2 15 .................................................................. 2015 poverty target Increasedforeign aid a' 10 ............... ................................................. ~ ~ ~............................ ~~.................... 5 ..................................................................................................................... 0 1 6.16 Historically unseen levels of foreign aid would be necessary for Ethiopia to reach the poverty MDG. A five percentage point increase in the aid/GDP ratio would bring total net official development assistance (ODA) progressively up to about US$32 per capita in 2015 from the 2002 level o f US$ 19.4.*09 This would entail a substantial mark-up in the assistance historically given to Ethiopia (US$13 per capita during the 1980s and US$16 per capita during the 1 9 9 0 ~and would bringEthiopia in line with the US$ 32 average per capita ) ~ ODA assistance given to Sub Saharan African countries over the past decade (1991-2002).210 However, the aid/GDP ratio would continue at the 2002 level o f 21-22 percent, which i s substantial both by historical and comparative standards. While such high levels o f aid/GDP have been sporadically observed in some smaller high aid African countries such as Malawi, Mozambique, and Rwanda, they have rarely been sustained for a decade, especially not in populous countries like Ethiopia, and have never been experienced in Ethiopia's recent history (since 198l), not even for one year. The average aid/GDP ratio for Sub Saharan African countries was 6.1 percent in 2002, the highest level ever during the past 30 years (Figure 6.2). The highest average level ever experienced by the Highly Indebted Poor Countries was 13.5 percent in 1994. Clearly, sustaining the aid level needed to reach the poverty MDG over the coming decade i s a tall order, especially giventhat Ethiopia finds itself already at its highest levels historically. 209Total net official development assistance includes both concessional loans and grants. In 2002, about 46 percent o f total assistance consisted of loans and 54 percent o f grants. In dollar terms, Ethiopia received about US$ 19.4 aid per capita in 2002, o f which US$ 8.9 was in the form o f loans, US$ 5.4 was development grants, and US$ 5 was relief aid. 2'oThe figures mentioned here are incurrent dollar terms (Le. not corrected for inflation or changes in exchange rate), Average per capita assistance in current dollar terms to Sub Saharan African countries during the 1980s i s US$25. 127 Figure 6.2: Net official development assistance per GDP for Ethiopia and Sub-Saharan Africa 1981-2002 ODA per GDP for Ethiopia and Sub-SaharanAfrica EPRDFcame power 25 20 g 15 5B 10 0 @ 5 0 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 199920002001 2002 Year 6.17 Even if the international community could be rallied to the cause, could such a massive inflow of aid2" yield the intendedeconomic growth and poverty reduction? The problems related to massive aid inflows such as Dutch disease effects, adverse effects on tax collection, and deteriorating govemance,212all o f which negatively affect growth and poverty reduction, are well known. The international experience documented by Collier and Dollar (2001) suggests that Ethiopia may not yet have reached the levels o f aid at which such phenomena start to affect growthnegatively (Box 6.1). 'I1 During the late 199Os, which coincided with the border war with Eritrea, aid levels stood at about US$ 10 per capita, or about 10 percent of GDP. During the immediate post Derg period in the early 1990s (1991-1994), Ethiopia receivedabout US$20per capita, which corresponded to about 15 percent o foverall GDP. Brautigam and Knack, 2004. 128 Box 6.1: The relationship between growth, aid, and institutions Based on 349 (four-year) growth episodes from 62 developing countries between 1974 and 1997, Collier and Dollar estimate the relationship between economic growth, aid and the quality o f the institutional and policy environment. They follow Knack and Keefer (1995) and use the Intemational Country Risk Guide Economic (ICRGE) indicators to capture the quality o f the institutional environment. They use the World Bank's Country Policy and Institutional Assessment scores (CPIA) as a proxy for the quality o f the country's policy and institutional framework for long term growth and poverty reduction in line with Easterly (1999).2'3 They find that the marginal effect of aid on economic growth per capita turns negative once the aid/GDP ratio exceeds 2.5 times the CPIAS2l4Given the current CPIA for Ethiopia o f 3,3, this implies that marginal retums to aid would tum negative at aid to purchasing power parity (PPP) GDP levels o f 8.2 percent. According to Collier and Dollar, the aid to real PPP GDP ratio for Ethiopia stood at 2.9 percent in 1996. Following the recent influx o f aid, it increased to 3.3 percent in2002. The pro osed scenario o f a permanent five percent increase inaid would fiuther increase this ratio to about 4.2 percent8 , still within the estimated 8.2 percent turning point beyond 5 ' ' which aid affects growth negatively, though diminishing retums may begin to set in. 6.18 While the empirical analysis by Collier and Dollar i s clearly not without its problemsY2l6 order o f magnitude o fthe results appears sufficiently large to suggest that aid the could still be productively used even when increased beyond its historical high levels, provided that the current policy framework is maintained (or improved). Similarly, the Aghor-Bayraktar-El Aynaoui model does not find any evidence o f "Dutch disease" effects, adverse effects o f aid on tax efforts, or declining marginal retums o f (non-food) aid on GDP. 213 Annually the World Bank assesses the quality o f the policy and institutional environment o f its client countries for long term growth and poverty reduction. The assessment i s comparativeirelative in nature and i s carried out by World Bank staff through an extensive internal consultative process. It assigns scores from two (unsatisfactory) to five (good) to 20 equally weighted components grouped in four broad areas: (1) issues o f macro-economic policies (e.g. inflation and macro-economic policies, fiscal policies, debt management); (2) issues related to structural policies (such as trade, financial sector policies, competitive environment for private sector development, factor and product markets, environmental policies); (3) policies for social inclusiodequity (gender, equity o f public resource use, human capital development, social protection and labor, monitoring and analysis o f poverty); and (4) public sector management issues and institutions (property rights and governance, quality o f budgetary and financial management, efficiency o f revenue mobilization, quality o f public administration, transparency, accountability and corruption). In case o f an extended period (more than three years) of a very low (=2) or a very high (=5) score on a particular area, the scores get recoded to one (=unsatisfactory for an extended period) or six =(good for an extended period) respectively. In2003, as in2002, the overall rating for Ethiopia i s 3.3. For comparison, in 2003, Mozambique scored 3.3, Malawi 3.4, Tanzania 3.8, Uganda 3.9 and Botswana 4.4. 214 Ga=0.185P-O.O72A with Ga denoting the marginal effect o f aid on (four-year average) GNP growth per capita, P, the policy level as captured by the CPIA index, and A, the aid/GDP ratio as captured by the net official development assistance (from OECD) divided by the real PPP GDP per capita from Summers and Heston (1991). The marginal effect of aid on economic growth turns negative if A/P > 0.185/0.072=2.5. Collier and Dollar estimated 1996 aid per real PPP GDP in Ethiopia at 2.9 percent. Using the World Development Indicator series on net official development assistance, population growth and real purchasing power per capita GDP, we observed that aid per capita increased f i o m US$ 14.04 per capita in 1996 to US$ 19.44 per capita in 2002, and that PPP per capita GDP rose from 568 to 693 respectively. Application o f the ratio [19.44/14.04]/[693/568] to the 1996 aid per real PPP GDP ratio o f 2.9 yields 3.3 percent. Similarly, assuming US$ 25 per capita aid as in scenario 3, we obtain an aid to real GDP ratio o f 4.2 percent, which is likely an overestimate as it does not account for the increase inGDP as a result o f increased aid. 2'6 While the analysis implicitly assumes that the causality runs from good policies and institutions to enhanced growth, it could equally run the other way, namely prospering economies cause good policies and institutions. The analysis does not control for this potential source o f endogeneity. This may bias the estimated coefficients if both are determinedby some unobserved third factor. 129 Nonetheless, caution i s warranted given that experience with high aid levels has been mixed.217 6.19 The shift to higher aid flows will need to be properly paced and sequenced to account for Ethiopia's absorption capacity. In sum, based on the available evidence, the simulations suggest that an increase ininternational aid by US$ 5 per capita or about US$ 350 million per year over the coming decade should enable Ethiopia to reduce its poverty head count by half from the 1990 level. Nonetheless, rallying the international community to this cause will pose a challenge, despite the relatively small amounts. Moreover, while admittedly small from an international perspective, the proposed aid flows are massive compared to the overall Ethiopian economy (21-22 percent of GDP) and highly unusual even in other highly indebted countries. The shift to higher aid flows will need to be properly paced and sequenced. While budgetary and program support programs are mechanisms by which substantial aid flows can be channeled at relatively low transaction cost for donors and government alike, the conditions for these aid modalities to be effective in reducing poverty depend critically on the government's own administrative capabilities to absorb this massive influx o f aid, as well as the private sector's capacity to supply the necessary goods and services. 6.20 Important processes have been set in motion to facilitate increasing fiscal flows. In particular, the intergovernmental system needed to facilitate fiscal flows across tiers o f government necessary to finance investment and recurrent costs o f service delivery has been made more transparent, and the responsibilities o f the different layers o f government have been clarified. First, the role o f the federal subsidy has been clarified and two additional specific purpose transfers from the federal to the regional governments have been created (the Public Sector Capacity Building Program [PSCAP] and the food security grants). Second, fiscal flows from regional to local governments are becoming more predictable and transparent through the system o f Woreda block grants and other municipality grants. Moreover, not only have the right channels been created to facilitate increasing fiscal flows, substantial efforts are also underway to strengthen the channels under PSCAP and the financial sector reform. The planning, budgeting, accounting, and procurement capacities o f the different layers o f government are being reinforced to ensure allocative efficiency o f increased aid flows within acceptable fiduciary risks. 6.21 Movingforward, substantialrisks remain. Implementationof the differentcapacity buildingand reform programs to reinforce the fiscal pipelineshas only just begun. Important constraints to make the current system o f government operations translate these additional aid flows into outputs and poverty reduction remain. They include among others: (1) the limited availability o f qualified manpower in government positions; (2) the limited ability to contract out the supply o f services to the indigenous private sector; and (3) the limited capacity o f the private sector itself to provide the necessary supply response. Furthermore, if aid flows are to increase along the lines outlined above, the tax generating capacity o f the different layers o f govemment needs to be strengthened to stem potential relaxation o f tax collection efforts. Clearly, during the first years of increased aid flows, much attention will have to go to strengthening the capacity o f the public and private sector. The processes set inmotion to do 2'7 BrautigamandKnack, 2004. 130 so will need to be closely monitored. The role of reinforcing tertiary education inthis process will also be critical. Without well directed and properly functioning pipelines, increased aid flows needed to finance Ethiopia's massive investment needs may end up fueling wage inflation and corruption, but not poverty reduction. Moreover, the success o f increased aid flows in terms o f poverty reduction critically hinge on the assumption that a stable macro- economic and policy framework i s maintained and that aid actually translates into growth. Both tendencies will have to be equally monitoredvery closely. 6.22 Halvingthe povertyheadcount from its 1990 level by 2015 emerges as a daunting task, which may not be feasible. Irrespectively, agriculturaldevelopmentand increased aid flows to help finance the necessary public investments will have to be core components of any roadtowards achievingthis objective. The task poses fierce challenges to the Ethiopian government and its development partners. Not only must the institutional architecture to channel large aid flows be strengthened,difficult choices interms of the nature and location o f public investments as well as in terms o f economic and institutional policies will have to be made to promote both agricultural and non-agricultural development to facilitate the exodus of labor and resources out of agriculture into non-agriculture as the former grows. Before returning to the micro-evidence presented in Chapter 4 to provide further guidance as to the optimal composition of the public investmentportfolio to reduce poverty, we first explore further inthe next section the role o f agriculture, its sub sectors, and its critical linkages with non-agriculture in fostering poverty reduction. 6.2 Toward a FeasiblePro-Poor Agricultural Growth Strategy 6.23 The macro-simulations presentedabove clearly indicate that it is hard to conceive how even spectacular growth inthe non-agricultural sectors could be sufficient to generate enough employment to pull the rural poor, which make up 85 to 90 percent o f all poor people in Ethiopia, out o f poverty. Nonetheless, it remains useful to review the rationale behind agriculture led development industrialization as a poverty reduction strategy, to explore the assumptions underpinning this strategy and their validity in the Ethiopian context. Furthermore, agricultural growth can be brought about in many different ways, and the poverty reducing effects o f different strategies are likely to differ. Agricultural strategies may for example differ in their emphasis on location (high versus low potential), the nature o f production (food crops versus cash crops or livestock), the appropriate technology (imgation versus agricultural intensification through adoption o f modem inputs and soil conservation techniques), the optimal farm size (large commercial farms versus family farms), or the role ascribed to the public and the private sector. 6.24 This section will reflect on these strategic issues, taking into account the structural features o f the Ethiopian economy such as the net market position o f rural households, their cereal demand behavior as captured by cereal price and income elasticities o f household demand for cereals, as well as differences in comparative advantage across space. In light o f the collapse of maize prices in 2001/2 following a bumper crop and mismanagement of food aid, special attention will go to the need to avoid extreme food price fluctuations for ADLI to maximize poverty reduction. Attention will also be given to the need to develop markets and balance growth in staple crop production with growth in the non-food and non-agricultural 131 sectors to keep up demand for food and thus prevent cereal prices from collapsing. The role o f food aid inthis context will also be highlighted. 6.2.1 The theoreticalcase for agriculture leddevelopment 6.25 The contribution of increasedagricultural productivity to economic growth and poverty reduction works through consumption and production linkages. Inparticular, a productivity increase innon-tradable activities such as cereal production leads to lower prices, effectively increasing consumers' real incomes. There will be important direct gains through decreased food prices for all net cereal buyers, which in most years make up the majority o f the Ethiopian population. However, the greatest benefits are usually indirect, through the consumption linkages, whereby the increase in people's real income stimulates the demand for locally produced goods and services which in turn generates employment and subsequently increases the demand for food as well. To generate sizeable multiplier effects, the income elasticity for locally produced goods and services must be large, and local supply o f these non-food non-tradables must be elastic and labor intensive. Moreover, the productivity gain must concern a non-tradable with a high average budget share such as cereals, which constitute about 30 to 40 percent o f total expenditures among the poor. However, net cereal sellers could potentially lose ifdemandis inelastic. 6.26 Production linkages can occur when increased productivity or higher prices in the productionof tradables positivelyaffect the incomes of producers. The direct poverty reducing effect may be substantial if the assets necessary for production of these goods are equally distributed and access to complementary inputs (e.g. fertilizer, improved seeds) i s universal. To maximize the poverty reducing effect, promoted technologies should be scale- neutral and labor intensive. Multiplier effects through backward linkages (increased demand for inputs) are usually limited, since inputs are generally capital intensive and imported. Nonetheless, important externalities may exist through increased local availability o f inputs for other (non-cash crop) agricultural activities. When production does not happen by the poor themselves, it must have a high labor content to have strong poverty reducing effects. Apart from primary resource extraction (mining, forestry, fishery) for which there appear only limited opportunities in Ethiopia, it is hard to imagine many other non-farm activities which engender broad employment opportunities in Ethiopia at this stage. Expansion o f the production o f agricultural tradables (e.g. coffee, chat, horti- and floriculture and livestock) offers much more potential with direct gains from increased income and employment opportunities for the poor as well as gains in forward linkages (e.g. employment opportunities inprocessing). 6.27 There are important synergies from a simultaneous pursuit of productivity increases and market development in both agricultural tradables and non-tradables. While progress in cash crop production technologies offers important opportunities for poverty reduction, greater opportunities for the poor are usually expected from combining consumption linkages resulting from productivity increases in non-tradable food production. The majority o f households inrural Ethiopia are subsistence farmers and net food buyers. So are urban households. Food price decreases following technological change thus hold the promise of substantially increasing their incomes. Yet, in the face o f inelastic demand for cereals, (reversible) productivity increases may lead to sharp declines and fluctuations in food 132 prices, rendering the obtained price decline unsustainable, thereby eroding the poverty reducing effects from earlier reversible technology adoption. To prevent cereal prices from collapsing, a parallel increase in income and demand for food through growth inthe non-food (e.g. agricultural tradable) sector, reduction o f transaction costs through better development o f the market channels to cater to this increase in food demand, and better management o f food aid, must be pursued. We discuss in more detail below how these issues play out in the Ethiopian context. A more comprehensive discussion of the consumption and production linkages arising from increased agricultural productivity including further empirical and historical evidence i s provided inAppendix 4. 133 Box 6.2: Reflectionson four common critiques to agriculture led developmentin Sub-SaharanAfrica 1. The obstacles to agricultural development in Africa are so severe that agriculture is not viable as an engine of development. Major obstacles to agricultural development do exist due to limited water availability, soil deterioration, agro- ecological heterogeneity, lack o f agricultural research and information, population density not adequate to stimulate intensification in many areas, and difficulties in market access. However, these obstacles are not insurmountable. For example, irrigation inAfrica has often failed due partly to social, political and institutional reasons (e.g. poor management, gender discrimination, violent conflict) which can be overcome in time given a policy environment that supports transition from a traditional to a modern economy. In this context it is important to note that Ethiopia has been involved in internal or external conflicts totaling more than two decades over the past 40 years, a factor which has clearly hampered the development o f irrigation and the development o f the private sector more broadly. Moreover, many o f the obstacles that constrain the agricultural sector also limit the returns to other sectors. As such, the key pillars o f most stimulus packages, including improvements in transport, communication and financial infrastructure, are not sector specific. 2. As intensijkation proceeds, smallholder farms will be at an increasing disadvantage relative to large commercial farms due to economies of scale. However, a large farm based development strategy for Africa would be less equitable, and slower to deliver poverty reduction. The evidence on this is mixed. While transaction costs related to the monitoring and enforcement o f labor contracts inagriculture rapidly increase with the number o f workers, as intensification proceeds scale advantages in markets for outputs, inputs, and finance grow in importance relative to smallholders' labor advantage. To fully exploit the welfare creating potential o f smallholder agriculture, it is necessary to have policies and institutional supports (e.g. marketing cooperatives) to reduce the disadvantage suffered by smallholders in non- labor markets. In this regard, policy makers should begin to structure the smallholder sub-sector to take advantage o f the growing preference for organic produce and fair trade witnessed in key Western food markets. 3. An agriculture-based development strategy will not reach the rural poor, as they live in marginal areas, and/or have land holdings too small to provide more than a secondary contribution to livelihoods. First, it must be emphasized that not all areas are low potential-this is also true for Ethiopia-and different strategies may have to be pursued depending on the area's comparative advantage. Second, even in the low potential areas, agriculture must be compared with alternative ways to generate rural incomes. Inmost marginal rural areas, there are few viable alternatives: diversified (or diversifiable) rural economies with attractive alternatives to agriculture are the exception rather than the rule. While education and fostering migration provides one alternative, poor area development i s often found to be more pro-poor. The finding that higher yields and agricultural income are also possible on smaller farms in food insecure areas i s important, as this indicates that some development in these areas is still possible through agricultural intensification. There are basically two broad strategies to help with so-called low potential areas. One strategy seeks to promote out- migration toward high potential areas (and urban centers). This strategy was followed by Ethiopia in the 1970s and 198Os, and has often been promoted in other parts o f the world, including Punjab in India, Central Luzon in the Philippines, and Sonora in Mexico. Yet, this strategy is not without risk either, as greater dependence on a smaller area increases the risk o f nationwide harvest failure in case o f drought. Secondly, massive migration may lead to rapid overcrowding and increasing poverty in the high potential areas. A balanced strategy focused on promoting agricultural intensification and risk reducing inputs (irrigation, pest management, soil conservation), appropriate crop mixes, diversification out o f agriculture, and public work programs may thus be more 4. It is diflicult to envisage in what exportable agricultural products Africa will be able to develop competitive advantage on a large scale. Presently, Africa still holds competitive advantages in a range o f unprocessed primary products. Nevertheless, it will be important for policy makers to invest resources into improving the institutional environment supporting agriculture to boost productivity. This i s especially important as the market evolves to favor high levels of capital investment, information, and supervision as product specifications continue to get ever more complex, a trend that promises to diminish any existing competitive advantages. The recent boom in Ethiopia's floriculture for export illustrates that it holds competitive advantage in agricultural exportables. The role o f the government infacilitating investment in this sector by private entrepreneurs (as opposed to public investment per se) has been critical. Source: adaptedfrom Kydd et al., 2001 2`8 Lipton and Ravallion, 1995. 134 6.2.2 Patterns of agriculturalgrowth and poverty reduction 6.28 While it is important to qualitatively understand the channels through which productivity growth in the tradable and non-tradable agricultural sectors affect economic growth and poverty reduction, the size o f the effects i s ultimately an empirical matter, which depends on the structure o f the rural economy and the rural-urban linkages. Computative General Equilibrium Models are frequently used to model the economy and simulate the effect o f change in agricultural productivity in different sectors on poverty. This i s the approach taken in the World Bank Country Economic Memorandum, though it does not include an explicit consideration o f the poverty effect^.^" Here we limit ourselves to a discussion o f some key features of the rural economy in Ethiopia such as households' net market position with respect to cereals (net sellerbuyer) as well as their cereal demand and supply behavior, both o f which are critical in gauging the effect o f an increase in the productivity o f staplehon-tradable crop and non-staplehradable crop production on poverty. Based on the multi-market model by Diao, et al., (2004) we subsequently explore the poverty reducing effects o f productivity gains in different agricultural sub-sectors, separately as well as combined, and discuss the results in relation to the poverty MDG. The section concludes by reflecting on the optimality o f different agricultural growth patterns across different geographical locations. Evidence suggests significant numbers of net cereal buying poor households in rural Ethiopia 6.29 From our earlier discussion o f the size o f households' landholdings (Chapter 4) we recall that landholdings are increasingly too small for households to meet their cereal needs at current average production technology, which challenges conventional wisdom that the large majority o f the rural poor are net sellers o f grain and would therefore benefit from high cereal prices, If there are significant numbers o f net cereal buyers among poor households in rural areas, policy measures aimed at fostering a gradual decline in cereal prices (such as staple crop productivity increasing measures) would be welfare enhancing for those poor households, as well as for the urban poor. It i s thus critical to obtain solid empirical evidence o f the net cereal market position o f rural households in Ethiopia. We explore this question using (1) nationally representative data which provide direct information on households' cereal market position during 1995/96, and (2) indirect information on households' cereal production potential compared to their average cereal needs. These results are contrasted with those o f other African countries. We further provide a brief profile o f net buyers and sellers according to their location and welfare levels. 6.30 Net cereal buyers may outnumber net cereal sellers in rural Ethiopia. To explore the net market position o f rural households we examined the food security survey conducted by the Ministry o f Economic Development and Cooperation (MEDAC) and the CSA inJune 1996. Respondents were asked a detailed set o f questions regarding their production, sales and purchases o f the six major grains (maize, wheat, sorghum, teff, barley and millet) between October 1995 and September 1996. As the survey took place in June 1996, 2'9See also de Janvry and Sadoulet, 2002 for a stylized model o f African economies and the effect o f an increase intotal factor productivity inagriculture onpoverty. 135 respondents were asked to anticipate their sales and purchases between July and the end o f September. Table 6.2 categorizes each rural household as either net buyer, autarkic, or net seller on the basis o f the volume (kgs) and the value (ETB) o f their net sales220(purchases - sales) duringthe survey period, aggregated across all six grains.221 Table 6.2: Cereal market positionof rural householdsin 1995/96 Position Market Percent Production Purchases Sales Net Sales Net Buyers 53.75 506 585 44 -541 Volume Autarkic (Kgs) 8.09 629 23 23 0 Net Sellers 38.16 1472 75 533 458 l._" __._--___-__._.I I ~ ________-__l__l__,_l__l_ll.ll____l_..li._. N e t Buyers 53.23 556 782 65 -717 Value Autarkic (Birr) 7.86 593 0 0 0 Net Sellers 38.91 1579 88 601 513 Source: Own calculations from CSA Food Security Survey, 1996 6.31 Fifty-three percent of rural households are net buyers o f grain, while only 38 percent are net sellers. Or, in 1996, an average production year, there were 15 percentage points more net cereal buyers inrural areas than there were net cereal sellers. The remaining eight percent o f households were autarkic. This clearly goes against the prevailing belief that the majority o f rural households are net cereal sellers. Looking at values versus volumes does not change the emerging picture. Inthe remainder of the text the net buyedseller classification will be based on values.222 Note furthermore that the average sales volumedvalue among net buyers i s very small (around 10 percent o f their production) indicating that these classifications are not driven by "distress" sales at low prices immediately after the harvest to repay credit and relax liquidity constraints and repurchases at highprices during the hunger season. Also, the limited sales volumes among the net buyers suggest that while substitution among cereals may take place (e.g. sell teff to buy barley), this i s not driving these results. In other words, the majority of rural (in addition to urban) households would stand to benefit from gradual declines in cereal prices. Autarkic households are consuming far less than the others suggesting that they are also likely to benefit from a price decrease. Appendix 5 elaborates further on the effect o f intra-annualprice fluctuations on households' marketposition, andthe incidence o f substitution across cereals given its policy implications in terms o f credit access and repayment modalities, as well as the potential focus o f agricultural research and extension on different cereals. 220 Sales do not include food aid. Food aid is fairly evenly distributedamong net cereal buyers and sellers (Table A.6.1 inAppendix 3). 22' For volumes, net buyers (sellers) are d e f i e d as households who purchase (sell) more kilos o f all grains than they sell (buy). Kilos of all grains have been treated equally. To calculate values, we use individual producer (retail) grain prices to value sales (purchases) and define net buyers (sellers) as those households with less (more) total grain expenditure than revenues. Autarkic households are those with zero net sales inboth volume and value. 222 While there i s slight variation inthe outcomes between volume and value based definitions of market position due to price differences in producer and retail prices, positional and relative measures are fairly robust to the definition used. The correlation coefficient betweenboth variables i s 0.94, indicating a highdegree o f sameness. 136 6.32 Livestock, off-farm wages and businesses provide sources of cash income to buy food. Given that rural households largely depend on agriculture, it is natural to ask from where this large group of net cereal buyers get their cash income from. While reliable nationally representative data on income i s currently not available in Ethiopia, the direct and indirect evidence from three case studies show that especially livestock, but also off-farm work and income from business activities are important sources o f cash income for rural households. Table 6.3 presents a break-down o f income into the various components across six different communities inEthiopia that differ across agro-ecological conditions and market access. While crop income provides usually more than 50 percent o f household income, we find that on average most households obtain income from a plurality o f sources. Important alternative sources o f cash income include earnings from sales o f livestock or their products, from off-fann wages, or from business profits. Table 6.3: Incomeper capita and incomesourcesacross several communities Debre Koro- Adele Gara- Berhan Dinki Degaga Keke Goro Domma (67 hhs) (54 hhs) (89~hhs) (60 hhs) (56 hhs) (79 hhs) Meanannualpc. income (ETB) 236 145 71 163 46 40r Meanannualper capitainc (US$) 114 70 34 79 22 19 Composition of incomes (%) Crops (includingsubsistence) 43 78 54 51 66 73 Livestockproducts 14 0 0 8 4 0 LivestockLive Sales 26 4 6 13 2 I O Off- farm(wage &business) 16 10 38 24 19 7 Remittances 2 6 1 5 10 8 Source: Dercon and Krishnan, 1996 6.33 Grain trading, collection of water and fuel wood, food for work, weaving, pottery, beer brewing, and government employment are the more important off-farm activities. Analysis of the activities o f the 1,477 households in the 1994 E M S data shows that o f the 4,321 adults in the sample (> 14 years), 71 percent were either full time farmers or domestic workers not involved in any kind o f off-farm activity or paid work. Another 19 percent were mainly involved in farming or domestic work, but they also participated in off- farm activities either as wage laborer (7 percent) or other income earning activities (12 percent). About half o f those involved in part-time wage work were involved in food for work and about half of those involved inpart time income eaming activities were involved in collecting and selling wood, water or fuel; with about 20 percent involved in grain trading. Only 10 percent were full-time off-fann workers. About one third were self employed (potters, weavers, processed food sellers), one third o fthem were traders, and 15 percent were government employees (government officials/administrators, teachers, soldiers). Average monthly salaries for adult males who were part time wage workers, part time income earners and non-farmers amounted to 109, 178, and 234 ETB respectively.223 6.34 Evidence from TigraJ24 shows that while cereal production is the primary source o f income for 97 percent o f households, they usually also have other sources o f income. In about one-third o f the households the secondary source o f income i s cattle production. Production o f other crops, including perishable annual and perennial crops, i s the secondary 223Verwimp, 1999. 224Pender andGebremedhin,2004. 137 source of income in about seven percent o f households. Non-farm activities are an important source o f additional income for about one quarter of the population, including trading activities (6.5 percent o f households), food for work (6 percent o f households), salary employment (1.6 percent) and other activities (handicrafts, brewing beer, priest, local officials) (10 percent). About one fifth o f the sample did not have a secondary source o f income. 6.35 The observed evidence on the composition of rural households' income portfolios i s consistent with the finding of a large number of net cereal buyers in rural areas. While the available evidence indicates that cereal production i s the major activity for most rural households, it also shows that rural households usually hold some livestock and that they are often engaged in off-farm activities which help generate cash income to buy cereals. Furthermore, income from cash crops (coffee, chat) also forms an important source o f cash income insome regions. 6.36 Rural households' net cereal market positionfluctuates over time, though even in good agricultural years a non-negligible number of rural households are net cereal buyers. Our empirical results on rural households' net cereal marketpositionpresentedso far are based on experience in 1995/96, a reasonable agricultural year. It i s however likely that people's net cereal market position will change annually depending on the rainfall patterns that year. Nonetheless, a substantial portion o f the rural population i s made up o f net cereal buyers, even ingood years (see Box 6.3). 138 Box 6.3: Rural households' cereal market positions To investigate how rural households' net cereal market positionfluctuates over time, we present households' expectations regarding their expected market position for each o f the different cereals, assuming that they were either faced with a good, an average, or a bad cropping season (Table B6.3.1).225 As expected, the share o f households in a particular marketing season i s clearly a function o f the year's cropping conditions. The proportion o f net buying households increases from about one-third to about one-half the rural population when going from very good to good years and from one-half to more than three-quarters o f the population when going from good to bad years. Table B6.3.1: Households' expected net cereal market position given annual cropping conditions Expected Good Averaae .. Bad Market Position Per Percent Percent of Percent per Percent of Percent Percent of per cereal total buylng cereal total buying per cereal total buylng Cereal per transaction and selling transaction and selllng transaction and selling household transactions transactions transactions Net Buyers 2o o8 31 39 31 97 53 99 56 27 79 83 Autarkic 36,03 40.79 29.51 Net Sellers 43,89 68.61 27.25 46.01 14.22 20.17 Tota' 100.00 63.97 100.00 59.22 100.00 70.49 Source: Own calculationsfiom CSA Food Security Survey, 1996 6.37 The finding that there is a large number of net cereal buyers in rural Ethiopia is consistent with evidence from other surveys in Ethiopia and in line with international experience. The findings regarding household's net cereal market positions in Ethiopia are also consistent with the more indirect estimates derived from the relationship between aggregate land productivity and mean land sizes in rural areas. Calculations show that to meet the daily minimum cereal requirement given current dietary patterns, at least 0.148 ha per person i s needed for a household to be self sufficient in cereals. Only 48 percent o f the rural population has currently enough land to do so. In other words, more than half o f the rural population are currently net buyers o f and these may also be the poorer as poverty and land size are closely correlated especially among the lower half o f the land distribution.227 Add to this the urban population and on average close to 60 percent o f the 225Before turning to a more detailed interpretation o f Table 6.4, it must be noted that a much larger fraction o f the households are seemingly autarkic compared to the actual market positions reported above. This follows from the fact that we report here a household's expected position per cereal as opposed to its expected position across all cereals per household. Information on household's net cereal market position could not be compiled from the reported expected positions which were per cereal. While many more households are likely to be autarkic for particular grains, this will not hold when aggregating across the different cereals. To facilitate interpretation o f the table we will assume that proportion o f net buying (selling) transactions out o f the total cereal transactions corresponds to the proportion of net buying (selling) households (see column 3, 5, 7). This provides a reasonable approximation o f the expected changes in the proportion of net cereal buying and net cereal selling households across seasons, though it likely leads us to overestimate the proportion o f net selling households. To see this, note that the actual proportion o f net buying, autarkic and net selling households in 199516 was estimated at 53, 8 and 38 percent respectively, while the expected proportion o f net buying and net selling households in an average year according to our estimation i s 53 and 46 percent respectively. 226Evidence from a nationally representative sample o f 4,338 households conducted between October 1995 and June 1996 estimated that 48.2 percent o f the rural households inEthiopia were net buyers o f cereals, even though the 1995196 was a good production year (Demeke, Said, and Jayne, 1997). 227Jape et al., 2003. See also Section 6.2.2. 139 Ethiopian population are actually net buyers o f cereals.228These numbers on the existence o f a large group o f net staple crop buyers among rural households are very much inline with the empirical evidence from other Sub Saharan African countries (Table 6.4). The large group o f net cereal buyerswould directly benefit from lower cereal prices.229 Table 6.4: Market participation profile for rural households in selected African locations Sales Concentration Indicator Location Market Involvement Indicator Percent of Total (Percent of Total Market CroP(s) Production Marketed Sales) (year) Net Buyers _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 70 80 (yoofhoueholds) ____ _________ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Net ______ _ (yo)_ _ _ _ ____ no^,^^^^^^ (Excluding Gifts) Or Sellers 50 ~____(9/. ofhoueholds) ____ ~ Mali Course grain 39 13 48 8 8 16 23 (1985-86) Senegal Course grain 30 40 29 5 7 11 15 (1986-87) Somalia M a u e 61 0 39 57 13 20 (1986-87) Rwanda Beans 73 5 22 I O 2 4 6 (1986 -87) Sorghum 67 9 24 21 2 4 5 Zimbabwe Maize 15-25 18-30 67-45 40 I O (1984-85) (Communal sector) Source: Weber et al., 1988 6.38 The bulk of the marketedsurplus is producedby a minorityof producers,also in Ethiopia. Based on the current distribution o f land it can be shown that the majority o f the marketed surplus o f cereals i s produced and sold by a minority o f larger surplus producers. In particular, back o f the envelope calculations suggest that 80 percent o f the marketed surplus i s produced and sold by only 20 percent o f the producers.230 These findings are consistent with observed market behavior by rural households in 1995/96, which shows that 18 percent o f the net selling households capture 68 percent o f the total sales (Table 6.5). The pattern is also in line with the sales concentration o f marketed cereals observed in other poorer African countries (Table 6.4). These statistics are critical to price-policy debates for agricultural commodities, and suggest that measures to keep prices above the market rate would be tantamount to subsidizing a small fraction o f farmers who produce the bulk o f marketed output. Even in countries like Zimbabwe where between 45 and 67 percent o f rural households are net sellers o f maize, 70 percent o f these sales accrue to only 10 percent o f the sellers. 6.39 Moreover, the bulk of net purchases are bought by a small percentage of the net buyinghouseholds. To gauge the welfare effects o f cereal price evolution it is also important to further characterize the net buying households. To do so, we split net buyers (and net sellers) into quintiles o f net sales values. As with cereal sales we also find that the bulk o f the net purchases are concentrated among a small group o f net buying households. While net buyers in the lowest quintile o f net purchases only account for 0.74 percent o f the total purchases, net buyers in the top quintile account for 80 percent o f the total purchases (Table 228 This is likely to be an underestimate as 15 to 20 percent o fthe cultivated land is allocated to permanent crops and not all land allocated to temporary crops i s allocated to cereals. A small amount i s also cultivated with oilseeds. 229 Evidence from the ERHS indicates that the households in the sample were on average 10 weeks per year without homegrown food, with 31 percent o f the households more than three months per year without homegrown food. Many people depend on the market at least some part of the year (Dercon, 2002). 230 It i s further estimated that 40 percent o f the marketed surplus i s produced by only five percent of the farmers. 140 6.5). Inother words, both the supply and demand sides o f the market are heavily concentrated among a small subset o f richer households inrural Ethiopia. Table 6.5: Profile of market position and degree of concentration by net sales quintiles __ ____ Values Net Percent Sales Market Position Mean Percent Share o f Quintiles hhs per No' Of Ofhhs PAE Harvest Dry Market Quintile Quintile per Expend 1 Net Buyers 427 18 1103 -4 -27 0.74 Net Sellers 306 21 1113 40 -11 1.20 2 Net Buyers 434 19 1155 -41 -60 2.52 Net Sellers 287 20 1104 108 -8 4.00 3 Net Buyers 453 19 1131 -97 -113 5.40 Net Sellers 295 20 1121 182 30 8.78 4 Net Buyers 458 20 1202 -235 -197 11.24 Net Sellers 293 20 1168 354 87 17.84 5 Net Buyers 563 24 1251 -769 -2043 80.11 Net Sellers 265 18 1212 1559 232 68.19 Autarkic 306 100 1049 0.00 0.00 Source: Own calculationsfi-om CSA Food Security Survey, 1996 6.40 While data do not permit an estimate of the number of net cereal buying households which are poor, evidence suggests that an important subset of poor households would stand to gain from a gradual decline in cereal prices. Further characterization o f households by their net cereal marketing position and wealth shows that net buying households are more concentrated in the richest quintile, while net selling households are more concentrated in the fourth quintile (Table 6.6. These net buyers are most likely rural households who eam most of their living out of farming (traders, self-employed, government officials) and who tend to be richer, while the rich net sellers are likely households with larger landholdings. However, despite the fact that not strong correlations could be observed between wealth and poverty overall, given that net buyers outnumber net sellers among the poorest two quintiles, the rural poor would still stand to lose more fi-om an increase incereal prices. Table 6.6: Net market position by welfare level in 1995 Average proportion per Volume o f net Value o f net Quintiles Number of households of cereal sales per cereal sales per Average P A E household (kgs) household (ETB) Expenditure') Of Expenditure') Net Autar Net Net Autar Net Net Net Buyers Net Net buyers -kic sellers buyers -kic sellers sellers Buyers Sellers 1 630 898 132 521 58 9 34 -367 338 -500 460 2 926 796 81 508 57 6 37 -592 374 -753 436 3 1188 820 106 541 56 7 37 -380 341 -462 469 4 1529 765 96 576 53 7 40 -637 690 -848 760 5 2524 731 59 319 66 5 29 -437 366 -565 446 ''Average 1995 expenditure per adult equivalent per community inETB. 141 6.41 While significant numbers of (poor) households stand to benefit from a cereal price decline, large price fluctuations must be avoided for price declines to be sustainable, and welfare gains will be much more substantialif prices decline gradually. Large price fluctuations could come about as follows: Ifprices decline rapidly in the face o f (reversible) productivity increases and price inelastic demand, net cereal sellers will be hurt, which would induce them to reduce their use o f inputs and thus overall production in the subsequent year; a reversal o f productivity and overall production will lead to highprices in the subsequent year, given price inelastic demand, and thus large price fluctuations. To the extent that the productivity increases are irreversible (e.g. agricultural intensification through investment in irrigation, soil conservation, or better cultivation techniques as opposed to use o f modem inputs), adopting farmers will inthe face o f inelastic demand see most o f the gains from technological change being transferred to rural and urban consumers through falling prices, which will force them to either further increase their productivity to stay afloat, a phenomenon known as the price treadmill, or else leave agriculture. Mismanagement o f food aid may further compound the occurrence o fbothphenomena. 6.42 An increase in income through, for example, a productivity or production increase in non-foodproductionmay attenuate these phenomena (price fluctuations and price treadmill). The extent to which this will be effective depends largely on the price elasticities o f cereal supply and cereal demand as well as the income elasticity of cereal demand. Given that these behavioral parameters may differ across cereals, the threat o f price fluctuations and the phenomenon o f a price treadmill will differ somewhat across cereals as well. For example, if productivity gains in staple crop production are accompanied with productivity and income gains innon-staple crop production (e.g. livestock sector or off-farm employment opportunities), an increase inthe demand for staple crops will (partly) offset the decline in food prices resulting from the increase in food crop productivity. This underscores the critical importance of balanced growth inboththe food and the non-food sectors. Before we discuss how these different forces play out inthe Ethiopian context based on multi-market model simulations, we first present the key parameters describing the demand behavior o f rural and urban households for the different cereals. This will help us gauge the potential for large cereal price fluctuations and the danger o f the price treadmill following increases in agricultural productivity in a qualitative manner, and it will also help us to interpret the simulation outcomes from the multi-market model. Cereal demand behavior in Ethiopia,pricefluctuations and theprice tread mill 6.43 Demand for cereals is price inelastic, but there are substantial differences across cereals and location. Demandfor cereals is more price inelastic in urban areas than in rural areas.231The demand for teff (and also wheat) i s the least price inelastic (around -0.7 and -0.5 respectively) indicating that household consumption i s relatively more responsive to changes intheir availability and price. Both crops are also well traded inthe market (see Table 6.7). 231Demand is own price inelastic ifit is below one. This implies that a 10percent increase inprices would lead to a less than 10 percent decrease in demand. For example, if own price elasticity o f the demand for maize i s - 0.41, then an increase in the maize price by 10 percent would lead to a decrease in the aggregate demand for maize by 4.1 percent. Demand for goods tend to more price inelastic, i.e. less sensitive to price, the less substitutes there are, and the smaller the share o f total expenditures it takes. 142 The demand for barley and sorghum i s very price inelastic (estimated at around -0.1 to -0.2), followed by millet. The price elasticity for maize, a much traded crop i s estimated at -0.4. 6.44 Income elasticity is especially high for teff and wheat (around one and 0.7 respectively in rural areas), around 0.5 for maize and around 0.2-0.3 for sorghum and barley. This implies that increased income will especially result in increased demand for teff and wheat, which make up the largest share o f cereal expenditures inurban areas. While both make up about eight and seven percent respectively o f total expenditures inrural areas, maize i s more important in rural diets, making up 10 percent o f total expenditures on average (compared to only two percent in urban areas). Moreover, in both rural and urban areas the share o f maize intotal expenditures increases relative to wheat and teff the poorer households become.232 Table 6.7: Estimated cereal demand and supply elasticities in Ethiopia Demand'' Supply Ownprice elasticity Income elasticity ownprice elasticity Rural urban rural urban maize -0.41 -0.22 0.53 0.27 0.60 wheat -0.53 -0.29 0.70 0.35 0.20 teff -0.72 -0.44 0.96 0.53 0.40 sorghum -0.17 -0.13 0.21 0.16 0.20 barley -0.24 -0.14 0.32 0.17 millet -0.38 -0.22 0.54 0.27 T, Average elasticities across all zones estimated from Linear Expenditure Systems by Diao et al., 2004; differences in elasticities across zones were minor. Estimates inhigher potential areas from Abrar, Morrissey and Rayner, 2004. 6.45 In sum, while especially teff, but also wheat, emerge as the preferred cereals, they are also two of the more expensive sources of calories (see Table (A5)l Appendix 5). The fact that fertilizer i s also mostly used on these two crops and that they are also more widely marketed i s consistent with the observed demand behavior. However, their marketing pattern differs substantially. The majority o f marketed wheat is produced by a small number o f producers. On the contrary, it i s the consumption (as opposed to the production) o f teff which appears more concentrated among a smaller number o f (richer) rural farmers, consistent with its high income elasticity, and the number o f net selling teff farmers actually outnumbers the number o f net buying households (Table (A5)4, Appendix 5). In other words, teff i s likely a cash crop for many farmers. This i s consistent with the observed price elasticities o f supply, which are higher for teff than for wheat. 6.46 Maize, on the other hand, is generally less preferred (lower income elasticity), but also a cheaper source of calories (see Table (A5)l Appendix 5), and as a result it i s more consumed by poorer households. Maize generates higher yields than all the other cereals and i s more responsive to improved seeds and fertilizer application, though given that it i s a long maturing crop, it i s also more prone to drought shocks. Given its responsiveness to modern inputs, supply is also more price elastic. As a result, it i s widely marketed, with the production o f the marketed surplus concentrated in the hands o f smaller group, and the net 232 Diao, et al., 2004, Table 13. 143 buyinghouseholds of maize outnumberingthe net selling households (Table (A5)4 Appendix 5). 6.47 Sorghumand barley233emerge as the less preferredcaloric sources and are likely important subsistence crops, especially in the drylands, even though they are also traded, most likely more locally. Demand for sorghum and barley i s least responsive to prices and income changes, and their supply i s much less responsive to price changes or changes in market access.234 6.48 The danger of large price fluctuations or a price treadmill occurring in response to a reversible (e.g. modern input use) or irreversible (e.g. investment in irrigation) increase in productivity differs across crops, with maize most exposed to these threats and wheat and teff least exposed. According to the available evidence, the demand for maize i s price inelastic (around -0.4). As a result, an increase in the productivity o f maize production i s likely to induce a substantial decline in maize prices. While this i s likely to have a positive welfare effect in the first round, given that net maize buyers outnumber net maize sellers this i s likely unsustainable. The productivity increase may not suffice to compensate net sellers for the price decline, inducing them to reduce their production through diminisheduse o f modern inputs during the next cropping season. This has the potential to lead to a stark increase in prices during the next season and unstable price patterns more generally. This i s consistent with the observed pattern in maize prices in 2000-2001, when bumper crops resulted in a collapse o f maize prices, inducing farmers to use less fertilizer duringthe next cropping season, which compoundedby a drought shock ledto a large decline inoverall production and the threat o ffamine for up to 14millionpeople. Although food aid imports continuing in the face o f bumper crops exacerbated the collapse in maize prices235,it i s nonetheless important to note that maize i s especially susceptible to this phenomenon, given its particular responsiveness to the combined package o f fertilizer and improved seeds, allowing big gains in yields. 6.49 A simultaneous productivity increase in non-staple crop production will be necessary to facilitate the desired, gradual and sustainable decline in maize prices. While a decline inmaize prices is desirable, given that it constitutes a larger budget share in the diet o f poorer households and that there are twice as many net maize buyers than net maize sellers, a price collapse will have to be avoided for price declines to be sustainable. It will thus be important to increase productivity in the production o f non-staple tradables (e.g. livestock, coffee, chat, and other cash crops as well as non-agricultural products) in tandem with increases in maize productivity, to help raise households' incomes and foster their demand for maize, which will help offset the price decline resulting from the maize productivity increase. A decline in maize prices i s to be expected given that (1) demand for maize i s not very income elastic (0.5), (2) substantial gains in income will be needed to increase the demand for maize, and (3) such an increase indemand will usually only lead to a limited increase in maize prices since supply i s not as price inelastic as demand. A simultaneous productivity increase in non-staple production will thus be necessary to help 233Milletis the less important cerealofthe three. 234Abrar, Morrissey and Rayner, 2004. 235 See World Bank Country, 2004c, for a more detailed account o f the negative effect o f food aid on cereal prices duringthe 2001102 season. 144 facilitate a gradual and sustained decline inmaize prices. It will be equally important to better manage food aid through local purchases while at the same time fostering the development o f markets to lower transaction costs and facilitate interaction between sellers and buyers, 6.50 Teff, on the other hand, appearsto be less susceptible to a large price collapse in the face of an increase in productivity. First, given that the demand for teff is less price inelastic (around -0.7 in rural areas and around -0.5 in urban areas), an increase in productivity and production will not result in such a large decline in prices. Moreover, given a relatively high income elasticity (elasticity i s about one), increases in income will translate into substantial increases inthe demand for teff, which will help keep prices up. As there are more net seller transactions than there are net buyingtransactions for teff, and given that teff consumption i s relatively more important the richer the household, keeping prices up would appear to be welfare enhancing. Inother words, given current tastes and market conditions, increasing productivity o f teff (which i s more difficult to do given a lack o f improved varieties and less responsiveness to fertilizer) combined with simultaneous income enhancing measures outside staple crop production would pay off. This i s consistent with the fact that fertilizer is currently heavily used on teff and the observation that teff production i s highly elastic to increased market accessibility. 6.51 Demandfor wheat appearsto be moderatelypriceinelastic (-OS), and supply also appears to be price inelastic (0.2-0.3). Many more rural households are net wheat buyers than there are net wheat sellers, indicating that a gradual decline inprices would be desirable. Increasing productivity o f wheat would not lead to as large a decline in prices as increasing the productivity o f maize would, and given rather inelastic supply and relatively strong income elasticity (0.7), prices can be more easily kept up with simultaneous income growth due to activity expansion innon-staple crops. Inthis context, a continued focus on increasing productivity in wheat production (which seems quite feasible given the limited use o f improved seeds and irrigation)236would bejustified. 6.52 Demand and supply for sorghum and barley are price inelastic, and income elasticity is also low. Yet there are many more net buying transactions than net selling transactions, suggesting that the number o f households who would benefit from a decline in prices would exceed the number who would suffer, and these may also be the larger ones. Increase in productivity (e.g. through combined use o f fertilizer and improved seeds) would lead to a strong decline in prices and hurt the net sellers (who may be the larger farmers). Little demand stimulation i s to be expected from income growth outside the staple crop sector to dampen the decline in prices, given limited responsiveness o f supply to price and low income elasticities o f demand (especially for sorghum). Nonetheless, sorghum and barley are important staples in the rural diet, and many households appear to be net buyers and stand to benefit from a gradual and sustained decline inprices. To help sustain productivity increases, demand management could happen through local purchases o f these cereals as food aid. 236Wheat production appears most responsive to irrigation among the different cereals according to Diao et al., 2004. 145 6.53 In conclusion, this qualitativediscussion indicatesthe following: 1) A sustained gradual cereal price decline i s desirable from a poverty/welfare perspective; 2) Given that the demand for cereals i s price inelastic, productivity increases in cereal production could lead to strong declines in cereal prices, which would either (a) not be sustainable if brought about through reversible productivity increases (e.g. modem input use) or (b) lead to a price treadmill if brought about through irreversible productivity increases (e.g. irrigation); 3) These threats are most severe for maize, also for sorghum and barley, but less for teff and wheat; 4) Regardless o f these points, cereal price declines mustbe gradual to be sustainable andthus most effective in terms o f poverty reduction, which requires a balanced growth pattem whereby both an increase in staple crop productivity and an increase in non-staple tradables (livestock, traditional and non-traditional agricultural export crops) and off-farm employment is simultaneously pursued. Such a growth pattern must be accompanied with market development and proper management o f food aid which does not distort the markets. 146 Agricultural growth options andpoverty reduction 6.54 Given this characterization o f the rural economy and demand behavior, we now proceed to explore the effect o f different agricultural growth pattems on poverty based on a multi-market model (see Box 6.4 for a description o f the assumptions underpinning the model). Box 6.4: Assumptionsunderpinningthe multi-market agricultural growth-poverty model Geographic diversity inproduction and consumption patterns across different agricultural commodities i s taken into account and separate demand and supply functions have been estimated per zone. Rural and urban differences in demand pattems are accounted for. Growth in non-agriculture i s assumed exogenous. Total expenditures spent on all the commodities equal the rural or urban income at the zonal level. Zonal rural and urban total income i s endogenously determined and is equal to the sum o f production revenues from zonal level agricultural and non- agricultural activities. An integrated national market i s assumed with differential price levels across zones, with producer prices for commodities in food surplus zones lower than the price in Addis and prices for the same commodity produced in food deficit zones higher than the Addis price. Domestic markets are linked to the international markets. Though given high transaction costs, prices for most (food) commodities are endogenously determined by the domestic market. Changes in poverty have been calculated using distribution-neutral shifts in response to exogenous productivity shocks across the different agricultural sub sectors o f the zonal expenditure distributions. The basic assumptions underpinning the model are described inTable B6.4.1, For a full description o f the model, we refer to Diao et al., 2004. Table B6.4.1: Base line characteristicsand base run growth assumptions of the agriculturalmulti-market model. Base line characteristics Base run assumptions Share in Share in Share in (historical growth economy agriculture consumption among rates) ("A) (%) poor (YO) Agricultural sub-sectors rural urban area yield Staple crops" 33.6 64.7 50-60 40-50 1.3 0.8 Livestock 13.5 26.0 4 - 6 3 - 6 4.8 Non-traditional export crops 2, 2.3 4.4 4.6 Coffee 2.5 4.8 1.8 A g sector growth (%) 2.8 Non-ag sector (%) 3.7 GDP growth (%) 3.2 Poverty reduction(%points)(base=44.5) 46.2 Cereals, pulses, oil crops, and root and tubers. ') Exportable vegetables, fruits, other horticultural products, chat, cotton, sugar, and sesame seed. Source: Adaptedfrom Diao, et al., 2004. 6.55 Table 6.8 describes the effects on economic growth and poverty reduction o f different agricultural growth patterns. Four scenarios are considered. The first three scenarios explore the growth and poverty reducing effects o f 1.5 percent additional growth in staple crop productivity, 3.5 percent additional growth in the livestock sector, 8.7 percent additional 147 growth in non-traditional export able^,^^^ and 10.1 percent additional growth in the coffee sector respectively between 2003 and 2015. The fourth scenario examines the combined effect o f a simultaneous implementation o f the first three scenarios. The exogenous additional growth in productivity across the different sub-sectors i s chosen to yield comparable overall agricultural sectoral growth rates o f about 3.8 percent. Note that this i s comparable with the agricultural growth rates considered inthe key macro scenario presented inSection 6.1,where we considered agriculture growing at 4.1 percent. When combined with the assumed historical growth rates in each of the sub sectors (see Box 6.4), the scenarios considered here yield a productivity growth rate o f staple crop production o f 2.3 percent (+ 1.3 percent historical area expansion), 8.3 percent growth (4.8+3.5) in the livestock sector, 13.3 percent growth (4.6 +8.7) inthe non-traditional export sector, and 11.9 percent growth in the coffee sector (1.8 +lO.l). Growth in the non-agricultural sector is!exogenously assumed at 3.7 percent, which i s somewhat below the historical average observed over the past decade and a half. The 1999 poverty head count rate o f 44.5 percent reported by the government is taken as the base line. Table 6.8: Growth and poverty reducing effects o f different agricultural growth patterns Combined Staple crop Livestock Coffee productivity growth only traditional. ') only export on]$) only (staple, livestock and non-trad exDort) Additional productivity growth (YO)(only in sub ~ector)~) 1.5 3.5 8.7 10.1 Additional growth and poverty reduction A g Sector growth (%) 3.8 3.9 3.6 3.5 5.5 Non-ag sector growth (%) 3.7 3.7 3.7 3.7 3.7 GDP growth (%) 4 4.1 3.7 3.6 5.3 Poverty headcount after the intervention (base=44.5) 37.7 40 41 42.8 28.4 r, Cereals, pulses, oil crops, and root and tubers. 2, Exportable vegetables, h i t s , other horticultural products, chat, cotton, sugar, and sesame seed. 3, Productivity growth in addition to historical (Le. base run) annual growth, which is estimated at 1.3 percent area expansion for staple crops and 0.8 percent yield increase (or 2.1 percent total growth); 4.8 percent for livestock; 4.6 percent for non-traditional export crops and 1.8 percent for coffee. 6.56 Growth in staple crop production is critical for poverty reduction. We begin by considering the growth and poverty reducing effects o f separate interventions in each o f the agricultural sub sectors. From Table 6.8, it i s clear that the largest poverty reducing effects are brought about by productivity growth in staple crop production. This follows from the fact that productivity increases in staple crop production directly benefit many small farmers, thereby generating a substantial drop in rural poverty. Moreover, given that the poor inrural and urban areas spend respectively 70 and 50 percent o f their total income on staple crop food, increased food supply which lowers food prices benefits the poor indirectly. 6.57 Two caveats must be noted here. First, the implicit assumption o f perfect market integration underpinningthe model i s tenuous, and an increase instaple crop productivity will need to be complemented with better incentive structures for the traders, improved 237These include exportable vegetables, fruits, other horticultural products, chat, cotton, sugar, and sesame seed. 148 institutional arrangements, and the development of transport and storage infrastructure to bring about well-functioning cereal markets.238 Second, as discussed above, while the threat o f collapsing cereal prices in response to productivity enhancements should not be exaggerated, and a gradual decline in food prices i s indeed desirable, these cannot be automatically assumed. Both better market development and an increase in the demand for food through development in the non-food sector will be necessary to better facilitate a gradual and sustainable decline in food prices. In the model, productivity growth in staple crop production i s imposed exogenously and gradually, and does not allow for a "corrective" reaction of surplus producers through which they could reduce the use o f inputs and thus output production, leading to large price fluctuations in the face o f price inelastic cereal demand. 6.58 Growth in the livestock sector alone has the largest effect on overall economic growth, but a smaller effect on poverty reduction. The livestock sector has historically been growing at a faster pace (4.8 percent between 1995 and 2002) than staple crop production and because it still has more growth potential than staple crop production, an additional growth rate o f 3.5 percent i s assumed. While in this scenario more overall growth i s generated, the poverty reducing effects are smaller because livestock accounts for a relatively small share o f poor farmers' income. Further, poor consumers benefit much less from price declines in livestock products (meat and dairy) induced by increased livestock production, as these constitute only a small share o f their total expenditures (four and three percent respectively). Nonetheless, productivity gains in the livestock sector could be a route out o f poverty for subsets o f households which rely heavily on the livestock as a major source o f income. 6.59 Finally, growth in the non-traditional export sector may have less poverty reducing effects. However, growth in this sector may currently not require much additional public investment and it could help relieve pressure on land by generating employment and demand for food while at the same time preventing food prices from collapsing. Growth in some nontraditional exportables (e.g. export o f roses) has been booming in recent years, indicating that there i s still a lot o f growth potential in this sector. An additional annual growth rate of 8.7 percent is comparable to an additional growth rate of 3.5 percent in livestock and 1.5 percent in staple crop production. International experience suggests that segments o f traditional exports have the potential to multiply in value, given the size o f export markets, once established. However, since such activities often only involve a small group of farmers, mostly concentrated in certain regions and around big cities, especially during the early stages due to technology and credit constraints, it will be important to examine the value chains involved to ensure broader involvement o f smallholders and their capture o f a sizeable share o f that value. Consumers tend to benefit little given that the products are intended for the export markets. 6.60 In addition, growth in labor intensive non-traditional exportables may generate important employment opportunities, thereby relieving land pressure on households in the more traditional cereal producing zones. This linkage effect has not been accounted for in this model, even though it is heavily emphasized in the Ethiopian Computable General 238Gabremadhin,2004. 149 Equilibrium For this linkage effect to be important, labor has to become much more mobile. It will also require timely expansion o f export markets to sustain the necessary annual increase in exports by 29 percent. This will in turn require a substantial reduction in transport and transaction costs. Nonetheless, as illustrated by the recent boom in floriculture, substantial progress can be made in these areas without substantial public investment by the government, especially around Addis Ababa. The role o f the government as facilitator, coordinator and thus generator o f a conducive investment climate has proven to be much more crucial than public investment to attract the necessary private investments at this stage. 6.61 Balanced growth strategies complemented with market development have the largest poverty reducing potential. The separate simulations underscore the critical importance o f continued and concerted efforts to increase staple crop productivity for maximum poverty reduction. Yet they are only part o f the solution. Growth in other agricultural sub-sectors (e.g. livestock and non-traditional agricultural exports) as well as non- agricultural sectors, must be simultaneously pursued. Simultaneous additional productivity growth in all three sectors would reduce poverty to by 16 percentage points to 28.4 percent. Balanced growth strategies and market development will facilitate the necessary gradual decline in food prices as well as the transition out o f agriculture into non-agricultural activities. Some o f the linkages between growth in the different sectors go as follows. Growth in staple crop production and non-traditional exports increase domestic demand for livestock products and help support livestock prices. This helps farmers raise income from livestock production. Growth in the livestock sector generates demand for animal feed and increased income from livestock and non-traditional exports raises demand for cereals, which both help support cereal prices, preventinga collapse incereal prices. 6.62 Lower agricultural growth rates than the ones assumed in the multi-market model may actually be necessary to reach the poverty MDG if one also accounts for the consumptionlinkages.Note that the multi-market model under discussion here assumed that the non-agricultural sectors grow exogenously at 3.7 percent. Indoing so, it has neglected the important consumption linkage effects discussed above. As food prices decline, households' real income increases, which will generate demand for locally produced goods and services and thus generate off-farm employment, which inturn will stimulate the demand for food and thus help prevent food prices from collapsing. As a result, lower agricultural growth rates may be necessary to reach the poverty MDG, as assumed inthe macro simulations in Section 6.1, Table 6.1 where it was found that an agricultural growth rate o f 4.1 percent would almost suffice to reduce poverty by half from its 1990 level, given that the non-agricultural sector continued to grow at its historically observed average pace o f 6.5 percent per year since 1992. Yet while non-agricultural growth has so far been largelydrivenby government expansion, it will be critical inthe future that off-farm supply is sufficiently elastic to the increased demand for local goods and services to hlly capture the consumption linkages. 239World Bank Country, 2004c. An undesirable feature of the current scenarios based on the Ethiopian Computable General Equilibrium Model is that they lead to less food production, higher food prices, and less food imports. This seems at odds with the needfill the foodgap inEthiopia. 150 Toward more spatially differentiated growth strategies, alsofor agriculture 6.63 From Section 5.2 we recall that a decade of three to four percent sustained annual growth in staple crop production is technically within reach with intensifieduse o f fertilizer and improved seed packages coupled with soil conservation (and water management). Irrigation i s especially profitable when applied to non-traditional exportables (vegetables) and coffee, but less so when applied to cereal production.240 Moreover, giventhe limited area expansion under irrigation which can be expected compared to the total cereal area cultivated, it i s unlikely to make an important contribution to overall cereal production. Nonetheless, while it i s crucial to know that such productivity increases are technically possible, it will be equally critical for the success o f an agricultural growth-based poverty reduction strategy to better understand the reasons behindthe current limited adoption rates of these technologies (modem inputs, soil conservation techniques, and water management). Moreover, given the geographical diversity which characterizes Ethiopia, it will be important to further reflect on the optimality o f different (agricultural) growth strategies across different locations. 6.64 Staple crop production growth appears to be more poverty reducing in food surplus areas, while livestock production growth appears to be more poverty reducing in food deficit areas. Further disaggregation of the poverty reducing effects o f the different scenarios under the multi-market model by food deficit and food surplus areas241indicates that while productivity growth in staple crop production has significant poverty reducing effects across Ethiopia, its effects are more pronounced inthe food surplus areas. The poverty head count in rural areas in the food surplus zones falls from 30 to 23 percent by 2015, while it drops from 62 to 54 percent in the food deficit rural areas. Under the livestock growth scenario on the other hand, poverty falls only from 30 to 27.6 percent inthe rural food surplus areas, while it drops from 62 to 53 percent inthe food deficit areas. 6.65 In the northern dry lands of Ethiopia,while there is still scope for intensification of food productionwhich will continueto be important for food security purposes, there is more potential for promoting livestock production and tree planting (for construction poles and fuel wood), indicatin the desirability o f the broader extension packages currently pursued by the government."W Moreover, to successfully intensify food production, promotion o f the use of fertilizer and improved seed packages (e.g. for sorghum) will need to be complemented with the adoption o f soil conservation structures (stone terraces) and sustainable land management practices such as reduced (or zero) tillage, reduced burning and application of manure, and better water management (water harvesting, micro dams and small scale irrigation). Better land management practices help reduce soil erosion and increase the soil moisture content. Zero tillage could also greatly benefit female headed households, since women are culturally not allowed to plough and thus often constrained to share cropping, or other households without oxen. Small scale irrigation would increase the use o f fertilizer and ~~ 240 See World Bank, 2004c and World Bank, 2003b for more detailed discussions o n the costs and benefits of different water management techniques across different crops. 24' A zone i s considered food deficitlsurplus if the zonal level per rural household cereal equivalent output i s 20 percent belowiabove the national average. Otherwise the zone i s considered food balanced. 242For more details, see Ehui and Pender, 2004. 151 offer the potential for higher productivity. It would also facilitate the shift toward higher value crops such as vegetables. 6.66 Improvement o f livestock production will require institutional changes in the management o f grazing and crop lands. Investments in chicken and beehives appear to be profitable, with returns exceeding 30 percent. Allocation o f unused degraded land for private tree planting or other economic uses to communities and households has proven to be quite successful in fostering tree planting, wood lot management, and wood production. Food and cash for work programs will remain an important complement to farm income, yielding on average about 40 percent o f off-farm income and about 10 percent o f total income in Tigray in 1998-99. Finally, continued investmentin education is key, both to increase incomes and human development outcomes locally as well as to foster out-migration. 6.67 In the high potential cereal producing areas of the central and northwestern highlands, a continued focus on intensive cereal production through increased use of inputpackagesis warranted given the existingscope for substantialyield increases. This strategy could be usefully complemented with the development o f dairy production in areas closer to urban markets. Such a strategy will also need to be complemented with further market development to enhance the tradability o f the produced surplus. 6.68 In the humid high potential perennial zones in the southern and western highlands,continuedfocus on intensificationof food productionthrough improvedseeds and fertilizer use, along with concerted efforts to increase productivity and marketing efficiency for coffee hold promise. Intensification o f food production, which is little marketed given high transaction costs, would alleviate landpressure inthese areas, where it i s the highest in all o f Ethiopia, as well as the tension between food and cash crop (coffee or chat). There appears ample scope for increasing productivity in coffee production especially through the adoption o f higher yielding improved cultivars resistant to coffee berry disease, which i s faced by more than half o f the coffee planted in Ethiopia. Better management o f coffee quality and reduction o f the marketing margins, estimated at 50-60 percent o f export prices in 2000 and 2001 could further substantially enhance agricultural incomes. 6.69 In Central Ethiopia around Addis Ababa, further intensification of cereal crop production, developmentof the dairy industry in peri-urbanareas, tree plantingand the promotion of non traditional agricultural exports hold much promise. To further promote the intensification of cereal production through adoption o f modern input packages, the provision o f these inputs along with credit for their purchase may have to be complemented with rainfall based insurance to protect farmers from shocks. This also holds for the other regions, and experiments are currently underway to provide interlinked credit- insurance packages to farmers in Tigray. While the agro-ecological conditions inthis part o f the country are similar to those o f central Kenya, dairy production in Ethiopia is very underdeveloped compared to Kenya. Development o f the dairy sector will require addressing some o f the following constraints: (1) hightransaction costs incollecting and marketing dairy products; (2) limited adoption o f higher yielding cross-bred cows, despite evidence that they are profitable, due to lack o f credit to buy them as well as concerns about larger susceptibility o f cross-breds to diseases; (3) limited quality and irregular. supply o f fodder. Eucalyptus production in areas unsuitable for crop production or on uncultivated slopes also appears to be 152 quite profitable, especially in areas with good market access.243Finally, given the proximity to export markets, excellent airport facilities, good road infrastructure around Addis, and favorable ago-ecological conditions, non-traditional exports such as horti- and floriculture appear intemationally competitive and have excellent growth potential if the business environment i s accommodating. Promotion o f these activities will generate employment, increase the demand for food, relieve land pressure and help facilitate the transition out o f agriculture. 6.3 Toward an Optimal PublicInvestmentMix-a Micro Perspective 6.70 While an agricultural led growth strategy does require continued agriculture- specific investments in such areas as research and extension, many of the necessary investments to foster agricultural growth (both in the food and the non-food sector) actually lie outside agriculture, such as investments in infrastructure, education, health, access to information, etc. As both agricultural as well as non-agricultural activity typically stand to benefit from these investments, the debate about agriculture versus non-agriculture i s often ill conceived from this perspective. To gauge the relative marginal benefits from different non-agricultural specific public interventions, we use the estimated results from the regression analysis presented in Tables 4.9 and 4.11, Chapter 4. While we do not have information on the cost o f the various interventions and thus cannot evaluate their cost effectiveness, these simulations nonetheless help illuminate which interventions are likely to be critical for reducing poverty. Furthermore, it i s implicitly assumed that the estimated marginal retums remain stable over time and that they do not change as the variables themselves change. The further inthe future one projects and the larger the simulated change in a particular variable, the more tenuous this assumption becomes.244 Nonetheless, these simulations can provide useful guidance to inform changes at the margin inthe current public investment mix, along with insight into the size and optimal mix o f investments needed to reach the poverty MDG. Simulation scenarios 1-10 are based on the pooled woreda Fixed Effect Model (Column 1 o f Table 4.9), while we use the estimated results in Table 4.11 with an explicit geographic characterization o f the woredas for scenarios 11-14.245 6.71 There are substantial poverty reducing effects associated with increasing educational achievements. We begin by simulating the effect o f attaining the literacy- centered MDG: having all adults educated to the primary level. However, to mitigate the Lucas we only simulate the effect o f endowing all adults with at least a fourth grade level o f education (the first level o f primary education). The results are impressive and 243Holden et al., 2003; Okumuet al., 2002. 244T o see this, note that it i s likely that the returns to education will change if the political and economic climate in a country change. Similarly, when many people start to obtain primary education, the returns to primary education inthe labor market will probably decline compared to when virtually nobody hadprimaryeducation. 245For the education simulations, we use a quadratic specification, since we simulate the effect o f an increase in primaryeducation and we want to control for the differential returns between primary and secondary education, even though they are limited. 246The Lucas Critique argues that a large increase in educational outcomes may affect the returns to education, which would bias our simulations (Lucas, 1976). The increase in mean education levels may reduce individual retums to education following an increase inthe supply. Conversely, a more educated population might unlock new growth opportunities, stimulate a culture o f entrepreneurs, etc., and by these positive externalities increase the retums to education. Furthermore, changes in education might also indirectly affect the retums to other endowments in the model. 153 lend substantial support to massive investment inprimary education in Ethiopia as a primary policy objective. As simulation 3 shows, providing all adults with a fourth grade education would raise consumption levels by eight percent and reduce poverty incidence by eighteen percent. The poverty gap would be reduced by almost a quarter. Moreover, the poverty alleviation response to female education is almost twice as high as the poverty response to male education. What it would take to drastically enhance male and especially female enrollment and completion rates i s explored in Chapter 9. Given its immediate effects, the role for adult literacy campaigns in imparting literacy on adults should also be explored as a complementary and timely intervention. Table 6.9: Simulatedeffects on consumptionand poverty reductionof different policy interventions ExpenditureETB HeadCount Poverty'' Poverty Gap'' Pre Inter- Post % Pre Inter- Post % Pre Post Inter- % vention Inter- change vention Inter- change Inter- vention change vention vention vention Woreda Fixed Effects: Pooled 1995/99 Education') 1) all male adults with at least 4" grade education 1435 1474 2.70 38 36 -6.24 9.94 9.07 -8.79 2) all female adultswth at least 4" grade education 1435 1514 5.47 38 34 -11.62 9.94 8.37 -15.79 3) joint intervention(I)& (2) 1435 1555 8.34 38 31 -17.64 9.94 7.59 -23.60 Infrastructure 4) providing poorest quintile with access to electricity 1435 1443 0.52 38 38 0.00 9.94 9.23 -7.17 5) providing poorestquintile with access to proper sanitation 1435 1450 1.04 38 38 0.00 9.94 8.52 -14.29 6) bringing everybody within lOkmof budtaxi service 1435 1461 1.76 38 37 -3.84 9.94 9.39 -5.52 7) bringing everybody within 2km of clean water source 1435 1440 0.32 38 38 -0.57 9.94 9.86 -0,84 8) joint intervention(4), (5), (6) & (7) 1435 1489 3.75 38 36 -5.15 9.94 7.05 -29.10 9) full intervention(3) & (8) 1435 1614 12.45 38 27 -27.96 9.94 4.51 -52.73 Information 10) providing poorest quintile with a radio 1435 1506 4.96 38 34 -10.82 9.94 5.82 -41.45 Unbundling woredas: Model 13' Nurient Loss 11) reducing DAP nutrient loss in all woredas to presentaverage 1398 1418 1.41 39 38 -3.28 10.14 9.74 -3.94 12) full intervention(9) & (11)4) 1398 1596 14.16 39 26 -34.33 10.14 4.52 -55.43 Unbundling woredas: Model Z5) Vulnerability 13) limiting crop damage to amaximum of 20% 1348 1356 0.61 40 40 -1.23 10.52 10.32 -1.91 14) limiting crop damage to amaximum of 10% 1348 1365 1.29 40 39 -2.54 10.52 10.11 -3.99 Woreda Fixed Effects: Extended 99" Cultivationpractices 15) Limiting minimumuse of commercial fertilizer to 32% of cultivatedland 1351 1402 3.61 39.18 36.04 -8.73 10.03 8.88 -12.95 16) Limiting minimum use of natural fertilizer to 32% of cultivatedland 1351 1434 5.79 39.18 34.56 -13.37 10.03 8.51 -17.89 154 ExpenditureETB HeadCount Poverty" PovertyGap" Pre Inter- Post % Pre Inter- Post % Pre Post Inter- % vention Inter- change vention Inter- change Inter- vention change vention vention vention Expectedpoverty reducing effects of public investment in education and roads between 1999and 2004 17)adding I 9ofa grade to eachadult , male" 1435 1437 0.14 38.08 37.98 -0.25 9.94 9.90 -0.42 18)adding .22ofa grade to eachadult female') 1435 1436 0.06 38.08 38.05 -0.07 9.94 9.92 -0.16 19)joint intervention (1 8) & (1 9) ') 1435 1438 0.21 38.08 37.95 -0.33 9.94 9.88 -0.57 20) Decreasingdistanceto dry weather roadby I.2 km 1351 1353 0.14 39.18 39.11 -0.20 10.03 9.99 -0.43 21) joint intervention (21) & (30) 1351 1356 0.35 39.18 38.96 -0.58 10.03 9.93 -I.03 22) Decreasingdistanceto dry weather readby 3.8 km 1351 1356 0.39 39.18 38.90 -0.72 10.03 9.92 -1.18 ') We use the lower poverty line as our poverty threshold. *) Instead o f using the reported regression results in Table 4.9, Chapter 4, the simulations on education use a quadratic specification for education to capture non-linearities in retums (Le. lower returns to primary than to secondary education). 3, Note that the full intervention inthis case uses the coefficients for the education and infrastructure variables from the regression on Unbundling woreda Effects: Model 1. As we did not have information on DAP nutrient loss and most o f the additional variables used in Model 1, for all woredas, we use approximately 17,000 observations for Model 1 as opposed to 28,000 for the woreda Fixed Effects Pooled Model. Note also that the intervention for education consists o f providing universal primary education for all adults and the full interventionsimulation used coefficients from the reduced observations model. 4, The quadratic specification was only significant for the pooled woreda Fixed Effect Model and as such we use the linear, grade 4 education intervention (4) & (5) for the education component o f this full intervention. 5, As we only had crop damage information for rural areas, Model 2 is limited only to our rural sample, which consists o fjust over 11,500 observations. 6, This specification only uses 1999 data as no data on cultivationpractices was collected in 1995. 6.72 Improvementsin infrastructureaffect consumptionnot only directly by lowering transactions costs and increasing access to opportunities and information, but also indirectly through their effect on the returns to endowments. We take a direct targeting approach in our simulations, even though that may not be the most cost-effective pro-poor approach overall given the costs of public service delivery to certain poor areas. We either target our interventions to the poorest households in our sample when the intervention is dichotomous (either you have electricity or access to proper sanitation or you do not), or we improve accessibility for those who are most remote from the service first. Providing all households in the poorest quintile with electricity such that 35 percent of the population, as opposed to the current 10 percent247, have access to electricity only increases average consumptionby O S percent. Providing the households in the poorest quintile with proper sanitation, an intervention that increases the percent of householdswith proper sanitation from 16 to 39 percent,248results ina one percent increase in consumption. When taken inisolation, 247 Note that the intervention does not result in a full 20 percent addition o f households with electricity as a s m a l l fraction o f the poorest quintile already had access to electricity. Recall too that all household level means we "'resent are population weighted and thus affect the result. While maximum 20 percent o f the households would get access to sanitation, more than 20 percent o f the people may gain access to sanitation since the intervention is targeted to the poorest households which also tend to be larger. 155 neither of these two interventions singularly affects the poverty incidence. This is clearly a function o f targeting the poorest households. Despite the fact that their mean consumption levels will increase, their poverty i s too deeply rooted for it to be erasedby these interventions alone. Indeed, when we look at the effect o f these interventions on the poverty gap, we notice that these interventions reduce the poverty gap by seven and 14 percent respectively. 6.73 Next, we simulate the effect o f investments in transport infrastructure and water provision. We bring everybody within at least 10 kilometers o f budtaxi service. This drops the mean from 16.69 to 6.47 km and benefits 31 percent o f the sample. Inthe case o f water, we place a two kilometer limit on distance to water, dropping the mean from 1.57 to 0.51 km and affecting 12 percent o f the sample. Improved access to transport has a more substantial effect, reducing poverty incidence by 3.84 percent, as opposed to 0.57 percent for improved access to clean drinking water. 6.74 Direct poverty targeted investment in public infrastructure could substantially reducethe poverty gap, though it will need to be complementedwith other investments to substantially reduce poverty incidence. While each o f the infrastructure interventions has somewhat modest, though non-negligible, effects on consumption, their joint effect i s much larger than the sum o f the parts. Together, the simulated infrastructure investments would reduce poverty incidence by 5.15 percent. The simulations would further suggest that when the infrastructure interventions are combined with bringing all adults to 4th grade completion, Ethiopia would be able to reduce poverty incidence by about 28 percent , i.e. more than halfthe distance o f reaching the poverty MDG. The poverty gap would be reduced by one half. 6.75 Estimated benefits from reduced dung collection are significant. Our regression results already pointed to sizeable negative effects of dung collection on expenditure. According to our simulations (1l), bringing all woredas with DAP nutrient loss due to dung (and crop residue) collection greater than the mean back down to the mean would reduce poverty by 3.28 percent. Clearly, actions to reduce dung collection and increase soil fertility are likely to have substantial effects on poverty. For comparison, bringing DAP loss due to dung collection everywhere back to the current nationwide average is estimated to have an effect similar to that o f bringing everybody within 10 km o f budtaxi service, or giving all male adults two extra years o f primary schooling. The combined effect o f 4th grade completion for all adults, the simulated infrastructure improvements, and reduction in dung collection would reduce poverty by 34 percent to 26 percentage points. While these results are somewhat less optimistic than those based on the pooled fixed effect simulations, they nonetheless indicate that substantial reductions in poverty could be achieved by improving educational achievements, increasing access to infrastructure and promoting alternative fuel use. 6.76 Tremendous gains are possible from increasing access to information by providing a radio to the poorest quintile of the population, with effects similar to bringing all female adults to qfhgrade completion (see simulation 10 in Table 6.9). As indicated in Chapter 4 (see also Appendix 3, Tables A.4.1 and A.4.2), households and communities with a radio appear substantially better off than those without a radio. Information i s a powerful transformer indeed, especially when people are illiterate and 156 completely isolated from the rest o f the world, virtually excluding them from any source o f information and knowledge, as inEthiopia. Insuch circumstances, radio is a powerful tool to promote better farming techniques, to enhance people's knowledge o f health and nutrition and promote better hygiene and nutritional practices, to facilitate market integration by communicating food prices, and to foster development more generally throughout Ethiopia. Radio i s also a major dialogue initiator, explaining the important observed externality effects o f radio ownership within communities. It i s estimated that providing the poorest quintile o f the population with a radio would increase average consumption by five percent, it would reduce poverty incidence by 11 percent, and it would reduce the poverty gap by about 40 percent. These effects are similar to those o f bringing all female adults to 4`h grade completion, and exceed the combined effects o f the infrastructure package. 6.77 Marginal beneficial effects of radio ownership are real, high compared to the marginal costs, and not simply a reflection of people's wealth position. While one would be justified in being suspicious about the magnitude of the estimated effects and one could argue that these estimated effects merely capture a person's wealth position (those who are wealthier are more likely to have a radio), it must be underscored that the estimated effects o f radio ownership on household welfare are over and above people's wealth position as captured by their possession o f livestock, farm equipment, consumer durables (bicycle and TV) as well as the quality o ftheir housing (quality o fmaterial of walls and roofs). Moreover, this effect is inaddition to the effect o f other public infrastructure(electricity, access to roads, health centers, markets, water and sanitation) as well as the overall wealth position o f the woreda (woreda dummy variables have been used) (see Appendix 3, Tables A.4.1 and A.4.2 for a detailed specification). The effect furthermore holds in both rural and urban areas, though the externality effect i s especially strong in rural areas. Even if the effects would be slightly overestimated, enhancing radio ownership and access to information more generally emerges as a very cost effective intervention, with the marginal benefits largely exceeding the marginal gains at the current stage o f development inEthiopia. 6.78 While the direct estimated poverty reducing effect may appear small, providing better risk management tools will be critical given the long-lasting damage of shocks. Providing farmers with irrigation to protect themselves against droughts, increased use o f pesticides, or ex post insurance schemes such as rainfall based insurance so that their crop damage would not exceed 20 percent o f their total cultivated area would reduce poverty 1.2 percent, This appears rather small, Yet this i s likely to be an underestimate o f the negative effects o f shocks on human welfare, as shocks do not only affect current consumption but are also likely to suppress future consumption. The micro-evidence from Ethiopia presented by Dercon (2004) indicates a long lasting growth penalty o f two to three percentage points for households heavily affected by the 1984-85 drought shock. Moreover, consumption smoothing by households may cause long lasting damage to the hture earnings stream o f their children. As indicated in Chapter 10, households in Ethiopia often withdraw their children, and especially girls, from school in response to crop damage shocks. Moreover, crop damage also appears to negatively affect early child growth, causing long-lasting permanent damage to their cognitive ability and future earningpotential.249 249Yamano, Alderman and Christiaensen, 2005. 157 6.79 Increasing the share of cultivated land covered by commercial fertilizer to 32 percent-the current average in food secure areas-promises to increase overall consumptionlevels by 3.6 percentand reduce povertyby about eight percent. While our simulations may admittedly also capture the effects o f other technologies coming along with fertilizer use (such as the effects o f improved seeds, pesticides, better cultivation practices, et^.),^^' agriculture in Ethiopia has clearly not yet reached its current production frontier, and there remains substantial scope for wider diffusion and adoption o f basic agricultural inputs. This finding is consistent with those presented in Chapter 5, especially since the estimated coefficients are based on current practices while the findings from Chapter 5 indicate that many gains will come from combined fertilizer and improved seed packages, which are currently still only applied on about four percent o f the cultivated cereal areas. Enhanced formal education, better access to roads and transport services, improved availability o f insurance schemes, and increased access to information about farming techniques, expected timing o f the rainfall (weather forecasting) and market prices through increased radio ownership and broadcasting will all be instrumental in promoting the adoption o f modem inputs and the use ofmore efficient application rates and farming techniques. 6.80 Additional improvements in policies and the institutional environment will be necessary to maintain and further enhance the estimated marginal returns to these public investments.251Inparticular, further improvements inthe investment climate will be needed to facilitate a private sector response and reinforce the expected benefits. The government has recently introduced new policies in favor o f the private sector and has been very active inimproving its dialogue with the sector. Progress has been achieved inthe areas o f land availability, business registration, tax management and customs. Sector specific policies granting incentives and removing key obstacles to private investment have already paid off, for example in the significant growth o f the floriculture sector. Nonetheless, significant obstacles remain to the development o f the private sector. The system o f laws and regulations still deters private sector investment. Weak domestic competition and contestability inmany sectors, the nature o f state interventions, and the numerous state-owned enterprises taken together result in an uneven playing field. Inefficiency o f the banking sector, which i s almost wholly state-owned, leads to high cost o f capital and difficult access to credit. A particularly critical issue to facilitate private sector response (and urban growth) relates to the availability and cost o f (urban) land. While the reform o f urban land i s now in full swing, progress has been slow and substantial residual uncertainties for private investors persist, hampering private sector investment. 250Note however, that the use o f these practices is so limited that our coefficients are unlikely to pick up their effects. 25'See World Bank, 2004c for more details. 158 * CHAPTER 7. ENHANCINGPEOPLE'S INCOME-CONCLUDING REMARKS Households startfrom an extremely low endowment base 7.1 Ethiopia's private endowment base is extremely low and has largely remainedso over the past decade. While the educational status o f Ethiopia's population has beensteadily improving over the past decade, educational attainment remains limited. Male adults completed on average only 1.8 grades, and female adults only 0.88 grades, with the gender gap somewhat less pronounced in Tigray. Disease and malnutrition (see further Chapter 8) further erode labor productivity. Land pressure has increased tremendously over the past decades, with average landholdings declining from 0.5 ha per person in the 1960s to 0.11 ha per person in 1999. While about 40 percent o f the population have at least one ox, only 30 percent have two, the number necessary for an ox-span to plough the fields. 7.2 A land poor class living on "hunger" plots has emerged. Since the land reform under the Derg regime, Ethiopian agriculture is essentially characterized by smallholder farming. Nonetheless, land inequality i s high, as a result o f continuous fragmentation o f landholdings and the emergence o f a rapidly growing group o f people living on "hunger plots". For example, given current technology, about one-fifth o f all rural households (excluding SNNPR) do not manage to produce half o f their annual cereal caloric needs from their plots, despite beingmainly dependent on agriculture. 7.3 Soil nutrient depletion continues at an alarming pace. The most important source o f cooking fuel i s firewood, used by almost 75 percent o f households. About one in six households uses mainly dung cakes as source o f cooking fuel, resulting in a continuous depletion o f the soil at alarming rates. Preliminary estimates suggest that the annual phosphorus and nitrogen loss nationwide due to dung removal i s about equivalent to the total amount o f commercial fertilizer annually applied. While the extent to which the nutrient content o f dung translates into actual alimentation o f the soil with phosphorus and nitrogen depends on the handling of the dung and the cultivation methods used, these results are sufficiently significant to warrant a closer investigation o f the effect o f dung collection on agricultural productivity and poverty. 7.4 Remoteness epitomizes daily life in rural Ethiopia. Only 14 percent o f the rural population has a radio, exemplifying the sheer disconnect o f Ethiopia's population from the rest o f the world, not only in terms o f market access, but also in terms o f access to information. Rural households are on average 10 kilometers away from a dry weather road and 18 kilometers from any public transport, renderingpeople immobile. Average distance to drinking water, however, improved substantially from 2.8 km in 1995 to 0.8 km in 1999. On the other hand, access to services inurban areas deteriorated, likely due to an influx of rural immigrants. 7.5 Risk and drought shocks have severe and longlasting effects on poverty. Ceteris paribus, households in areas characterized by larger rainfall fluctuations were found to be poorer. The effects o f crop damage due to droughts, pests, insects, frosts, or other causes on 159 consumption were substantial. For example, it was estimated that 2.7 percent o f consumption per adult equivalent was lost in 1999 due to crop damage, a year characterized by average rainfall. This corresponds to about 1.5 years o f GDP growth per capita. Moreover, not only are households unable to protect their consumption from continuously recurring shocks, but the effect o f these shocks are often long lasting. Micro-econometric evidence shows that households that suffered substantially during the 1984-5 drought, which resulted in a nationwide famine, continued to experience two to three percent less annual per capita growth duringthe 1990sthan those who weren't hit as hard. Livelihoods are agriculture based, but labor productivity in agriculture is low 7.6 Ethiopia still finds itself at the very beginning of its structural transformation, with rural households continuing to rely heavily on low input, low output, subsistence- oriented, rained agriculture and agriculture related activities. Agriculture is responsible for 85 percent of employment, 45 percent o f national income and more than 90 percent o f exports. O f the total area under temporary crops in the 1990s, cereals, pulses, and oilseeds accounted for 88.7 percent, 8.7 percent and 2.7 percent respectively. Commercial fertilizer i s applied to approximately 40 percent o f total farmlandunder cereals and heavily concentrated on a few cereals (wheat, teff, and maize). Improved seeds are only applied on about five percent and pesticides on about seven percent o f the total cultivated cereal area. Less than one percent o f the total cultivated area in Ethiopia i s irrigated, despite massive fluctuations in rainfall. The limited (and only slowly expanding) use o f inputs and modern technology combined with the small and decreasing landholding sizes i s consistent with the consistently low yields in cereal production and the low marginal productivity o f labor observed in the data. 7.7 Evidence suggests significant numbers of net cereal buying poor households in rural Ethiopia. This is consistent with the evidence from other poor SSA countries. Poorer households are more likely to engage in multiple (often non-remunerative) activities as a coping strategy. Especially livestock products, but also business activities (collection o f water and fuel wood, artisanal activities, grain trading,) and off-farm wage work (especially food and cash for work) provide sources o f cash income to buy food. Inaddition to the urban poor population, these rural households stand to gain from a (gradual) decline incereal prices. This is in contrast to another smaller set o f poor rural households which are net cereal sellers, a poverty reduction dilemma. 7.8 Substantial potential remains for increasing productivity in staple crop production through agricultural intensification, especially in food secure areas. Based on a review o f the available evidence, rough estimates suggest that doubling cereal yields in the more food secure areas and increasing cereal yields by 50 percent in the food insecure areas lie well within the realm o f the possible. Much could be gained from broader adoption o f both fertilizer and improved seeds, and increasing market access (better access to roads and rural towns) in the more food secure areas, while broader adoption o f fertilizer-improved seed packages will have to be complemented with the promotion of soil conservation and better water and risk management techniques as well as improved market access inthe food insecure areas. To realize these benefits, it will be important to further our understanding o f the major factors constraining wider adoption and diffusion o f land saving technologies. The role o f 160 political and land tenure security inthe adoption o f irrigation and environmentally sustainable cultivation practices respectively, as well as the role o f having effective risk management strategies in adopting modern inputs deserve fbrther investigation. Anecdotal evidence suggests that the demand by poorer households for certain packages is limitedbecause o f the downside risks involved. Similarly, input delivery systems may have to be made more efficient and demand driven. 7.9 Nonetheless,the low marginal value of labor in terms of additional agricultural incomefrom cereal production, given current landholdingsize, and the higher marginal value of expanding landholdings, suggest complementary policy routes to expand households' income. As noted, one option is raising marginal productivity o f labor and land through agricultural intensification in cereal production. Second, the land frontier could be pushedfurther and new areas could be exploited, i.e. agricultural extensification. Third, labor productivity could be increased through diversification into non-cereal (tradable) agricultural production. Fourth, land pressure could be reduced and labor productivity enhanced through diversification and migration out o f agriculture into highly remunerative non-agricultural activities. The optimal combination will obviously differ across space depending on the region's comparative advantage in terms o f agro-ecological potential, and market access as determinedby population density and accessto infrastructure. Balanced agriculturalgrowth is criticalfor achieving overallgrowth withpoverty reduction 7.10 Macroeconomic projections suggest that reaching the MDG of halving poverty incidence by 2015 from its 1990 level will not be possible without buoyant agricultural growth. Simulations indicate that a decade of 4.1 percent agricultural growth would bringus close to reaching the poverty MDG. While such growth in agriculture will be necessary to sustain the required growth in the non-agricultural sectors and facilitate the structural transformation with labor shifting out o f agriculture into industry and services over time, it will not be sufficient. Across all these sectoral dimensions, enhancement o f individual agency, both in the economic and social space, and increased foreign aid will be equally necessary to achieve growth with poverty reduction. 7.11 This raises the question of how to bring about robust and sustained agricultural growth. Agricultural extensification has so far been the key factor driving growth in agriculture,though its potentialto further boost agriculturalgrowth is limited. Not only has pushing the land frontier not been sufficient for growth in agriculture to keep up with population growth, let alone to sustain growth o f 4.1 percent, the scope for further agricultural extensification i s rapidly decreasing, especially in light o f continued population growth o f 2.5 to three percent per year. There remains some potential in the lowlands when they become more accessible through malaria and tse-tse fly eradication, and given better public services. This i s consistent with the philosophy behind the ongoing resettlement program, though sustainable population movements would be better stimulated by fostering voluntary and less programmatic migration. 7.12 Agricultural intensification in cereal production through the promotion of increased modern input use will continue to have to play an important role in raising incomes and reducing poverty. Given the large share o f cereal consumption in (poor) 161 people's budget, and the existence o f a substantial group o f (poor) net cereal buyers (rural and urban), it will be important to sustain the focus on increasing cereal productivity to prevent cereal prices from rising rapidly which would hurt the poor (as well as many non-poor). Evidence suggests that there i s still a lot o f scope to do so through increased use o f modem input packages (fertilizer and improved seeds), especially inthe food secure areas. However, agricultural intensification in cereal production through increased modern input use alone will clearly not suffice and the constraints to further technology adoption (risk management, input delivery systems, market development) mustbe better understood. 7.13 Increasingcerealproductionin the face of priceinelasticcereal demand may lead to large cereal price declines. When increased production follows from reversible productivity increases (e.g. through the use o f fertilizer and improved seeds), this may subsequentlygenerate large (and undesirable) cereal price fluctuations. Or, whenproductivity increases follow from irreversible investments (e.g. infrastructure facilitating better water and soil management), it may put cereal producers on a price treadmill whereby producers see the gains from their investments being eroded by lower cereal prices, which will force them to either retire from agriculture or engage inanother round o f productivity increases. 7.14 However, these risks must be put in the right context and can also be managed. First, for the 2000-2002 experience, it is generally agreed that the observed collapse in (especially) maize prices was compounded by food aid mismanagement, with food aid being importedwhile it could have beenlocally procured to help support local cereal prices. Clearly there is a role for more effective and non-distortive management o f food aid. Second, in contrast to maize, the demand for other cereals such as teff and wheat i s much less price inelastic (and their supply i s also more price elastic)-the observed price collapse in 2000- 2002 was indeed the largest for maize. Third, it i s important to recognize that for many (rural) poorer households who are net cereal buyers as well as the urbanpopulation, declining cereal prices increase their real incomes. 7.15 To avoid large price fluctuations and/or a price treadmill in cereals, better food aid management and complementary actions especially in market development and agricultural diversification are needed. A simultaneous increase in the production and productivity o f non-staple tradables (livestock, traditional and non-traditional agricultural export crops), in addition to increased cereal production, can foster the production linkages, help generate off-farm employment, and generate demand for food which will prevent food prices from collapsing. A more balanced agricultural growth pattern will also facilitate migration out o f the food insecure areas and maximize the linkage effects and thus poverty reduction. To do so, focused interventions in staple and non-staple agriculture such as agricultural research and extension will need to be complemented with market development, i.e. proper incentives for farmers and traders, a facilitating institutional environment, and infrastructureto improve market connectivity. 7.16 Furthermore, agricultural growth strategies will need to be spatially diversified. While there is still some scope for intensification o f food production inthe northerndry lands o f Ethiopia, this should be complemented with promoting livestock production and tree planting, indicating the desirability o f the broader extension packages currently pursued by the government. Moreover, to successfully intensify food production, promotion o f the use o f 162 fertilizer and improved seed packages (e.g. for sorghum) will need to be complemented with the adoption o f soil conservation structures (e.g. stone terraces) and sustainable land management practices as well as better water and risk management. The current food for work programs when complemented with sufficient technical assistance could be usefully usedto help build these infrastructures. 7.17 Inthe highpotentialcereal producing areas ofthe central andnorthwestem highlands, a continued focus on intensive cereal production through increased use o f input packages i s warranted given the existing scope for substantial yield increases. This strategy could be usefully complemented with the development of dairy production in areas closer to urban markets. Continued focus on intensification o f food production through improved seeds and fertilizer use and concerted efforts to increase productivity o f coffee production and marketing efficiency hold promise to foster agricultural growth and reduce poverty in the humid high potential perennial zones in the southern and western highlands. In addition to further intensification of cereal crop production, development of the dairy industry in pen- urban areas, tree planting, and the promotion o f non-traditional agricultural exports, including floriculture and horticulture products also hold promise in Central Ethiopia around Addis Ababa, initially without much additional public investment, Returns topublic investment at the margin and the role of foreign aid 7.18 Many public investmentsbenefit both agricultural and non-agriculturalgrowth. While an agricultural led growth strategy does require continued investments in agricultural research and extension and other agricultural specific investments (e.g. soil conservation), many o f the necessary investments to foster agricultural growth (both in the food and non- food sector) lie actually outside o f agriculture, such as investments in infrastructure, education, health, and access to information. As both agricultural as well as non-agricultural activity usually stand to benefit from these investments, the debate about agriculture versus non-agriculture seems often illconceived from this perspective. 7.19 The pay-offs to education both for rural male and especially for rural female adults are still huge, warranting a continued focus on public investment in education, even though the benefits will only be felt over time. The micro-simulations suggest that bringing all male and female adults up to at least a 4th grade education could potentially reduce poverty incidence by 18 percent, with the gains from enhancing female primary education twice those o f enhancing male primary education. Education will be necessary to help households adopt new technologies and thus enhance their agricultural productivity. In addition, the more educated are usually more likely to migrate, and education i s strongly correlated with the adoption of off-farm work. Targeting education and training efforts to the landless as well as the more food insecure areas may facilitate their migration out o f agriculture and poverty, and provide the necessary trained workforce to respond to increased demand for locally produced goods and services following agricultural intensification. Yet while the next generation is beingeducated, the scope for adult literacy campaigns shouldalso be explored to generate a more immediate impact. The determinants o f primary school enrollments and the policy implications o f rapidly increasing school enrollments will be discussed inmore detail inChapter 10. 163 7.20 Tremendous gains are possible from increasingaccess to information. Providing a radio to the poorest quintile o f the population having similar poverty reducing effects as bringing all female adults to 4thgrade completion, warranting some reallocation of the current investment portfolio at the margin. Information i s a powerful transformer, especially when the majority o f the population is illiterate, living in isolation, and virtually excluded from any source o f information or knowledge, as in Ethiopia. In such circumstances, radio is a powerful tool to promote better farming techniques, to enable better timing o f farming through weather forecasting, to facilitate market integration by communicating food prices, to enhance people's knowledge o f health and nutrition and promote better hygiene and nutritional practices, and to foster development more generally throughout Ethiopia. Radio i s also a major dialogue initiator, explaining the important observed externality effects o f radio ownership within communities. It i s estimated that providing the poorest quintile o f the population with a radio would increase average consumption by five percent, reduce poverty incidence by 11 percent, and reduce the poverty gap by about 40 percent. These effects are similar to those o f bringingall female adults to 4'h grade completion. Moreover, the marginal beneficial effects o f radio ownership are real, empirically robust, and high compared to the marginal costs, making investment in increasing access to information a very cost effective and especially timely intervention to connect rural Ethiopia to the rest o f the world and reduce their poverty. 7.21 Promoting market connectivity through improved access to roads in addition to access to information will be critical to stimulate and distribute the benefits from increased agriculturaland non-agriculturalproduction. Access to markets as proxied by proximity to urban centers and roads contributes greatly to increased agricultural production, diversification o f agriculture into non-food production and thus overall poverty reduction. Roads bring direct short term employment, generate access to markets and services, facilitate migration and exchange o f information and ideas, and bring long term off-farm employment opportunities. Continued emphasis by the government on expanding especially the rural road network as envisaged inthe current road development plans i s warranted. 7.22 Policies to strengthen households' asset base should be supplemented with promoting a broad range of ex-ante and ex-post risk management strategies. The evidence on the large impact o f shocks, and especially their ratcheting, persistent effects highlight the high benefits o f containing any crisis and the need to find ways of supporting those affected by a crisis well beyond the initial crisis period. Public works programs and other safety nets clearly have a role to play, but targeting i s often difficult and timing issues continue to impact their ability to handle local level crisis. Moreover, both in food secure and food insecure areas risks and the absence o f efficient tools to cope with them ex post may prevent many poorer people from adopting more productive, but higher risk, production technologies such as fertilizer and more remunerative crop portfolios. While the empirical existence o f such a risk-induced poverty trap needs to be further documented, there are nonetheless good indications that many synergies could be obtained from the joint provision o f insurance (e.g. rainfall-based insurance), credit, and inputs. Effective insurance mechanisms (including productive safety nets and public work programs) would also reduce asset depletion in the face o f a shock, preventinghouseholds from falling into a poverty trap. Reducing household vulnerability to crisis will also have to implyreducing the dependence on 164 a small number o f agricultural based livelihood strategies, and effectively promoting the use o f water management techniques. 7.23 Institutional and resource obstacles to the generation of off-farm employment and the enhancement of returns to the current endowment base must be removed. While agricultural intensification should generate demand for locally produced goods and services, and thus stimulate local employment, institutional supply constraints in factor markets must be addressed. As testified by several citizens close to rural towns, there i s increasing competition among farmers and non-farmers for land, which i s often allocated in favor o f the former. Limitedavailability o f land for residential and business purposes appears to be an important constraint to the development o f off-farm employment. Other obstacles include access to credit and knowledge to start off-farm businesses. The needfor increased aid flows must be carefully balanced against absorption capacity constraints 7.24 The need for massive public investment is undisputed. As indicated, people's private and public endowment bases are extremely low. Moreover, many o f the necessary investmentssuch as those in children's education, children's nutritional status and health, and those in infrastructure will only pay off in the long or medium term respectively, while the immediate demands (e.g. food crises, health expenditures) are huge compared to the available resources. The necessary trade-offs betweeninvestmentinthe hture and the need to cater to immediate demands partly traps the country in poverty. Along with the recognition o f these needs and discrepancy intiming comes the imperative for increasing foreign assistance. 7.25 To be sustainable and effective, increased aid flows will need to be carefully balanced against simultaneously increased absorption capacity. While the simulations indicate that substantial aid flows will indeed be needed to reach the poverty MDG, the capacity o f the current system o f government to translate these increased aid flows into poverty reduction within acceptable fiduciary risks will need to be closely monitored, Important processes have already been put in motion over the past years to facilitate increasing aid flows under the public sector reform program and the Public Sector Capacity Building Program. Yet concerns remain regarding the limited number o f qualified government employees to properly channel these additional fiscal flows, the limited ability o f the government to contract out the supply o f services to the iocal private sector, and the limitedcapacity of the private sector to supply such services. How the tensions between the need for massive public investmentand foreign aid and the limited absorption capacity can be properly managed will need to be at the center o f our attention for increased aid flows to be effective inreducing poverty. 165 Part111:DeterminantsofNon-MonetaryWell-Being As discussed in Part I,assessing people's non-monetary dimensions o f well-being (their human capabilities and their empowerment status), as well as exploring how these dimensions could be improved form an integral part o f an evaluation o f people's overall well-being. In this part of the report, we will focus particularly on the determinants of human capabilities, which are both intrinsically and instrumentally valuable, and comment briefly on other interventions necessary to enable people to make effective choices intheir lives. Indeed, only individuals who are able to live free from hunger and preventable illnesses, and with adequate numeracy and literacy skills to participate in civic and market activities, are able to realize their maximum potential as human beings. Widespread subscription to the Millennium Development Goals i s evidence o f the intrinsic value ascribed to achievements in human development across countries and cultures. From Part I1 it i s also clear that improvements in human development outcomes have tremendous instrumental value and are infact essential for inclusive, equitable and sustainable economic growth, vulnerability mitigation and poverty reduction. This i s especially true o f the nutrition, health and education outcomes o f women and children. Early childhood malnutrition creates a nutrition-productivity poverty trap from which escape later in life may be impossible. Poor nutritional status among pregnant women and infants can have irreversible impacts on the cognitive development of children, reduce their educational attainment, and thereby negatively affect their labor productivity and earnings as adult workers, While in many countries, including Ethiopia, households maintain a bias against investing in female schooling, the individual and social returns to female education are extremely high, as discussed in Chapter 4. Moreover educated mothers are also less likely to have malnourished children and children who die prematurely from infectious disease, and are more likely to enroll their children in school. In identifying causal pathways and policy levers that can be used to improve human capabilities in Ethiopia, this part o f the report also seeks to fill some of the knowledge gaps identified in Ethiopia's SDPRP. We will try to shed light on the relative importance o f demand and supply side factors in reaching the MDGs, especially in the education sector. Building additional schools and supplying new teachers and financial resources to the educational system may not be enough to engender broad-based expansion o f the human capital base o f the country's workforce if the demand for schooling remains low. Low enrollment rates may be perpetuated by demand side factors such as income poverty, labor shortages on smallholder farms, gender-based discrimination within the household, and risk and vulnerability considerations which force poor people into making tradeoffs between meeting immediate consumption needs versus investing in their children's schooling. Poor people themselves identify education as a road out o f poverty over the l o n g - r ~ nand ~ ~ ~this i s also borne out inour quantitative analysis inChapter 4.253 ~ 252Rahmato and Kidanu, 1999; Legovini, 2004. 253A large beneficial effect o f education in household welfare i s reported by Kronlid (2001). Data comes from three waves o f the Ethiopian urban socio-economic survey for the years 1994, 1995 and 1997. The author estimates returns to education in terms o f the effect o f main income earner's education on log o f per adult monthly income o f the household, and finds returns o f 13 percent, 28 percent, 37 percent and 153 percent respectively for primary, secondary, post-secondary and tertiary education. These effects are all significant at 167 We also explore cross-sectoral linkages and spillover effects across MDGs, and comment on the trade-off between equity and efficiency in reaching these targets. For example, female educational attainment may influence demand for child education as well as child mortality outcomes. Health services interventions (such as deworming or improvinghealth care access) may boost primary school attendance among children. As governments and policy makers seek to maximize the human development impact o f public spending over the next decade in pursuitof the MDGs, it will be crucial to carefully consider synergies betweendifferent types o f interventions, as well as the sequencing o f interventions that will lead to the largest improvements in human development outcomes. While the cost calculation and budgetary impacts o f spending programs that seek to improve human development outcomes have been considered at length in Sectoral Country Status Reports, PERs, and macroeconomic management discussions, the microeconomic perspective o f what drives outcomes and which interventions might have the most direct impact on specific human development outcomes has been the missing link thus far in policy discussions. This part o f the study aims to fill this link. In doing so, we also highlight the magnitudes and nature of disparities in human development outcomes among poorer and richer households (as indicated by their asset or consumption levels). Part I11o f the study i s organized around three interlinked pathways to improving basic human capabilities o f people. These comprise: reducing child malnutrition (Chapter 8); improving health outcomes with an emphasis on child mortality (Chapter 9); and improving primary school enrollment and completion (Chapter 10). For each pathway, we highlight the magnitude o f disparities inkey human development outcomes across regional, ruralhrban and gender lines, explore their determinants usingmicroeconomic data, and pinpoint policy levers and interventions that would be most effective in reducing these disparities and accelerating progress toward the MDGs in Ethiopia. This chapter i s not meant to provide an exhaustive account o f the challenges facing the education and health sectors in Ethiopia.254Rather, our focus i s on the determinants and cross-sectoral linkages between key MDG outcomes and interventions, and the opportunities that policy-makers have to substantially improve people's human capabilities, and especially those o f the poor, beyond the policy levers immediately available intheir sector. Chapter 11concludes. CHAPTER8. CHILD MALNUTRITION,FOOD SECURITY AND ECONOMIC GROWTH 8.1 The Nutrition-PovertyTrap-Another Lost Generationin the Making 8.1 Not only is being well-nourished widely considered one of people's primary capabilities, malnutrition also poses a large burden on current and future economic growth. Early childhood malnutrition (among children between six and 36 months) can cause irreversible damage to brain and motor-skill development, stifle human capital the conventional level. However, other studies have found low private returns to primary education inEthiopia (Appleton, 1995; Krishnan, et. al., 1998). 254For a detailed account ofthe educationand health sector please see World Bank, 2004a and 2004b. 168 formation by causing delays in enrollment and later increasing the probability o f grade repetition and drop-out, lower current health status, and increase the lifetime risk o f chronic diseases associated with premature mortality. Through impairing the cognitive function and process o f skill formation in the next generation of workers, and increasing vulnerability to chronic illness, child malnutrition can have lasting impacts on long-run economic growth and poverty alleviation. Even among adults, hunger and malnutrition can have severe economic consequences. Poor nutritional status reduces an individual's ability to concentrate and engage in strenuous physical activity necessary to grow crops. Micronutrient deficiencies such as anemia are associated with lower worker productivity and reduced Adult workers who do not consume sufficient calories to be productive in manual labor jobs will earn less256and may also face greater risk o f unemployment and illness. As a result their children may also run a higher risk o f experiencing severe malnutrition. 8.2 Ethiopiahas one of the highest child malnutritionrates in the world and due to its extreme vulnerability to recurring droughts coupled with its dependence on rain-fed agricultural production for survival, the Ethiopian population i s especially susceptible to both long-term and transient malnutrition.257 Nonetheless, child malnutrition has slipped from the policy agenda over the past two decades. Ethiopia's SDPRP does not even consider the problem o f child malnutrition. 8.3 The current lack of focus on malnutrition in the policy arena is in stark contrast to the attention child malnutrition received prior to the 1990s, when Ethiopia was recognized as an African center o f excellence in the area o f nutrition policy, research, and practice. Many of Ethiopia's neighbors looked to the Ethiopia Health and Nutrition Research Institute as among the best o f its kind. After the early 1990s, however, advocates saw a decline inthe prominence givento nutrition inboth programs and policy debate, accompanied by a "brain-drain" of top professionals in related disciplines out o f the country. In addition, the status o f the Ethiopian Health and Nutrition Research Institute was downgraded, and responsibility for formulating nutrition policy was dispersed across three different agencies (the Ministry of Health, the Bureau of Food Security under the Ministry o f Rural Development, and the Disaster Prevention and Preparedness Commission) each o f which has significant other responsibilities and agendas. 'j5Li.,et al., 2003; Basta, et.al., 1979. 256Croppenstedt and Muller (2000), provide Ethiopia-specific evidence from the 1994 EHRS of the negative effects of adult undemutrition on agricultural production and wages. They find an output elasticity of 1.90 to 2.26 with respect to adult weight for height of the household head (a measure of short run malnutrition), which means that at the mean an increase in weight for height by one standarddeviation would increase total value of agricultural output by 27 percent. Similarly, they find an elasticity of weight for'height and Body Mass Index (BMI=body weight body height squared) of 3.02 and of 3.04 respectively on wages. This means that at the mean an increase in weight for height and the BMIby one standard deviation would increase wages earned in rural areas by 29 percent and 26 percent respectively. In either case, undemutrition appears to have a strong negative effect on adult labor productivity (production or wages). These figures are toward the upper end of the range reported in the literature, consistent with the observed severity of adult undernutrition in Ethiopia. The average BMI among wage workers inthe 1994 EHRS sample is 19,8 with a standarderror of two. People with a BMIbelow 1.85 are consideredundernourished. 'j7Children between six and 24 months were found to experience about 0.9 cm less growth over a six-month periodincommunities where halfthe crop area was damaged comparedto those without crop damage (Yamano, Alderman and Christiaensen,2005). 169 8.4 This dispersion of responsibility has prevented the formulation of, and agency ownership over, a coherent and comprehensive agenda aimed at reducing child malnutrition during both drought and non-drought years. At the institution level, the problem o f malnutrition has been given second or third priority behind competing sector objectives. Even in cases where malnutrition has received attention, the weak linkages betweenthe various agencies working on different aspects o f nutrition policy-including food security, disaster response planning, health, sanitation, and agriculture-has exacerbated the disconnected way in which nutrition policy i s formulated in Ethiopia. The lack o f transparency concerning the institutional responsibilities regarding nutrition has left nutrition policy largely abandoned despite its importance for long-run economic growth. 8.5 What is malnutrition and how do we measure it? Malnutrition is a condition in which an individual lacks adequate macronutrients (protein and calories) and micronutrients (including zinc, iodine, and iron) to live a fully productive life. Poor nutritional status i s an immediate function o f dietary intake and health status. Infectious diseases such as malaria, intestinal parasites and infectious diarrhea can lead to malnutrition among children even if micro and macronutrients are available by preventing absorption o f key nutrients and reducing effective caloric intake. Broader underlying causes o f child malnutrition at the household level include household food security, the availability and quality o fhealth services and the sanitary environment, as well as the use o f appropriate caring practices by the household. Physiological measures used to identify an individual who is suffering from malnutrition include: anthropometric indicators (such as low birth weight, stunting, wasting and underweight status in preschool or school-age children, and low body mass index among adults or adolescents); clinical indicators (including iodine deficiency); and laboratory indicators (including low hemoglobin in preschool or school-age children and non-lactating, non-pregnant women).258 Key measures o f nutritional status and their interpretation are shown in Table 8.1 below. 258Behnnan,Alderman, andHoddinott, 2003. 170 Table 8.1: Measures of nutritional status Indicator Interpretation Anthropometric Indicators L o w birthweight An indicator o f intrauterine growth retardation resulting from short maternal stature, poor maternal nutrition before or during pregnancy, infection and smoking. L o w height-for-age (stunting) in Children's skeletal (linear) growth compromised due to constraints to preschool or school-age children one or more o f nutrition, health, or mother-infant interactions. This i s an indicator o f chronic nutritional deprivation. L o w weight-for-age (undernutrition) Children suffer thinness resulting from energy deficit and/or disease- inpreschool or school-age children inducedpoor appetite, malabsorption, or loss o f nutrients. This i s an indicator o f transitory nutritional deprivation, though it usually correlates well with l o w height for age. This i s an MDG indicator. L o w weight-for-height (wasting) in Children suffer thinness resulting from energy deficit and/or disease- preschool or school-age children induced poor appetite, malabsorption, or loss o f nutrients. This i s an indicator o f transitory nutritional deprivation. L o w body mass index in adults or Adults suffer thinness as a result o f inadequate energy intake, an adolescents uncompensated increase inphysical activity, or (severe) illness. Clinical Indicators Iodine deficiency Iodine deficiency results from low intake o f iodine in the diet. Laboratory Indicators L o w hemoglobin (anemia) in Children suffer from anemia either as a result o f low iron intake or preschool or school-age children poor absorption, or as a result o f illness. Severe protein-energy malnutrition and vitamin B12ifolate deficiency can also lead to anemia. L o w hemoglobin (anemia) innon- Women suffer from anemia as a result o f low iron intake, poor lactating, non-pregnant women. absorption, illness, or excessive loss o f blood. Severe protein-energy malnutrition and vitamin BlYfolate deficiency can also lead to anemia. Anemia is rare in adult men except inconditions o f extreme iron deficient diets. Source: Behrman, Alderman, and Hoddinott, 2003, quoting ACC/SCN, 2000a, Morris, 2001 8.6 There are three key channels through which child malnutrition can create povertytraps. First, severe malnutrition during the first few years of life negatively impacts the cognitive and psycho-motor development of infants, often causing irreversible damage. Fetal iodine deficiency i s known to cause irreversible damage to central nervous system de~elopment.~~'Children who have suffered from nutritional deficiencies early in life perform worse on aptitude tests.260 Some studies suggest that the retardation o f psycho-motor development in young children due to early childhood malnutrition may also be permanent. Inparticular, the motor development skills ofinfants hospitalized for severe malnutrition were found not to recover at all, evenafter six months of intensivenutritional rehabilitation.261 259 Dasgupta, 1997. 260 Pollitt, 1990; Dasgupta, 1997. 26' Chavez and Martinez, 1984. 171 Second, child malnutritiondirectly impacts healthstatus as well as risk of illness during childhood and adult years. Low-birth-weight (LBW) babies tend to be sicker and have higher mortality rates than non-LBW babies. Alderman and Behnnan (2003) find that in low income countries for every birth that occurs in the 2,500-3,000 gram range instead o f the 2,000-2,499 gram range, the probability o f neonatal or post-neonatal infant mortality decreases by 7.8 percent. Likewise, a study on both fratemal and identical twins suggests that an additional pound at birthdecreases the probability o f infant death duringthe period 28 days to one year by 14 percent.262 In addition, malnutrition during childhood i s found to be associated with chronic diseases that can cause premature death among adults, including coronary heart disease, non-insulin dependent diabetes, high blood pressure, obstructive lung disease, highblood cholesterol, and renal damage.263 Third, malnutrition and poor nutritional status among children directly impacts their educationalattainment, learningeffectiveness and efficiency,as well as their probability of grade repetition. Malnutrition and poor nutritional status may cause caregivers to be less willing to invest in 'schooling for a child (due to poor cognitive capacity). Malnourished children tend to have delayed entry into primary school (which is associated with lower lifetime earnings), and are more likely to repeat grades. Empirical research from Ghana suggests that for each year o f delay in entry to primary school a student loses three percent o f lifetime wealth.264 Girls who have low educational attainment are more likely to have malnourished children later in life and children who die prematurely.265 Malnourished childrenare less likely to pay attentioninschool and less able to studyproductively thanwell- nourishedpeers.266This i s true even ifthe cognitive capacity o fmalnourished children has not been permanently impaired. Malnutrition among school children thus may impede human capital formation and reduce educational effectiveness. 8.7 Early childhoodmalnutritionhas long-runimpacts on the economic and physical well-being of the poor throughout their lifetimes. Physical stunting caused by severe malnutrition during childhood i s associated with lower earnings as an adult. Thomas and Strauss (1997) find that a one percent increase in height among adult workers inurban Brazil leads to a two-2.4 percent increase in lifelong wages or eamings. Inaddition, physical height has been found to be a statistically significant explanatory variable for wages in the United States.267 Workers who suffered from early childhood malnutrition are likely to have low cognitive achievement as adults, and low schooling attainment which translates into lower labor productivity and lifetime eamings2@' Lowered stature as a pre-schooler following 262Conley, Strully and Bennett, 2003. 263Barker, 1998. 264Glewwe and Jacoby (1995) find delayed enrollments among the malnourished in their cross-sectionalstudy, but not different in the total years completed. By contrast, Alderman, Hoddinott and Kinsey (2003) track a cohort of Zimbabweans over two decades and find both delayed school entry and lower grade completed for malnourished children. Corroborating evidence from India suggests that better nourished children start school earlier and repeat fewer grades (Glewwe, Jacoby and King, 2001). 265Alderman, Hoddinott, and Kinsey, 2003. 266Behrman,Alderman, and Hoddinott, 2003. 267Strauss and Thomas, 1998. 268 See Martorell, 1995; Martorell, Rivera and Kaplowitz, 1989; Haas et al., 1996; Martorell, 1999; and Martorell, Khan and Schroeder, 1994. 172 exposure to the 1982-84 drought in Zimbabwe was found to result in a permanent loss o f stature o f 2.3 cm, a delay in starting school o f 3.7 months, and 0.4 grades less of completed schooling. The combined effect o f these different factors was estimated to reduce lifetime earnings by seven percent.269 Against this background, we consider recent trends in, and causal determinants of, child malnutrition in Ethiopia, and seek to highlight inexpensive policy options that could potentially enable the country to significantly reduce malnutrition, especially among children. 8.2 A Profile of ChildMalnutritionin Ethiopia 8.8 Since the early 198Os, child stunting prevalence in Ethiopia has persisted at around 60 percent, and is among the highest in the Child stunting, which i s measured as abnormally low height for age in children, i s an indicator o f poor long-run nutritional status. Although the prevalence o f child stunting in Ethiopia decreased during the second half o f the 1990s, the prevalence in2000 was still significantly above the Sub Saharan African average o f 34 percent (estimate excludes Ethiopia), and only slightly below the 1983 level o f 60 percent, indicating significant scope for further reduction over the next decade (Figure 8.1). Child wasting (or abnormally low weight for height) i s a measure o f short-run nutritional deprivation in children. Child wasting prevalence in Ethiopia has recently increased, from eight percent during the 1990s to 10 percent in 2000, which i s slightly above the sub-Saharan average o fninepercent. Figure 8.1: Percentof childrenwasted and stunted in Ethiopia, 1983-2000 80 b 50 s 10 0 1983 1992 1996 )'ear 1997 1998 2000 ') stunted = height-for-age z score < -2; wasted = weight-for height z score -2. *) The 1983 nutrition survey covers about 9,000 children across all rural areas except Tigray; the 1992 nutrition survey covers more than 20,000 children across all rural areas; the 1995196, 1997, 1998, and 1999/2000Welfare MonitoringSurveys cover about 6,000, 8,000, 29,000, and 15,000 children respectively. 3, 1983 and 1992 surveys comprise 6-59 months old children, the other surveys 3-59 months old children. 4, 1996 and 2000 figures based on lst 2"dround of the W M S (June 1995iJanuary 1996, and June 1999/January 2000 and respectively) Source: CSA, Rural Nutrition Survey, 1983; CSA, Rural Nutrition Survey, 1992; CSA, Health and Nutrition Survey, 1998; CSA, WeEfareMonitoring Surveys, 1995196, 97, and 1999/2000 269Alderman, Hoddinott, andKinsey, 2003. 270ChristiaensenandAlderman, 2004. 173 8.9 The nationalestimates of child stunting and wasting prevalence mask significant urban-rural and regional variations (Figures 8.2 and 8.3). Perhaps the largest and most persistent disparities in child nutritional outcomes in Ethiopia exist between rural and urban areas. Taking averages over the four most recent years for which we have data, child stunting prevalence in rural Ethiopia i s 63 percent, versus 50 percent in urban areas. The average disparity in wasting prevalence betweenrural and urban areas i s not as large, with prevalence averaging ninepercent inrural areas and seven percent inurban areas. Figure 8.2: Child stunting and wasting in urbanvs. rural areas, 1996-2000 70 60 50 -g 40 -d6m z 30 ri 20 10 9 Rural Wasting 0 Average Stunting Prevalence Average Wasting Prevalence ('96, '97, '98, '00) ('96, '97, '98, '00) Nutrltional Indicator 8.10 The regions with the poorest long-run nutritional indicators are Amhara, Tigray, Oromiya and SNNP (Figure 813). Amhara i s the only region with a higher child stunting prevalence than the national average (65 percent versus 57 percent) in 2000. Addis Ababa, Dire Dawa, Afar and Gambela all have relatively lower child stunting prevalence (falling in the range o f 37 to 42 percent). Improvements in the long-run nutritional status o f children have been slowest inAmhara and Oromiya, two regions with among the highest child stunting prevalence in Ethiopia. The regional variations in child wasting prevalence are not as large. While Gambela had a relatively low child stunting prevalence in 2000 (41 percent versus 57 percent for the country overall) child wasting prevalence i s above the national average (13 percent versus a 10 percent national average). Most regions in Ethiopia have child wasting prevalence inthe 11-12 percent range. Notable exceptions are Oromiya and S N " (each with ninepercent child wasting prevalence), and Harari and Addis Ababa (each with five percent prevalence, half the national average). Child wasting and stunting prevalence are not perfectly correlated across Ethiopia's regions, suggesting that transient malnutrition i s more pervasive insome regions o f Ethiopia, and chronic malnutrition in others. Regional variations inchild nutritional outcomes may be linked to differences indrought incidence and severity, illness prevalence, and coverage o f food security programs. 174 Figure8.3: Child wasting and stuntinginEthiopia by region, 2000.'' lm PercentChiidrenWasted in 2000 ~ 1 1 "-I 56 57 51 40 37 7 Child stunting i s defined as height-for-agez score <-2. Source: CSA, Rural Nutrition Survey, 1983: CSA,Rural Nutrition Survey, 1992; CSA, Health and Nutrition Survey, 1998: CSA, Welfare Monitoring Surveys, 1995/96, 97, and 1999/2000 8.11 Gender-based disparities are relatively small in Ethiopia, and opposite to the commonly expected direction. Malnourishment among boys is larger than among girls. Averaging across four years for which we have data, male stunting prevalence has averaged 63 percent, or four percentage points higher than the prevalence among girls. Similarly, while nine percent o f boys in Ethiopia are wasted, only eight pertent o f girls are. The finding that boys are more likely to be malnourished than girls in Ethiopia i s consistent with nutritional outcomes in other African countries27', but opposite the patterns observed in Asia and Latin America, where girls are consistently discriminated against inthe allocation o f food resources within the household. The reasons for this difference are not fully understood. 271Svedberg, 1990; Sahn and Stifel, 2000. 175 8.3 Determinantsof ChildMalnutrition 8.12 Empirical analysis using three years of Welfare Monitoring Surveys augmented with data on food prices was used to model the determinants o f children's long-run nutritional status in Ethiopia as measured by height for age z-scores ( H A Z - s ~ o r e s )among children ~ ~ ~ aged three to 60 months.273 The results suggest that the key determinants of long-run child nutrition outcomes in Ethiopia include the following: household income, female adult education, community nutritional knowledge, and food prices. W e review the empirical findings o f this study in greater detail below. Full regression results are presented in Table A.8.1 inAppendix 3. 8.13 There is only limited catch up growth after the age of three. The coefficients on the age variables suggest that a malnourishedchild's standardized height deteriorates up to the age o f three, and only slightly improves thereafter. This shows that there is very limited scope for catching up on child growth perfomance after age three, and i s consistent with the growing body o f evidence indicating that growth lost in early years is at best only partially regained during childhood and adolescence, especially when children grow up in poor environments. In other words, interventions to improve children's nutritional status must be targeted to those below three years. 8.14 The effect of female education is about twice as large as that of male education, though both have a large positive effect. Each year o f primary or secondary education o f the most educated female adult in the household increases, ceteris paribus, a child's HAZ- score by 0.03. In other words, were all women to complete six years o f primary schooling, the current average gap between the nutritional status o f Ethiopia's children and the nutritional status o f the reference population would be reduced by seven percent. 8.15 Post secondary schooling also has a strong and positiveeffect on child nutritional outcomes. The HAZ-score o f children inhouseholds where a female adult has obtained post secondary education i s on average 0.23 larger than those in households where the most educated female adult completed secondary schooling, and 0.60 larger than those in households with no formally educated female adults. The current average HAZ-score o f Ethiopian children between six and 60 months old i s -2.48. Children with HAZ-scores below two are considered stunted and those with HAZ-scores below three are considered severely '''To assess chronic child malnutrition, a child's height for age score is compared to the average height for age score of a child o f the same sex ina well nourished reference population. By dividing the difference between the child's height for age and the average o f the reference population by the standard deviation, one obtains the HAZ-score. Children with a HAZ-score below minus two, implying that that 95 percent o f the reference populationhas a higher height for age score, are usually considered stunted; those with HAZ-scores below minus three, implying that 99 percent o f the reference population has larger HAZ-scores, are considered severely stunted. 273 Data sources used in the analysis include individual, household and community level data used in the empirical analysis are taken from the 1995196, 1997 and 1998 Welfare Monitoring Surveys (WMS), the 1995196 household income and expenditure survey (HICES), the 1998 Health and Nutrition Survey (HNS), and regional price data from various quarterly statistical bulletins o f the Ethiopian Central Statistical Authority (Christiaensen and Alderman, 2004). 176 stunted, The education effects discussed here are supplementary to the positive effects o f education on child malnutrition through the enhancement o f income. 8.16 Increasing income helps reducing child malnutrition, but other complementary actions will be needed. As expected, households with greater financial resources have less stunted children, though the effects of income are not large. A 10 percent increase in expenditures removes only 0.7 percent o f the gap between the current average HAZ-score in Ethiopia and the reference standard. This estimated effect o f income on nutritional status i s within the range estimated in other countries across the world, highlighting the fact that economic growth alone will not suffice to reduce child malnutrition inEthiopia. 8.17 The presence of a private tap and flush toilet in a household has a positive effect on child height.274This finding suggests that there may be important health linkages with child nutritional outcomes. Inboth cases the effect i s substantial, though the coefficients on flushtoilets are estimated with less precision. No correlation was found betweenchildheight and indicators that capture access to communication infrastructure, such as ownership o f radios or televisions. 8.18 Children's HAZ-scoresare very responsiveto relativepricevariations for staples and key inputs in the cooking process. Higher teff, kerosene, and charcoal prices are associated with shorter children and the effect i s large and statistically significant. The negative effect o f higher teff prices on child malnutrition holds after controlling for long run geographical price differences, though it becomes less pronounced. Higher maize, sorghum, beef, and milk prices on the other hand are associated with taller children.275However, when accounting for geographical price differences, the signs on the sorghum and beef coefficients reverse. 8.19 The most striking finding is the positive effect of maternal nutritional knowledge276on child growth faltering, over and above the effect o f maternal formal education, household income, and other determinants o f child malnutrition. As indicated by other if mothers obtain their nutritional information outside school, communal nutritional practices and attitudes must be key in shaping individuals' nutritional knowledge, especially in areas where communication infrastructure (like radios and television sets) are virtually absent. The empirical analysis suggests that increasing the community's ability to correctly diagnose stunted and nonstunted children by.25 percentage points would have effects similar to providing at least one female adult per household with primary education. 8.20 Harvest failure causes substantial child growth faltering. Empirical examination of the impact o f shocks on child growth inEthiopia indicates that children betweensix and 24 274Access to other sources o f drinking water which are generally deemed safe, such as public taps and protected wells, did not (positively) affect children's height. They were subsequently omitted from the analysis. 275To understand the positive effect o f beef prices on chronic child malnutrition, note that people tend to substitute beef for other foods when beef prices decline. As calories and proteins from animal products are relatively more expensive, people's net nutrient intake tends to drop in the process. A similar process would lead to a positive association between milk prices and the height o f children. 276Proxiedby the community's diagnostic capability o f growth faltering. 277Glewwe, 1999. 177 months experienced about 0.9 cm less growth over a six month period in communities where half the crop area was damaged compared to those without crop damage.278 Furthermore, contrary to the usual gender bias against girls, the growth o f girls under two suffered less from income shocks than the growth o f boys under two. This could be related to greater biological resilience or intra-household dynamics, and i s consistent with the observation that boys are more malnourished than girls. 8.21 Food aid279has helped reduce child malnutrition, though its effectiveness in protectingchild growth from shocks has been somewhat muteddue to inflexible targeting rules in the face o f shocks. Children in communities that received food aid grew on average two cm faster in a six month period than ifno food aid would have been available. This helps compensate poor child nutrition and growth in communities that are targeted for food aid. In addition, the total amount o f food aid appears on average sufficient to offset the negative effects o f plot damage on child growth in food aid receiving communities. This result i s encouraging as it indicates that food aid has indeed been effective in protecting early childhood growth from droughts and other income shocks in food aid receiving communities. Yet at the same time food aid reception appears to be largely determinedby factors other than shocks, and many communities that experience shocks do not receive food aid. Insum, while food aid has helped to reduce child malnutrition in Ethiopia, inflexible food aid targeting in the face o f shocks, together with the endemic nature o f poverty and extremely low levels of maternal education in Ethiopia, has also left child stunting rates at alarming levels. Given the large impact of shocks on child malnutrition, food aid targeting rules more responsive to shocks, as well as other insurance mechanisms, are called for. 8.4 PolicyActions Neededto Halve the Prevalenceof Pre-SchoolStunting 8.22 What levers are available to government and donors to help reduce the incidence and severity of child malnutrition inEthiopia? The authors o f the nutrition study discussed above also simulate the effect o f different policy interventions on long-term child nutritional outcomes (as measured by child HAZ-scores) in Ethiopia. These simulations, described in greater detail below,280enable us to determine the relative importance of enhancing private incomes versus expanding publicly provided services such as education, direct nutrition interventions, and programs to protect against large price and income shocks within the Ethiopian context. 8.23 First, the study simulates the impact of sustained income growth of 2.5 percent per adult equivalent over a 15 year period. Though challenging, this is similar to the per capita growth rate neededto reach the poverty MDG. The results suggest that fifteen years o f sustained per adult equivalent income growth o f 2.5 percent would reduce chronic child malnutrition prevalence by only up to four percent. Income growth alone will clearly not suffice to eliminate child malnutrition inEthiopia. 278Yamano, Alderman, and Christiaensen,2005. 279Ethiopia is one of the major recipients of food aid worldwide. Approximately one-fifth to one-quarter of all food aid deliveries to Africa over the past decade have gone to Ethiopia, with food aid estimated at 20 percent of domesticproduction indrought years (Jayne et al., 2003). 280Complete regressionresults are provided inTable A.8.2 inAppendix 3. 178 Figure 8.4: Prevalence of child ~t~~~~~~~in Ethiopia 1.71) 8.26 Fourth, the study explored the effect of a 25 percent increase in real prices of maize, teff, and sorghum, correspondingto observed price increases during drier years. This simulation is important given the drought-prone nature o f the Ethiopian climate, and climate dependent nature o f its economy. It sheds light on the malnutrition reducing potential of controlling price fluctuations by for example fostering cereal marketintegration andthe use o f grain reserves and warehouse receipts. The results suggests that an increase in cereal prices by 25 percent, not at all uncommon even within the time span o f one year (Table A.8.2) would increase the prevalence o f child malnutrition by three to four percent. While higher cereal prices would diminish the nutritional status o f children in net cereal consuming households, they may also positively affect the nutritional status o f children in net cereal producing households if the indirect income effect exceeds the direct consumption effect. However, as discussed in Chapter a substantial group o f poor Ethiopian households are estimated to be net cereal buyers. Moreover, both net producers and net buyers may often sell at low prices during the immediate post harvest season and buy at high prices during the hunger season. 8.27 Fifth, interventions reducing crop damage could substantially increase early childhood growth in the critical age range for intervention. For example, interventions which reduced the crop area damaged by 25 percentage points could increase height growth among children between six and 24 months old by almost 0.5 cm over a six month period. Better targeting o f food aid in response to covariant and idiosyncratic shocks at the household level i s called for, though additional interventions to strengthen households' capacity to cope with shocks such as rainfall based insurance and strengthening o f informal insurance mechanisms through local institutions (e.g. iddirs) should be explored as well. In addition, better pest and water management techniques would be important avenues through which to prevent the occurrence o f crop damage in the first place. The relative cost effectiveness o f strengthening ex ante risk mitigation versus ex post coping capacity needs to be further examined. 8.5 ConcludingRemarks 8.28 Reducing the prevalence of child stunting and other forms of malnutrition is imperativefor long-run economic growth. Early childhood malnutrition impairs children's cognitive ability, delays their enrollment in school and reduces the number of grades completed. It further increases the risk o f future illness. The combined effect on future earnings can be substantial. And yet despitethe alarming prevalence o f child stuntingrates in Ethiopia which have persisted over the past two decades, child malnutrition has largely been neglected inthe policy debate inEthiopia. 8.29 Simulations suggest that Ethiopia is unlikely to reach the MDG of halving the prevalence of child stunting from its 1990 level. Nonetheless, substantial progress toward this goal can be made by expanding household income and increasingparental and especially female adult education. In addition, it appears that substantial benefits can be derived from strengtheninghouseholds' ability to reduce crop damage caused by pests and droughts as well as from strengthening their coping capacities to deal with shocks ex post, either by more shock responsive targeting o f food aid or through the development o f other insurance mechanisms such as weather based insurance. Similarly, actions which reduce cereal price 180 increases during drier years would be helpful in reducing child stunting. Finally, interventions aimed at improving sanitary conditions and health infrastructure are likely to have positive effects as well, though this was not confirmed by the data and may in effect be related to the poor quality of current health services.282 8.30 Child growth monitoring and maternal nutritional education programs could play an important complementary role to other development actions such as promotion o f food security, income growth, and parental education, which are already underway. Moreover, while it will take a considerable amount o f time before these other interventions substantially affect pre-school child growth faltering, child growth monitoring and nutritional education programs could take effect immediately. The role o f other direct nutrition interventions such as micronutrient supplements, promotion of exclusive breast-feeding, and appropriate complementary feeding should be equally considered. 8.3 1 Most importantly, increasingawareness about the long term detrimental effects of early childhood malnutrition on future economic growth is necessary. A comprehensive and coherent multi-sectoral nutrition policy will need to be developed, and the institutional responsibilities o f the different ministries and coordination o f their actions and interventions in the field o f nutrition will need to be clarified. Given its critical importance for economic growth, this agenda should be o f great concern to the Ministry of Finance and Economic Development. 282World Bank, 2004b. 181 CHAPTER 9. LIVINGAND GROWINGINTHE FACE OFDISEASE 9.1 Health and Poverty 9.1 Life expectancy and the quality of life associated with good health form the most fundamental human capabilities. In addition to its intrinsic value, good health has also instrumental value by enhancing worker productivity, facilitating human capital formation, and thus paving the way for long-run increases in living standards and the rate o f economic growth. The private and social costs o f poor health to society are very high. Children who suffer from infectious disease are more likely to be malnourished and less able to pay attention in school and to learn effectively.283 Poor health among older workers can negatively impact labor productivity and reduce lifetime income. Infectious diseases that reduce labor supply on smallholder farms, and potentially deplete household savings as a result o f increased burial and health care costs, may reduce food security and perpetuate the vicious cycle o f illness, poverty and hunger. Certain illnesses, including malaria and HIV/AIDS, pose large threats to both the physical and economic well-being o f the overall population indeveloping countries. 9.2 Illness and chronic health problemshave the ability to trap workers in a vicious cycle of poverty. There are often large feedback effects between health and income for the poor. Workers ingood health are better able to increase their income levels over time. Rising income makes it more likely that they will undertake health investments, thus ensuringbetter health status over the short and long run. Among children, poor health often interacts with malnutrition to prevent skill formation. Illness among school children i s associated with poor performance on aptitude tests.284 9.3 Achieving the healthMDGs in Ethiopia will be a dauntingtask, and will require both supply and demand side interventions that not only increase access to health care services among the poor but also increase individual awareness o f the negative consequences and causes o f poor health. Ethiopia currently has approximately 119 hospitals, 45 1 health centers and 311 pharmacies. Not only does a large proportion o f the population thus lack access to a basic health but the number o f physicians and nurses per capita is among the lowest in the world. Presently there are approximately 1,888 physicians and 12,838 nurses for a population o f around 67 million people, corresponding to a doctor-patient ratio o f one to over 35,000, possibly the lowest in the world. Not only are health professionals in Ethiopia extremely scare, but they are also predominantly male and concentrated in urban areas, which reinforces gender and rural-urban disparities in health access, utilization and outcomes. Attempts to increase the cadre o f trained physicians will be constrained by the paucity o f secondary school graduates, as well as the low wages among surgeons and general practitioners compared to what they could earn in other countries. This leads to high emigration and staff attrition rates among key health care workers. 283Dasgupta, 1997. 284Pollitt, 1990; Dasgupta, 1997. 285Overall, 61 percent of the population (75 percent of urban households and 42 percent o f rural households) i s within 10 kilometeres of a health facility. 183 9.4 Even if the poor have access to health care services, their utilization rates are very low. For example, results reported inthe Health CSR (World Bank 2004b) suggest that 25.4 percent o f households in the poorest quintile versus 19.3 percent o f households in the richest quintile reported recent child diarrhea incidents within the two months prior to the survey. While only 21 percent o f those in the poorest quintile sought treatment, a much higher 43 percent o f households in the richest quintile sought treatment (see Table A.9.1, Appendix 3). Among those households that did seek treatment, households inthe top quintile were more likely to seek treatment inhospitals and health centers where the quality o f care is likely to be higher than households in the bottom quintile.286Poor access and low health service utilization rates reinforce poor health outcomes among the poor in Ethiopia. Policies that aim to improve the health outcomes o f the poor will thus not only need to consider their level o f access but also behavioral and institutional factors that influence their utilization o f health services when they are available. Availability/access (38 percent) and quality o f care (23 percent) were the two main reasons for choosing a facility. 9.5 In light of these observations, we would ideally like to examine the effects of both health supply and demand side determinants on health outcomes, as we will do in the case o f education outcomes in the next section. However, data limitations and methodological considerations prevent us from doing so. Instead, this chapter will focus on one key health outcome, the prevalence o f child mortality under five, one o f the health MDGs. It will in particular examine the relative importance o f a series o f household and community characteristics that are commonly found to be important inreducing child mortality, though it will not consider the effect o f the quality o f the health system per se. Before discussing the determinants o f child mortality, we first present a brief health profile o f the Ethiopian population. 9.2 HealthProfileof the EthiopianPeople 9.6 Illness incidence may be increasing, and women tend to be sick slightly more often than men. Illness events are both widespread and frequent in Ethiopia, and a recent household survey suggests that their prevalence i s increasing over time. While the percent o f household heads that reported being sick within the two months prior to being surveyed was 26 percent in 1995, the illness prevalence had increased by 1999 to 37.7 percent. The average incidence o f poor health varies regionally as well as betweenmen and women. As illustrated in Table 9.1 below, women are typically more likely to have experienced illness events in Ethiopia than menduring the two months prior to the survey. 286World Bank2004b. 184 Table 9.1: Self reportedincidenceof healthproblemsduring the past two months 1995 (%) 1999 (%) Male Female Both Male Female Both Tigray 19.7 20.1 19.8 30.5 31.1 30.8 Afar 19.0 17.5 18.8 23.8 26.5 25.1 Amhara 19.8 24.2 20.5 28.4 30.3 29.3 Oromiya 17.0 16.3 16.9 24.8 27.9 26.4 Somali 13.6 16.6 14.0 29.7 37.3 33.5 Benishangul 36.3 44.8 37.7 40.0 36.3 38.1 SNNPR 16.6 23.8 17.8 24.2 27.1 25.7 Gambela 24.8 30.3 26.3 31.0 34.9 32.8 Harari 11.0 8.6 10.4 23.6 26.2 25.0 Addis Ababa 8.9 8.6 8.8 15.7 18.0 17.0 DireDawa 9.1 5.6 8.1 34.5 37.6 36.2 Urban 14.1 13.9 14.0 17.5 21.2 19.5 Rural 18.2 22.0 18.8 27.2 29.7 28.4 National 17.8 19.8 18.1 25.9 28.4 27.2 Source: Ethiopia Welfare Monitoring Surveys 1995 & 2000 9.7 Malaria is one of the leading causes of inpatient days in hospitals and outpatient visits in Ethiopia. The disease burdeninEthiopia is heavily biasedtoward physical ailments that are associated with poverty, economic destitution and low health access and utilization. Communicable diseases are among the top 10 reasons for outpatient visits, inpatient admissions and death inEthiopia, many o f which can be easily preventedusing a combination o f targeted health marketing and inexpensive public interventions (Table 9.2). The top causes of death in Ethiopia imply an age distribution o f illness that i s heavily tilted toward infants and children. Again, the main causes o f death among children in Ethiopia are communicable diseases. Approximately 28 percent o f under-five child deaths in Ethiopia are attributed to pneumonia-a disappearing cause o f death in many poor countries-and 24 percent to diarrhea. HIV/AIDS and malaria account for 6.2 and 4.5 percent of child deaths re~pectively.~'~HIV/AIDS i s a growing cause o f death o f children in Ethiopia. 287World Bank2004b. 185 Table 9.2: Top 10 reasonsfor outpatientvisits, inpatientadmissions and death Outpatient Visits Inpatient Admissions Causes o fDeath Rank Number % Number % Number % 1 All types o f 328760 10.4 All types o f 16782 14.8 Tuberculosis o f 1005 10.0 malaria malaria respiratory system 2 Helminthiasis 213195 6.7 Pneumonia 10090 8.9 Pneumonia 734 7.3 3 Acute upper 205129 6.5 Tuberculosis of 8881 7.8 All types o f 462 4.6 respiratory respiratory malaria infection system 4 Bronchopneum 173123 5.5 Accidents 6976 6.2 Bacillary 224 2.2 onia dysentery 5 Infections o f 145680 4.6 Abortion 4449 3.9 Accidents 156 1.6 skinand subcutaneous tissue 6 Gastric and 137942 4.4 Pregnancy, 4326 3.8 Meningitis 149 1.5 duodenites childbirth& the puerperium 7 Dysentery 111938 3.5 Cataract 2735 2.4 Hypertension 142 1.4 8 Tuberculosis o f 70526 2.2 Bacillary 1848 1.6 Gastroenteritis 109 1.1 respiratory dysentery & colitis system 9 Sexually 68733 2.2 Gastroenteritis 1707 1.5 AIDS 83 0.8 transmitted & colitis infection 10 Bronchitis, 58594 1.8 Meningitis 1015 0.9 Leishmaniasis 48 0.5 chronic and unqualified Total o f all the 1513620 47.8 Total o f above 58809 51.9 Totalofallthe 3113 31.1 above cases cases above cases Total ofall 3167514 100.0 Total o f all 113365 100.0 Total o f all 10006 100.0 cases cases cases 1) PPD M o H 2002 version didnot provide updated information based on the above format Source: World Bank 2004b 9.8 The health MDGs cover four key outcomes in Ethiopia that are directly or indirectly related to the major causes o f hospital visits and mortality in Ethiopia. We highlight these MDGsbelow. 9.2.1 HIV/AIDS 9.9 The MDGs call for halting the spread o f HIV/AIDS by 2015, and beginning to reverse its spread. Currently, Ethiopia has a generalized2@HN/AIDS epidemic with about 1.5 million people living with the virus according to the most recent estimates289(see Table A.9.2 for further details on the burden o f HIV/AIDS to Ethiopia and similar figures for other Sub Saharan African countries). As discussed in Chapter 1, unabated progression o f the epidemic 288 A generalized HIV/AIDS epidemic refers to a situation where HIV has spread far beyond the original subpopulations with high-risk, which are now heavily infected. Prevalence among women attending ante-natal clinics is five percent or more. 289Federal Democratic Republic o f Ethiopia, 2004. 186 will bring not only greater human suffering, but also threatens Ethiopia's prospects for economic growth and poverty alleviation. The government has established a National HIV/AIDS Prevention and Control Office under the Prime Minister's Office. It includes members from across the different sectors and societal organizations and it has developed a multi-sectoral approach to combat the pandemic (The Ethiopian Multi-Sectoral Action Plan). The approach is unique inthat it sought active community participation. Its implementationis anchored around three guiding principles: flexibility (learning by doing), speed (quick implementation) and coverage (scaling up). 9.10 Despite a slow start, the program is now in full swing. A functional institutional framework has been created with HIV/AIDS Councils and Offices at the national, regional, woreda and community levels coordinating and facilitating the different interventions. About 30 federal government offices and 125 regional bureaus have accessed funds to implement their HIV/AIDS prevention plans. 7,960 in-school and 400 out o f school anti-AIDS youth clubs have been formed. The level o f awareness on HIV/AIDS i s now nearly 97 percent, and the demand for condoms has grown tremendously from 4.1 million in 1999 to 67.6 million in 2002, with kiosks, health institutions, NGOs, workplaces and anti-AIDS youth clubs as major outlets. There are currently 185 Voluntary Counseling and Testing (VCT) centers and the demand for VCTs i s on the rise. It is, however, as o f yet still unclear how all o f these positive developments are affecting the incidence of HIV/AIDS given the absence o freliable data. 9.2.2 Malaria 9.11 Similarly, the MDG for malaria i s to halt its spread and to begin to reverse its incidence by 2015. Malaria i s one o f the leading causes o f inpatient days in hospitals and outpatient visits in Ethiopia. Controlling the disease also has the potential to expand access to cultivable land, and to alleviate the intense land pressure that characterizes Ethiopian agriculture. The impact o f malaria on the poor i s further discussed in Chapter 4. 9.2.3 Maternal mortality 9.12 Ethiopia has one o f the highest matemal mortality rates in Sub Saharan Africa (if not the world), ranging from 790 to 3,200 (per 100,000 live births). Highmatemal mortality rates are related to the low levels o f anti-natal care and birth attendance by health professionals- especially among the poor-in Ethiopia. The MDGs call for reducing the matemal mortality rate by three-quarters between 1990 and 2015. Reducing the maternal mortality rate in Ethiopia by the amount envisioned by the MDGs will be challenging given the country's extreme shortage o f obstetricians and highly trained professional staff to attend births. Expansion o f clinical services i s indeed the only route to significantly reduce the rate, but this will take more time and resources. 9.2.4 Child mortality 9.13 Ethiopia's under-five mortality rate i s among the highest in the world, presently estimated at 166 per 1,000, Over the past decade, Ethiopia has made improvements in reducing child mortality. In fact, child mortality rates in Ethiopia have fallen steadily since 187 the 1960s, and Ethiopia has now caught up to the Sub Saharan African regional average. However, the rate o f decline i s far from sufficient to reach MDG goals in the foreseeable future. We turn now to a more careful examination o f the variation in child mortality rates across different groups within Ethiopia, and the underlyingdeterminants. 9.3 Determinantsof ChildMortality 9.14 Under-five child mortality rates vary by wealth and are highest in rural areas, where low health access, malnutrition and extreme poverty interact to cause premature death among children. In Figure 9.1 we present child mortality rates by wealth quintile for rural households.290 Figure 9.1: Under-five child mortality by wealth quintile 200 , ea 180 i 160 - 140 - =5 120 0 - 8 r 100 - &n 80 - 5 60 - 8 40 - 20 - 0 Poorest 2nd Poorest Middle 2nd Richest Richest Wealth Quintiles Source: World Bank, 2004b, using DHS 2000 data 9.15 Except for the lowest quintile, which surprisingly has a relatively low rate o f child mortality, there i s a steady decline inunder-five mortality as wealth increases. However, even the richest families lose almost one in eight children before the age o f five. Incontrast to the gender bias observed ineducation, girls do not have higher mortality rates relative to boys. In fact, similar to other African countries, the child mortality rate i s slightly higher for boys than girls. There are, however, significant differences across regions, with children in Addis Ababa twice as likely to survive to the age o f five as those inGambela (Figure 9.2). , 290 We draw upon data from Ethiopia DHS 2000. Note results are from rural households who have children born at least five years prior to the survey (to control for censoring). 188 Figure 9.2: Regional variation in infant mortality and under five mortality rate') 2 5 0 45 ,200 .-;I50 e, 0 0 k z 1 0 0 0, B 3z P .r 5 0 0 1) Number o f infants out of 1000 live births dyingbefore their first birthday. Source: World Bank 2004b 9.16 Inorder to examine the relative importance of key determinants of child mortality, we estimate a reduced-form equation (Table 9.3). The results presentedinTable 9.3 highlight the fact that some o f the key policy levers for achieving health outcomes, such as water and education, lie outside the health sector. Descriptive statistics on the determinants are reported inTable A.9.1 inAppendix 3. 9.17 The strong positive effect of mother's education is robust across these specifications: mothers in rural areas with an additional four years o f education are 4.8 to 6.4 percentage points less likely to see their children die before the age o f five. Inurban settings, the effect i s less but still strong: four more years o f education reduces under-five mortality by between 2.4 and 3.2 percentage points. The beneficial effect o f mothers' education on various dimensions of child health i s one o f the most robust empirical relationships ever established in the development literat~re.~''Inour rural sample, 92 percent of mothers have not completed primary schooling (see Table A.9.3, Appendix 3). This impliesthat even a modest increase in female education will have significant payoffs in improving child health, particularly in rural areas. Unfortunately, our empirical analysis o f primary school enrollment decisions also suggests that Ethiopian households have a strong bias against investments ingirls' education. 9.18 Access to clean water emerges as an extremely important determinant of child mortality, Children inrural households using uncovered ground water (spring or uncovered well) are two to three percentage points less likely to die before reaching the age o f five, 29'As indicated in our discussion o f child malnutrition, the pathways through which schooling leads to better child health outcomes are still more o f an empirical mystery. 189 compared to those in households using surface or rainfall water. In our rural sample 38 percent of households rely on surface or rainfall water, so this represents an area o f great opportunity for effective intervention. Access to piped water (including communal taps) brings down the risk of childhood mortality even further, by five percentage points. The effect o f safe water access i s most pronounced inurban areas, where access to surface water i s limited-children from urbanhouseholds with access to a coveredwater source are 13 percent less likely to die before reaching their fifth birthday than those relying on rain and surface water. 9.19 Somewhat surprisingly, we do not find a discemable effect on child mortality rates in rural areas, nor in all but one o f the urban specifications o f access to sanitation. This may be because of low variation in access to sanitation facilities, or because o f important threshold effects such that at least a minimum proportion o f the community needs to use sanitation facilitates to have an effect. In rural areas, 92 percent o f households have no sanitation facility and the remaining eight percent use a traditional pit toilet. Even in urban areas, no facility i s reported by 30 percent o f households, with most o f the rest (66 percent) using traditional pit latrines. The result i s especially surprising and alarming in urban areas, given that households often live quite dispersed in rural areas, reducing the likelihood o f contamination. 9.20 Electricity is associated with child mortality in both rural and urban areas. Children from rural households with access to electricity are 10 percent less likely to die before reaching the age o f five; children from urban households with electricity are 6.5 percent less likely to die before the age o f five. However, the effect disappears once we control for community effects, indicating that electricity may be a proxy for access to other community services. 9.21 Girls from rural areas are 2.2 percentage points less likely to die before reaching age five compared to boys. This finding i s consistent with the empirical literature from Africa which suggests strong gender biases in investments in education, but no gender bias in investments in health. The `boy bias' in child mortality, however, does not hold in urban areas, where there i s no difference inchild mortality by gender o f the child. 9.22 While there was a strong bi-variate relationship between under-five child mortality and wealth in the previous descriptive section, after controlling for a host o f other factors (many o f which are strongly correlated with wealth), the wealth effect on under -five child mortality rates is no longer significant. 9.23 Especially in rural areas, child survival would benefit from postponement of motherhood. A rural mother who i s five years older i s 10percent less likely to suffer the loss of her young child (the median age o f first birth i s 19 years in rural areas and 20 in urban areas).292 Mothers' nutritional status also affects under-five child mortality rates. We find that the mother's height, which i s proxy for nutritional status during childhood and adolescence, has a small but significant effect in all but one specification, indicating 292Note that while the sign o f mother's age reverses in the urban sample, its effect is overwhelmed by the squared term. 190 persistence in health outcomes across generations. However, other measures o f maternal nutritional status, including mother's weight and Body-Mass-Index, were not significant. Table 9.3: Determinants of under-five child mortalitv in rural and urban areas of EthioDia Rural Urban IndependentVariables Rural Rural Rural Urban Urban Urban (1) (2) (3) 4, (4) (5) (6) Child'sCharacteristics Female Child -0.022* -0.022* -0.022" 0.005 0.004 0.016 (2.08)* (2.05)* (1.9)" -0.26 -0.2 -0.71 Mother'sCharacteristics Mother'sAge -0.02 -0.02 -0.016 0.044 0.042 0.053 (2.65)** (2.64)** (1.96)" (2.89)** (2.76)** (2.81)** Mother'sAge Squared 0.00 0.00 0.00 -0.001 -0.001 -0.001 (2.55)* (2.54)* (1 .91)" (2.83)** (2.71)** (2.70)** Mother's Weight 0.00 0.00 0.00 0.00 0.00 0.00 -0.45 -0.42 -0.89 -0.87 -0.78 -1.85 Mother'sHeight 0.00 0.00 0.00 0.000 0.000 -0.001 (2.33)* (2.39)* -1.37 (2.63)** (2.39)* (3.70)** Mother'sEducation(Years) -0.016 -0.016 -0.012 -0.006 -0.008 -0.007 (2.87)** (2.79)** (1.83)" (2.16)* (2.71)** (2.08)* FemaleHeadedHouseholdDummy -0.013 -0.012 -0.007 0.008 0.016 0.022 -0.85 -0.79 -0.39 -0.36 -0.75 -0.81 Wealth Quintile' Quintile 2 -0.039 -0.034 -1.32 -0.9 Quintile 3 -0.014 -0.017 0.051 0.081 - 1.09 -1.19 -1.68 (2.08)* Quintile 4 0,001 -0.015 -0.015 -0.011 -0.03 -0.78 -0.51 -0.29 Quintile 5 -0.027 -0.006 -0.023 -0.028 -1.37 -0.29 -0.61 -0.6 Wealth Score Wealth Score 0.292 0.124 -1.24 -1.12 Wealth Score Squared -0.097 -0,101 -1.27 -0.92 InfrastructureAccess PipedWater -0.049 -0.049 -0.05 1 -0.027 -0.017 -0.05 (1.97)* -1.95 -1.33 -0.72 -0.47 -0.94 UncoveredGroundWater -0.028 -0.03 -0.025 -0.054 -0.047 0.025 (2.32)* (2.43)* -1.44 -1.26 -1.09 -0.32 CoveredWater -0.023 -0.024 -0.003 -0.132 -0.131 -0.099 -1.22 -1.28 -0.09 (3.10)** (3.04)** -1.34 No Sanitation 0.011 0.01 0.003 -0.63 -0.58 -0.14 Flush-Toilet -0.039 -0.045 -0.128 -0.54 -0.64 (2.02)* 191 Rural Urban IndependentVariables Rural Rural Rural Urban Urban Urban (1) (2) (3)41 (4) (5) (6) Pit-Toilet -0.014 -0.022 -0.034 -0.53 -0.84 -0.98 Electricity -0.099 -0.1 -0.101 -0.065 -0.082 -0.065 (1.82)" (1.82)" -1.2 (1.99)* (2.52)* -1.45 Religion & Region Orthodox 0.017 0.016 -0.014 0.061 0.057 0.091 -0.89 -0.85 -0.47 -1.39 -1.31 -1.67 Muslim 0.051 0.05 0.048 0.137 0.134 0.18 (2.53)* (2.53)* -1.2 (2.63)** (2.58)** (2.73)** Tigray 0.024 0.025 0.074 -0.043 -0.04 -0.194 -0.72 -0.76 -0.33 -1.07 -1 -1.32 Afar 0.04 0.034 0.247 -0.1 -0.IO3 -0.181 -1.31 -1.14 -0.8 -1.69 -1.76 (1.99)* Amhara -0.015 -0.013 0.477 -0.037 -0.03 -0.184 -0.5 1 -0.43 -1.36 -0.9 -0.71 -1.5 Oromiya 0.004 0.004 0.347 -0.013 0.001 -0.113 -0.14 -0.14 -1.23 -0.32 -0.02 -0.59 Benishangul -0.013 -0.012 0.137 0.081 0.075 -0.172 -0.42 -0.39 -0.42 -0.9 -0.83 -1.34 SNNP 0.034 0.034 0.354 -0.05 -0.033 -0.178 -1.11 -1.11 -0.92 -0.97 -0.6 -1.52 Gambela 0.08 0.082 0.36 0.103 0.122 -0.175 (1.98)* (2.02)* -1.09 -1.6 (1.85)" -1.38 Harari 0.066 0.066 0.063 0.017 0.022 -0.082 (1 3)" (1.77)" -0.27 -0.49 -0.61 -0.32 Dire Dawa -0.02 -0.021 0.173 0.006 0.008 -0.179 -0.46 -0.5 -0.52 -0.21 -0.27 -0.82 Observations 5876 5876 5501 1451 1451 1194 'I Coefficients are reported as marginal probabilities; dependent variable takes o n the value o f one if the child died before reaching the age o f five (0 otherwise). *) A significant at 10 percent; * significant at 5 percent; ** significant at 1 percent; robust (Huber-White) z- statistics inrows under coefficients. 3, Controlling for censoring (selection at least five years prior to the survey). 4, Specification (3) and (6) are runusing cluster dummies to proxy for infrastructure, prices, and environment. The cluster dummies are not significant inthe urban specification (6). The poorest and second poorest households are the two left-out quintiles in the rural sample; the poorest quintile i s the left-out quintile inthe urban sample; households using surface or rainfall water form the left- out water category; household without access to any sanitation are the left-out categories; Protestant and others are the left-out categories; Somali i s the left out region. 9.24 There are also large and statistically significant ethnic and regional effects on child mortality. Inparticular, being Muslimincreases the probability of premature death by five percentage points relative to being Protestant (the omitted category) in the rural sample and between a shocking 13 and 18 percent among urban dwellers. There are also large and statistically significant regional variations in under-five child mortality rates. Relative to 192 children living in rural Somali communities, children inrural Gambela have an eight percent higherprobability o f dyingbefore age five. 9.4 PolicyActions Neededto Reachthe ChildMortalityMDG 9.25 While Ethiopia has made considerable progress in reducing child mortality over the past decade, the rate o f decline i s far from sufficient to reach the MDG goals in the foreseeable future.293Unless the pace o fprogress inimproving health outcomes inEthiopia i s drastically accelerated, it i s highly unlikely that Ethiopia will be able to meet its child mortality MDG goal by 2015 (Figure 9.3). Figure 9.3: AchievingMDG goal for reductionin child mortality 250 1 Years -Achieve M G D (Decrease of 5.2 percent points per year) I -Current Trend (Decrease of 1.9 percent points per year) +Required trend (7.4 percent point per year from 2000) 9.26 Ethiopia's overall progress in achieving the health MDGs will be largely determined by its success in reducing child mortality in rural areas, given the concentration o f the population in rural areas as well as the disproportionate share of premature child deaths that take place there. We undertake three simple policy simulations that illustrate the challenges that lie ahead inachieving the envisioned reductions inunder-five child mortality inEthiopia. 9.27 First, we simulate the impact on under-five child mortality rates in Ethiopia of giving all mothersfour years of education. The average mother inour rural sample has less than 0.8 years o f schooling; increasing this by 3.2 years increases the average rural child's chance o f survival by 5.1 percentage points, bringing down the rural under-five mortality rate from 192 to 141 deaths per 1,000 live births. In urban areas, with higher current educational attainment (1.3 years) and a lower marginal effect o f mother's education on child mortality, the decrease is a more modest 1.6 percentage points. 293World Bank 2004b. 193 9.28 Second, we simulate the impact of providing access to safe drinking water (piped water) to those households which are currently using surface or rainfall water.294This would reduce child mortality in rural areas by 1.7 percentage points, but in the already well- serviced urban areas by only 0.9 percentage points. Currently, 35 and seven percent o f all rural and urban households respectively use surface or rainfall water. 9.29 Together these two interventionswould reduce child mortality by 6.8 percentage points in rural areas (or 68 deaths less per 1,000 live births) and 2.5 percentagepoints (or 25 deaths less per 1,000 live births) in urban areas. From Figure9.4, it i s clear that any hope for achieving the child mortality MDG inEthiopia lies in a focus on rural areas. This i s where child mortality is highest, and also where the greatest gains from expanding education and infrastructure can be realized. These results also illustrate the considerable scope for making progress toward health targets through interventions in other areas, particularly in women's education. Figure 9.4: Effect of selectedinterventionson rural under-five childhoodmortality z e 120 8 0 100 w 8 80 a 2 60 5g '2 40 a 20 g o safe water alone education alone combined needed to achieve interventions MDG 9.30 The simulations presented so far do not consider direct interventions in health such as immunizations, better treatment o f diarrhea, a major cause o f child mortality, and improved access to medical care, which also prove to have much potential to reduce child mortality. The ongoing expansion o f access to key professional preventive services delivered by two female health extension workers in every kebele, trained in one year, through the outreach component o f the health extension package, i s to be the backbone o f the revised health system o f Ethiopia. Increasing access to this basic extension package (which includes immunization and Vitamin A supplementation) to 65 percent o f the population i s estimated to buy a reduction o f twelve percent in under-five mortality. It would cost an additional US$ 0.59 per capita. 294For the rural households we simulate the effect o f providing piped water; for the urban, covered. 194 9.5 ConcludingRemarks 9.31 The spread of HIV/AIDS (and malaria) threaten Ethiopia's prospects for economic growth and poverty alleviation. Over 50 percent o f child deaths in Ethiopia are caused by pneumonia and diarrhea alone, ailments that are preventable at a fairly low cost. While there has been positive progress toward the health MDGs in Ethiopia, the challenges ahead are substantial and will require cross-sectoral interventions. Due to the legacy o f both malnutrition and poor health access among the poor, a concerted effort on the part o f government and its development partners will be necessary if Ethiopia i s to achieve the reduction in under-five child mortality envisioned by the Millennium Development Goals. This effort would be most efficiently targeted at the rural areas. Interventions to expand access to safe water and to increase mother's education hold especially strong promise, as does the new health extension package mentioned above. 195 CHAPTER 10. TOWARD AN EDUCATED PEOPLE 10.1 Literacy, Developmentand OngoingPolicyChallenges in Ethiopia 10.1 As is the case with improving people's health, pursuing education is not only instrumentally valuable, but it is also important in its own right. Like being well- nourished and healthy, being able to read and write i s one o f our primary capabilities as human beings. Education enables people to acquire information and convert it into knowledge, both critical components to make effective choices and translate them into desired actions. As illustrated throughout this document, education also has tremendous instrumental value. The large beneficial effects o f education on economic growth and consumption poverty in Ethiopia have been highlighted in Chapter 4. Its critical role in facilitating the adoption o f productivity-enhancing farming techniques and technologies,29s which in turn leads to improved food security and agricultural development, has been discussed in Chapter 5. The importance o f parental education in reaching many human capabilities cannot be sufficiently underscored, as illustrated with respect to child malnutrition and child mortality in the previous chapters o f Part 111. Inparticular, the pervasive effects o f female adult education on human development outcomes (including education itself) has beenone o f the most robust empirical relations estimated inthe empirical human development literature. 10.2 Since 1994 Ethiopia has experienced dramatic increases in primary school enrollments. Gross primary school enrollment rates havemore than doubled from 24 percent in 1994 to 57 percent in2000 for grades 1-8, and net enrollment rates have gone up from a mere 17.8 percent in 1994-5 to 48.8 percent in 2000. The boom in primary school enrollments has been largely attributed to growth in the number o f grade 1 entrants. During the period 1994-1996, the number o fnew entrants exceeded those inthe previous year by 20, 26, and 15 percent respectively, or by about 200,000 children in absolute terms.296 Since 1996-97, the number of new grade 1 entrants has grown at a slower pace o fjust four percent each year. 10.3 Ethiopia's phenomenal success in expanding primary school enrollment since 1994 had its foundations in the government's New Education Policy, which was formulated and adopted in 1994. The major reforms included significant investments in government schools at the primary and secondary levels-which enabled the education sector to accommodate rapid growth in the number o f children entering grade 1-as well as other policy reforms that sought to remove demand-side and institutional constraints to increasing educational attainment, especially among the poor. Key supply side reforms included: (1) the construction o f new public schools; (2) an increase inthe supply o f private schools; and (3) an increase in the number o f double-shift schools which essentially enabled the government to boost enrollments using its existing capital stock in the education sector. Inparticular, large public investments in government schools under the government's 1994 Education and Training Policy and Strategy program has been cited as the engine behind the massive increase in primary school enrollment between 1994 and 2001.297 To foster the demand for ~~ ~~ 295Weir, 1999;Knight et al, 2003. 296World Bank, 2004a. 297World Bank 2004a. 197 schooling and remove barriers to school entry, the government's New Education Policy abolished school fees in govemment schools and also implemented food-aid programs in primary schools. 10.4 Other institutional reforms that help to explain the success of the government's efforts at boosting primary school enrollment in Ethiopia include a rigorous program o f education sector decentralization that allowed for the introduction and use o f native languages in the classroom, and greater community involvement in the construction o f schools. Inaddition, the government has made efforts to increase the number o f female teachers (which has been shown to increase female enrollments inprimary and secondary schools) and the provision of school suppliesthat are important for learning, such as textbooks. 10.5 Inefficiencies in spending levels remain and are attributed to the persistence of educational disparities and apparent deterioration in quality of schooling. Inparticular, the primary school share of recurrent spending has decreased in favor o f tertiary education, despite a dramatic increase inprimary school enrollments implying an increasing demand for recurrent finance. The share of recurrent spending going toward the primary school sector in Ethiopia (50 percent) i s far below the international benchmark o f 67 percent envisaged under the Education For All Fast Track Initiative.298 In addition, a large share o f the primary education budget i s consumed by teachers' salaries, leaving minimal amounts available for non-wage recurrent spending on supplies, textbooks and teaching materials which have direct impacts on the quality o f educational services provided. Non-salary spendingper student has beendramatically eroded over time from an average o f around 7 ETB per student in 1994 to around 4 ETB per student in 2001. In addition, since 1994-95 pupil-teacher ratios have trendedupward, averageing 70 children per classroom for grades 1-8, and 80 at the secondary school level, by 2001 (Figure 10.1). 298World Bank, 2004a. This target i s based on the pattern o f spending in countries that have made good progress toward universalizing primary school completion as outlined in Bruns, Mingat and Rakotomalala (2003). 198 Figure 10.1: Trends in pupil-teacher and pupil-section ratios and in non-salary public spending per primary and secondary student, Ethiopia, 1990-2001 _.- A 80 8 h I Ei u .-An 50 P n 5 40 0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 Note: The PTR is computed by dividing total enrollments by the total number of teachers, and the PSR by dividing total enrollmentsby the totalnumber of sections. These calculationsyield weightedratioswhich differ from the unweightedratios computed by averaging the ratios across schools. Non- salary public spending per student is computed by dividing the reportedbudget amount for this itemby the total number of primaryand secondary students in govemment schools. Source: World Bank 2004a. Based on data supplied by the Ministry of Education on enrolments, teachers and sections; and on MoFED budget data on expenditure 10.6 The education MDGs in Ethiopia call for achieving universal primary school completion and enrollment by 2015 and ensuring gender parity in primary and secondary education by 2005. Increases in primary school enrollment rates require expansions in the coverage o f the education system, while higher primary school completion rates require improvements in student flows through the education system. To further inform the appropriate mix o f policy interventions needed to achieve the education MDGs in Ethiopia, we jointly examine the relative importance o f both supply side (access to schools and the quality of schooling) and demand side (income, parental education, and gender) determinants of primary enrollment rates and fifth grade completion rates. We begin by exploring the nature o f education outcomes in Ethiopia and the magnitude o f the disparities across socio- economic groups. 10.2 EducationalProfileinEthiopia 10.7 Since 1994, Ethiopia has made significant progress toward achieving the educationMDGs, but significant improvements remain possible. In2000 gross primary school enrollment rates for the first cycle o f primary school (grades 1-4) were 83 percent, and were 57 percent for grades 1-8. However, net primary school enrollment rates still lag at 49 percent (Tables 10.1 and 10.2). This difference between gross and net primary enrollment 199 rates reflects the big backlog in educational achievement Ethiopia had built up and which i s now beingclosed as older children have also been enrolling in Grades 1-4 Grades 5-8 Grades 1-8 Boys Girls Total Boys Girls Total Boys Girls Total 1994-5 37.7 22.6 30.3 17 13.9 15.5 28.9 19 24.1 1995-6 58.3 33.2 46 19.3 13.8 16.6 39.8 24 32 2000-1 95.3 70.2 83 38.3 22.9 30.8 67.3 47 57.4 Table 10.2: Net primary enrollmentin Ethiopia Grades 1-8 Boys Girls Total 1994-5 II 20.7 14.7 17.8 1995-6 28 18.6 23.4 2000-1 55.7 41.7 48.8 10.8 Figure 10.2 shows both cross-sectional and cohort completion rates in Ethiopia. While there has been an increase in the cohort completion rate at all grade levels since 1995- 96, the overall completion rate remains extremely low. Presently only 60 percent o f each cohort enters grade 1, of which one-quarter drop out by grade 2 and one-half drop out by grade 5. Low completion rates may reflect the low quality of educational services inEthiopia (likely due to lack o fpublic funds), or low returns to education inthe labor market. Figure 10.2: Grade-specificcompletion rates, Ethiopia, 1995-96,2000, and 2001/2 100 1 *.* . 1 Cross-sectional completion rates, 2000 80 -* ./ l 20 I 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 Grade attained Source: WorldBank 2004a 299Gross primary enrollment i s defined as the number o f children enrolled inprimary school, regardless o f age, divided by the population of the age group that officially corresponds to the primary level; net enrollment i s the ratio o f the number of children o f official school age who are enrolled inprimary school to the total population of childreno f official school age. 200 10.9 While we do not see stark disparities in educational attainment across wealth quintileson average in Ethiopia,there is a wealth-gendernexus in educationaloutcomes which consists o f larger disparities in educational outcomes between girls across different wealth groups. In addition there are large disparities in educational outcomes between girls and boys within the same wealth quintile, which implies that households tend to under-invest in girls' education in Ethiopia. We find a gender-biased pattern in gross primary school enrollment rates for both 1996 and 2000 (Figure 10.3 and Figure 10.4). The wealth gradient for boys in primary school enrollment appears to have flattened between 1996 and 2000, while the wealth-enrollment gradient for girls has persisted over time. The gender gap between male and female primary school enrollment rates within the upper quintiles decreased between 1996 and 2000. Figure 10.3: Gross primary enrollment ratesby gender and wealth quintile, 1996 B 15% 0% Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Wealth quintile 201 Figure10.4: Gross primarv enrollment ratesbv gender and wealth quintile, 2000 25% 1 ... . 20% -2 P Q) W -% e 5 15% 3t e a" 8 10% 5% Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Wealth quintile 10.10 Wealth appears to be a stronger correlate o f primary school completion than primary school enrollment in Ethiopia. This i s particularly true for girls, as shown inFigures 10.5 and 10.6 below. Although there have been large increases in enrollments across all quintiles in Ethiopia between 1996 and 2000, overall primary school completion rates remain extremely low, less than five percent even for boys inwealthier families. Figure 10.5: Primary completionrates by gender and wealth quintile, 1996 - - __ .___- ____ - __ - - - 2 2 1 5 % .-e 0 C -E Q) 6 1 0% s .K $ .- 05% l t 0 0% Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Wealth quintile 202 Figure 10.6: Primarycompletion ratesby gender andwealthquintile,2000 4.5%,- I Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Wealth quintile 10.3 Determinantsof PrimarySchoolinginEthiopia 10.11 Schooling outcomes are shaped through a complex intersectionof factors, which often makes it extremely difficult to pinpoint causalpathways. From the supply side, the delivery o f schooling i s plagued by a plethora o f problems affecting the quantity and quality o f services.300 These include the possibilities that not enough money i s allocated to schools;301that allocated funds are siphoned off thereby further reducing the actual budget available at the school and that the money actually allocated to providing schooling i s not spent effectively (e.g. poor teacher quality). From the demand side, factors related to household income, alternate uses o f the child's time, gender preferences within the household, social norms, and market signals, all influence the household's decision o f whether or not to invest in the child's schooling. In the Ethiopian context, adverse weather shocks and credit and insurance market failures exacerbate the vulnerability o f poor households, with often deleterious consequences on human capital investments. Furthermore, factors outside o f the education sector have important consequences on educational outcomes.303 For example, Glewwe and Jacoby (1995) highlight the effect of childhood malnutrition on primary school enrollment in Ghana, while Miguel and Kremer (2005) find that in one district in Kenya the most cost effective intervention to increase primary school enrollment was a deworming program. 300World Bank, 2004h. Devarajan, Miller and Swanson, 2002. 302Reinikka and Svensson, 2001. '03Particularly the relationship between health and education. 203 10.12 We modeled the probability of primary school enrollment and completion in Ethiopia using data for the years 1996 and 2000.304While we cannot do justice to all of the complex relationships mentioned above, we present the most comprehensive analysis to date on several important determinants o f child schooling in Ethiopia. In particular, we highlight: (a) household factors which shape child schooling outcomes; (b) school quantity and quality factors; and (c) the impact o f adverse shocks on schooling investments. Inthis endeavor, we merge three major types o f data: (1) Welfare Monitoring Surveys (WMS) from the Central Bureau o f Statistics, from which we access household data (income, child schooling); (2) school census data from the Education Management Information System (EMIS), Ministry of Education, from which we obtain information on schools; and (3) various data sources on adverse shocks (e.g. household plot-level crop damage). We estimate separate specifications by rural and urban samples. 10.3.1 Householdand communitycharacteristics 10.13 There i s a strong bias against investmentin female education. The magnitude o f the gender bias i s undoubtedly one o f the striking findings of the analysis. Controlling for all other supply and demand side factors such as distance to and quality of schools, household resources, and parental education, girls who reside in rural areas are 11.6 percentage points less likely to be enrolled in school compared to boys (see Table A4.4 in the appendix). To better appreciate the astounding magnitude o f this effect, note that this impliesthat almost one million people of primary school age are denied schooling merelybecause o f their gender, i.e. irrespective o f the income o f the household, the parental educational status, the distance to the school or the quality o f the schooling.305 Even in urban areas, girls are 4.3 percentage points less likely to be enrolled in school than boys. Given that most common supply and demand determinants have been controlled for, the reasons for this strong bias against female education must partially relate to cultural factors, and perpetuate the disempowered position of female citizens observed and discussed in Chapter 1. To further explore the reasons for this gender bias inschool enrollment, we separately examine the effects o f the various known determinants of enrollment and completion separately for girls and boys (see Tables 10.3 and 10.4; detailed regression results are presented inTables A.10.1 - A.10.4 inAppendix 3). 304Non-pooled regressionresults are quite similar (separate 1996and2000 results are not reported). 305According to the DHS 2000, there are about 28.65 million females residing inrural areas of which 28 percent are between five and 14 years old. Multiplying the total number of school age people by their reduced percentage likelihoodof being enrolled yields 930,000. 204 Table 10.3: Determinantsof primary school enrollment in Ethiopia by gender and location'' Rural Urban Variable Girls Boys Girls Boys Statistical Coef- Statistical Coef- Statistical Coef- Statistical Coef- Significance ficient Significance ficient Significance ficient Significance ficient Individual and household characteristics Child age 8 Yes (1%) 0.114 Yes (1%) 0.103 Yes (1%) 0.07 Yes (1%) 0.08 Child age 9 Yes (1%) 0.214 Yes (1%) 0.24 Yes (1%) 0.115 Yes (1%) 0.116 Child age I O Yes (1%) 0.282 Yes (1%) 0.323 Yes (1%) 0.155 Yes (1%) 0.122 Child age 1I Yes (1%) 0.249 Yes (1%) 0.352 Yes (1%) 0.15 Yes (1%) 0.1I Child age 12 Yes (1%) 0.258 Yes (1%) 0.38 Yes (1%) 0.148 Yes (1%) 0.13 Child age 13 Yes (1%) 0.252 Yes (1%) 0.408 Yes (1%) 0.103 Yes (1%) 0.121 Child age 14 Yes (1%) 0.294 Yes (1%) 0.391 Yes (1%) 0.113 Yes (1%) 0.118 Female head of household No 0 No 0 No 0 No 0 Age of head of household No 0 Yes (10%) -0.001 Yes (5%) 0.002 No 0 Single head of household No 0 Yes (5%) -0.087 No 0 No 0 Schoolingof household head Yes (1%) 0.016 No 0 Yes (1%) 0.008 Yes (1%) 0.008 Other adult schooling Yes (1%) 0.012 No 0 No 0 Yes (1%) 0.007 Log of per capitahousehold Yes (1%) 0.051 Yes (1%) 0.065 No 0 Yes (1%) 0.041 expenditures Raindamage Yes (1%) -0.053 No 0 No 0 No 0 EA- Average male adult literacy Yes (10%) 0.055 Yes (1%) 0.17 No 0 Yes (1%) 0.206 rate EA- Average female adult literacy Yes (1 %) 0.454 Yes (I%) 0.374 Yes (1 %) 0.24 Yes (1 %) 0.I49 rate Community characteristicsand supply side factors DistancePrimary School 1-2 km Yes (1%) -0.037 Yes (10%) -0.033 No 0 No 0 DistancePrimary School 3-4 km Yes (1%) -0.067 Yes (5%) -0.046 No 0 No 0 DistancePrimary School 5-6 km Yes (1%) -0.104 Yes (1%) -0.093 No 0 No 0 DistancePrimary School 7-12 km Yes (1%) -0.122 Yes (1%) -0.177 No 0 No 0 DistancePrimary School > 13 km Yes (1%) -0.16 Yes (1%) -0.179 Distanceto secondary school No 0 No 0 No 0 No 0 Distance to food market Yes (5%) 0.003 No 0 No 0 No 0 Distanceto health clinic Yes (1%) -0.002 No 0 No 0 Yes (5%) -0.007 Distanceto post office Yes (10%) -0.001 No 0 No 0 No 0 Number of schoolsper capita Yes (1%) 0.363 Yes (1%) 0.706 Yes (5%) 0.179 No 0 Population density Yes (5%) 0 No No 0 Yes (1%) 0 Student-teacherratio Yes (1%) 0.001 Yes (1%) 0.001 Yes (1%) -0.001 Yes (1%) -0.002 Percent female teachers No 0 No 0 Yes (1%) 0.005 Yes (1%) 0.004 Percent teachers with certification No 0 Yes (10%) 0.106 No 0 No 0 Tigray Yes (10%) 0.09 No 0 Yes (10%) 0.132 No 0 Amhara Yes (5%) 0.105 Yes (10%) 0.118 Yes (5%) 0.166 No 0 Oromiya No 0 Yes (1%) 0.176 No 0 No 0 Benishangul Yes (10%) -0.065 Yes (1%) 0.212 No 0 No 0 SNPR No 0 Yes (1%) 0.189 Yes (10%) 0.149 No 0 Harari No 0 Yes (1%) 0.272 No 0 No 0 Year 2000 dummy Yes (1%) 0.14 Yes (1%) 0.157 Yes (1%) 0.103 Yes (1%) 0.069 ') All regression coefficients, when statistically significant at a level < 10 percent are reported as marginal probabilities.; 2, EA= enumeration area, which corresponds to a community. 205 10.14 Households with better educated adults and those in better educated communities are more likely to have children enrolled in primary school. For every additional year o f educational attainment o f the household head, the probability that a child i s enrolled in school increases by 1.1 percentage points inrural areas. The positive effect o f the household's education i s especially strong in determining girls' initial enrollment, where each additional year o f education increases the probability o f a girl's enrollment by 1.6 percentage points. Inurban areas, the effect i s estimated at 0.8 percentage points for both girls and boys. There are also strong positive externality effects o f educational attainment at the community level, particularly associated with female literacy. These are most pronounced in the rural areas, where a 10 percentage point increase in a community's female literacy rate results in a 4.5 percentage point increase in the probability o f enrollment for girls, and a 3.7 percentage point increase inthe probability o f enrollment for boys. The effects are about half as large in urban areas. Interestingly, adult male literacy rate o f the community seems to especially benefit the enrollment probability of boys. These results may indicate that the individual demand for education i s partially shaped by community preferences and social norms. Alternatively, or in addition, community educational attainments could be proxying for average community wealth. 10.15 Household income has a modest positive impact on the likelihood of primary school enrollment. The effect is slightly stronger in rural areas and for the enrollment o f boys. A ten percent increase in household income results in approximately a 0.5 percentage point increase inthe probability that a rural child is enrolled in school. 10.3.2 Risk and shocks 10.16 Idiosyncratic and covariant shocks affect primary school enrollment, especially among girls. Households with half o f their plot area damaged are 2 percentage points less likely to send their child o f school-going age to school. It further appears that harvest failure has asymmetric effects for male and female childreno f affected households. Inthe event that a household i s faced with an adverse production shock, especially females are less likely to attend primary school, and also less likely to complete primary schooling. This suggests that when vulnerable households are forced to make tradeoffs in educational investments o f their children they choose to protect their schooling investments in their sons rather than in their daughters. Strengthening farmers' risk management instruments will not only reduce their poverty and improve the nutritional status o f their children, it will also enhance the educational achievements o f their children, especially their daughters. 10.3.3 School characteristics 10.17 The distance of the nearest school from the homestead negatively impacts enrollment and completionprobabilities,especially in rural areas. The regression results suggest that households are reluctant to send their children to schools far from home. This effect may capture the opportunity cost o f primary school attendance, which will increase as the child's distance to school increases. Given that schoolchildren must generally walk to school, distance may also serve as a direct barrier to attendingprimary school among children living on remote farms, particularly young girls. To illustrate, households seven to 12 206 kilometers away from a school are 12 percentage points less likely to send their daughters to primary school, and 18 percentage points less likely to send their sons. If a school i s more than 13 kilometers away, children are 17.6 percentage points less likely to be enrolled in school, These findings suggest that supply side reforms in the education sector (e.g. construction o f new schools and classrooms in unserved or underserved areas) will continue to be a critical factor in improving primary school enrollment and completion rates in Ethiopia. 10.18 In general, student teacher ratios andthe proportionof female teacherspositively and substantiallyaffect enrollment and completion rates, especially in urban areas. To explore the effects of the quality o f schooling we look at the effects o f the student-teacher ratios, the percentage o f female teachers and the percentage o f teachers with a certificate. Overcrowding in the classroom decreases the chances o f enrollment in urban areas and the effects can be substantial. A reduction in the student-teacher ratio in urban areas from the current average o f 60 to 50 i s estimated to increase enrollment by two percentage points. Surprisingly, however, we find a positive relationship between the student teacher ratio and enrollment in rural areas. This may reflect the fact that the system i s still catching up with large increases in enrollment reflected in large student teacher ratios. Inboth urban and rural areas primary school completion rates (fifth grade) are negatively associated with student teacher ratios, though the effects are estimated with great imprecision. 10.19 The larger the percentage o f female teachers, the larger the probability that urban children will be enrolled in school. We do not find a relationship betweenthe gender o f the teacher and enrollment inrural areas. Interestingly,the presence o f female teachers has only a slightly larger positive effect on the enrollment o f girls relative to boys. Overall the effects are substantial: raising the percentage o f female teachers from its current average level o f almost 30 percent to 50 percent would increase enrollment rates in urban areas by 10 percentage points for girls and eight percentage points for boys. Moreover, while the presence o f female teachers positively affects the completion rates o f boys in urban areas, it negatively affects their completion rates in rural areas. No effects on the completion rates o f girls were discerned. These gender dynamics deserve fbrther exploration. No clear pattern was detected regarding the effect o f teacher certification. 10.20 Finally, the quality of school infrastructuregreatly increases the probability of male enrollment.306Availability o f water and toilets only affect boys' enrollment. Boys are 15 percentage points more likely to be enrolled if the closest school has drinking water and they are seven percentage points more likely to be enrolled if the closest school has a toilet. This not only highlights the importance o f the school environment, but also further points to the fact that households are less likely to send their daughters to school, whatever the school environment. ~ 306Results are based on 2000 only, for which more school infrastructure information was available. Regression results are not reported here. 207 10.3.4 Determinantsof fifth-gradecompletionin Ethiopia 10.21 Many of the demand and supply side determinants of enrollment have similar effects on primary school completionrates, though overallthe resultsare less significant, potentially related to the low completion rates observed in Ethiopia and thus the lack o f variation in our sample. Again, there exists a strong gender bias. Girls are eight to 10 percentage points less likely to complete fifth grade. Thus not only are girls less likely to be enrolled, when enrolled they are also less likely to complete primary schooling, controlling for a whole series o f other demand and supply side factors. Interestingly, schooling o f the household head only appears to affect completion rates in urban areas. However, there i s a strong positive education externality o f adult female literacy on the probability o f primary school completion for girls. Similarly, in the urban sample, adult male literacy has strong positive externalities inincreasing the probability o f primary school completion among boys, but not for girls. We find a perverse income effect inour urbansample on the probability o f primary school completion among girls. Distance to school and the number o f schools also have gender-specific impacts in the urban sample. Girls are less likely to complete primary schooling if they live far from a school, while boys are more likely to complete primary schooling incommunities with a greater availability o f schools. 208 Table 10.4: Primary school completionregression-rural and urban Ethiopia" Rural Urban Variable Girls Boys Girls Boys Statistical Statistical Statistical Statistical Significance Effect Significance Effect Effect ____- __ Effect Significance Significance . ...... .. .. __ Child age 13 No 0 No 0 Yes (1%) 0.147 Yes(]%) 0 252 Child age 14 N O 0 No 0 Yes (1%) 0.287 Yes (1%) 0 355 Femalehead of household No 0 No 0 No 0 No 0 Age of head of household No 0 No 0 Yes (1%) 0.003 Yes (5%) 0 002 Single headof household No 0 No 0 N O 0 No 0 Schoolingof householdhead No 0 No 0 No 0 Yes (1%) 0 009 Other adult schooling No 0 No 0 No 0 N O 0 Log of per capita household expenditures No 0 No 0 Yes (5%) -0.037 No 0 Rain damage No 0 No 0 No 0 N O 0 EA- Average male adult literacy rate No 0 No 0 No 0 Yes (5%) 0 203 EA- Average female adult literacy rate Yes (10%) 0.126 N O 0 No 0 No 0 DistancePrimary School 1-2km No 0 No 0 Yes (5%) -0.048 N O 0 Distance Primary School 3-4 km N O 0 No 0 Yes(lO%) -0.078 No 0 DistancePrimary School5-6 km No 0 No 0 No 0 No 0 Distance Primary School 7-12 km No 0 No 0 Distance Primary School > 13 km No 0 No 0 No 0 Distance to secondary school Yes (5%) -0.001 N O 0 No 0 No 0 Distance to food market Yes (5%) 0.003 No 0 No 0 N O 0 Distanceto healthclinic No 0 N O 0 No 0 No 0 Distance to postoffice No 0 No 0 No 0 No 0 Number of schools per capita No 0 Yes(l%) 0.199 No 0 Yes (5%) 0 28 Populationdensity No 0 No 0 N O 0 No 0 Student-teacherratio No 0 No 0 No 0 No 0 Percent female teachers No Yes (1%) -0.003 No 0 Yes(lO%) 0 003 Percentteachers with certification No 0 0 No 0 No 0 Tigray No 0 0 N O 0 No 0 Amhara No 0 0 N O 0 Yes (5%) 0.312 Oromiya No 0 0 No 0 Yes (10%) 0 26 Benishangul N O 0 0 Yes (10%) 0.246 No 0 SNPR N O 0 0 No No 0 Harari No 0 Yes(lO%) 0.174 Yes (10%) 0.287 N O 0 Year 2000 dummy Yes (1%) 0.104 Yes (1%) 0.115 Yes (1%) 0.084 No 0 Number of Observations 2014 2290 2927 273 1 I)Coefficients arereportedasmarginalprobabilities. 10.22 In conclusion, specific policy interventions to increase rural enrollment and completion rates should focus on: (1) increasing the availability o f schools; (2) strengthening the set o f risk management tools available to farmers (use o f pesticides and irrigation as well as ex post coping mechanisms), which would especially help in raising girls' education; and (3) improving the infrastructure in schools interms o f sanitation and water availability. 209 10.23 Actions to increaseurban enrollmentsshould focus more on improving the quality of schooling by: (1) reducing student-teacher ratios; (2) increasing the deployment o f female teachers; and (3) improving the school infrastructure in terms o f water and sanitation facilities. 10.24 Overall economic growth will also positively affect enrollment rates both in rural and urban areas. The potential impact on enrollment rates o f increasing awareness o f the importance of schooling, combined with adult literacy campaigns, deserves further investigation. The large externality effects o f overall educational attainments in the community and o f female adult literacy in particular, suggest promise in such an approach, especially if successfully targeted toward women. Finally, a more detailed ethnographic investigationo f why households are less likely to invest ingirls' education i s called for. 10.4 PolicyActionsNeededto Attain the EducationMDG 10.25 It is unlikely that Ethiopia will achieve universal primary completion by 2015 givencurrent educationoutcomes,which include a 60 percent cohort entry rate for Grade 1, a 25 percent attrition rate among entrants by Grade 2, and a 50 percent attrition rate by Grade 5.307 To further inform the relative effectiveness o f different interventions in increasing the gross enrollment rate, we perform a series o f simulations. Inparticular, we explore: (1) how overall economic growth would affect enrollment rates inboth rural and urban areas; (2) how much reducing the distance to schools and the occurrence o f harvest failure would raise enrollment rates in rural areas; and (3) how much increasing the quality o f schooling would increase enrollment rates in urban areas.3o8 Note that for a complete analysis o f the cost effectiveness o f these interventions the results must be complemented with cost estimates of the different interventions. 10.26 First, we simulate the impact of 11years of 3.3 percentper capita incomegrowth, which corresponds to the requiredGDP per capita growth rate to reach the poverty MDG (see Table 6.1, Chapter 6). This corresponds to an increase inhousehold income of 43 percent and would increase the enrollment rate in rural and urban areas by just two and one percentage points respectively. Clearly, income growth alone will not be sufficient to reach the enrollment MDG. 10.27 Second, we simulate the impact of bringing all children in rural areas within two km from a primary school, which currently is not the case for 54 percent o f all rural children. Targeted construction o f schools to bring all children within this distance would increase gross enrollment by 4.9 percentage points. 10.28 Third, we simulate the impact of a ten percentage point reduction in the crop area damaged. Weather and pest shocks are a constant source o f vulnerability and income loss in rural Ethiopia. A 10 percentage point reduction in the crop area damaged, which corresponds to a 50 percent reduction in crop damage from their 1995 and 1999 levels, both moderately good years, would increase rural enrollment rates by one-half percentage point. '07WorldBank 2004a. 308 For our simulations,we use the marginaleffects reported inTable A.10.3, Appendix 3. 210 Given that girls suffer disproportionately from crop damage, this would also help reduce the gender gap ineducational attainments. 10.29 Fourth, we examine the effect of increasing the quality of schooling in urban areas. Reducing the student teacher ratio from 60 to 50 would increase urban enrollment rates by two percentage points, and increasing the proportion o f female teachers from its current level o f about 30 percent to 40 percent would increase enrollment in urban areas by five percentage points. These effects are clearly substantial and would bring Ethiopia a good deal closer to universal primary enrollment. 10.30 While reaching universal primary education by 2015 lies beyond Ethiopia's reach, concertedand comprehensiveeffort will also be neededto reachthe more modest intermediategoal of 100 percent gross enrollment, a pre-condition for reaching universal primary education. The simulations suggest that the combined effect o f overall economic growth at a pace o f 3.3 percent per capita over the coming decade, increased access to schools, reduced exposure to harvest failure in rural areas, and improved quality o f schooling in urban areas would increase gross enrollment rates by 8.7 percentage points in rural areas and by 10.3 percentage points in urban areas. While this would bring Ethiopia close to reaching 100 percent gross enrollment in grades 1-4, the simulations suggest that in the absence o f additional interventions, Ethiopia would still remain far removed from universal gross enrollment ingrade 1-8. 10.5 ConcludingRemarks 10.31 Our empirical findings suggest that supply side interventions remain critical to increasing primary school enrollment and completion rates in Ethiopia. It further appears that in rural areas supply side interventions should be mainly focused on increasing accessibility to schooling, especially through the construction o f new schools inunserved and underserved areas. Remoteness from a school increases the opportunity cost o f primary school attendance, and given that school children must walk to school, distance may also serve as a direct barrier to attending primary school, especially for young girls. Supply side interventions in urban areas, where most children are already within two kilometers from a primary school, should focus on increasing the quality o f schooling, as captured by the student-teacher ratio. The presence o f female teachers may also have a substantial effect on school attendance and completion, especially by girls. 10.32 On the demand side, there is a need for social protection programs, which could help mitigate the negative impact o f idiosyncratic shocks on school enrollment and completion, again especially for girls. There may also be opportunities to expand and improve risk mitigation practices on smallholder farms (e.g. small scale irrigation) and to introduce insurance programs (e.g., crop insurance schemes) which would help buffer household food consumption and income from idiosyncratic production shocks. Given the strong association between parental (and especially female adult) education and the educational achievements o f the community, the role o f awareness and adult literacy campaigns should be further explored. A more detailed examination o f why households are less likely to invest in girls' education, particularly during times o f adverse shocks, would help to inform the design o f social risk management practices to protect female enrollment 211 during shocks, and to close the gender disparity in educational outcomes across wealth quintiles inEthiopia. Differential returns to education among men and women may influence household investment decisions. Empirical analyses aimed at understanding labor market segmentation and wage determination by gender could thus also improve our understanding o f the underlyingcauses o f household underinvestment in female education. Alternatively, or in addition, there may be a deep-rooted cultural bias as suggested by the disempowered position o f women in Ethiopian society generally. Measures to strengthen women's legal rights and expand their economic opportunities, as suggested in Chapter 1, may help to improve the standing o f women and to gradually erode this bias. Finally, action research should also be pursued whereby parents are encouraged to send their daughters to school for example through attendance fees or through greater involvement o f the communities. 212 CHAPTER 11. STRENGTHENINGPEOPLE'SAGENCY-CONCLUDING REMARKS 11.1 Part I11of the study explored the determinants o f non-monetary dimensions o f well- being and focused in particular on three key human development outcomes (malnutrition, health and mortality, and education). Improving human development outcomes i s indeed important in and o f itself to enhance people's well-being, and it i s also critical both to empower people-allowing them to make effective choices in life-and to engender economic growth and reduce consumptionpoverty. Not only will improvements innutritional status, health and educational outcomes among the poor unlock their human potential and stimulate economic growth, these improvements will also help break the vicious poverty cycles which keep individuals and families in conditions o f chronic poverty and economic destitution across generations. We review the key policy implications regarding appropriate strategies to enhance human development outcomes emerging from the analysis, and comment briefly on the need for other interventions, in particular the need to enhance access to information and to strengthen people's opportunity structure, necessary to further foster empowerment o f citizens ingeneral and women inparticular. 11.2 The spread of HIVIAIDS, but also continued high malaria incidence, must be reversedif Ethiopia's aspirationsfor economic growth and poverty alleviation are to be met. Ethiopia's HIVIAIDS epidemic is generalized, having spread far beyond the original high-risk subpopulations. Prevalence among women attending ante-natal clinics i s five percent or more, and approximately 1.5 million Ethiopians are living with the virus. Unabated progression o f the epidemic will undermine current and future development efforts. Malaria is one o f the leading causes o f both hospital stays and outpatient visits in Ethiopia, representing a huge burden on the country's strained health system. HIV/AIDS and malaria account for 6.2 and 4.5 percent of child deaths respectively. Controlling malaria will facilitate access to cultivable land, thus alleviating the intense land pressure that characterizes Ethiopian agriculture and providing opportunities to the rapidly emerging landless class. Immediate and concerted efforts to halt the spread o f both diseases and to provide effective treatment for those infected are a necessary condition for other development and poverty reduction interventions to bear any h i t . In doing so, it will be critical to rapidly build an accurate data base (for example through a National Biological Survey) to enable close monitoring o f the evolution of the incidence o f HIViAIDS incidence as well as a flexible response to the changing manifestation o f the disease. Successful implementation o f the multi-sectoral HIViAIDS strategy also requires continued support from the highest political levels, and the placement o f the program under the direction o f the Prime Minister's Office, as i s currently the case, i s thus warranted. 11.3 Enhancing female adult education should receive the highest priority given its key role in improving human development outcomes and consumption poverty reduction, as well as women's empowerment. The estimated large beneficial effect o f female adult education on human development outcomes and poverty i s probably one of the most robust empirical relationships ever established in the human development literature. This underscores the urgent need to better understand the strong bias against investment in female 213 education which continues to persist in rural Ethiopia, and to perform action research by evaluating programs to promote girls' enrollment. Girls inrural areas were estimated to be 11 percentage points less likely than boys to be enrolled in primary school and eight percentage points less likely to complete primary school once enrolled. Moreover, women's education was the major differentiating factor in women's attitude towards wife beating. While 88 percent o f all women with no education found it justified to receive a beating by their husbands either for burning the food, arguing with him, going out without telling him, neglecting the children, or refusing sexual relations, this reduced to 83 percent for women with primary education and 57 percent among women with secondary and higher education. Continued investment in expanding school access in rural areas will be critical to enhance primary enrollment. 11.4 The substantial negative impact of harvest failure on human development outcomes highlights the need to strengthen people's ability to cope with shocks. The empirical results for both child malnutrition and school enrollment clearly indicate that households' efforts to cope with harvest failure come at the expense o f their children's human capital development. Inthe case o f schooling girls are especially at a disadvantage inthe face o f shocks, while the nutritional status o f boys tends to suffer more than that o f girls. Given the permanent damage malnutrition-induced growth retardation and interruption o f schooling impose on future eamings and development, farmers' risk management strategies must be strengthened. This could be done either by helpingthem to mitigate the effects o f shocks (e.g. through irrigation, pest and plant disease management) or by increasing their capacity to cope with shocks ex post (e.g. through better targeting o f food aid in response to shocks, development weather index based crop insurance schemes, strengthening the existing informal insurance schemes, or through productive safety nets and contingent transfer programs such as foodcash for work or for education). 11.5 Increasingawareness about the long term detrimentaleffects of early childhood malnutrition on future economic growth is necessary. A comprehensive and coherent multi-sectoral nutrition policy will need to be developed, with the institutional responsibilities o f the various ministries and mechanisms for coordination o f their actions and interventions in the field o f nutrition clearly delineated. Given the critical importance of early childhood malnutrition for economic growth, this agenda should be o f great concem to the Ministry o f Finance and Development. 11.6 Maternal education and especially health and nutritional knowledge play a critical and timely role in reducingchild malnutritionand child mortality. The empirical evidence on the determinants o f child stunting suggests that child growth monitoring and matemal nutritional education programs could play an important complementary role to other development actions such as promotion of food security, income growth, and more generally female and male adult education, which are already underway. Other direct nutrition interventions such as micronutrient supplements, promotion o f exclusive breast-feeding, and appropriate complementary feeding have also proven to be very cost effective as indicated by the Copenhagen Consensus.309 Moreover, while it will take a considerable amount o f time 309 Initiative by The Economist whereby a group o f Nobel Prize winners were asked to evaluate and rank different development initiatives according to their cost effectiveness. 214 before the other development actions substantially affect pre-school child growth faltering and child mortality, child growth monitoring and nutritional education programs as well as complementary feeding and micronutrient supplementation could take effect immediately, as illustrated by the successful ongoing Vitamin A supplementation program. The most promising interventions to reach the MDG o f reducing child mortality by two-thirds by 2015 from the 1990 level are enhancing maternal education and increasing access to safe drinking water. Given that 24 percent o f under-five child deaths are attributed to diarrhea, maternal healthknowledge and behavioral change will be equally critical. 11.7 There is a clear need for a multi-sectoral approach to improve people's human capabilities, and its institutional implications warrant further attention. The empirical analysis clearly indicates that irrespective o f the particular humandevelopment outcome (be it malnutrition, mortality or education), important opportunities for improving these outcomes are to be found outside the particular sectoral realm. 11.8 A better understandingof the existence of synergies between and the appropriate sequencing of interventions is needed to inform a multi-sectoral approach toward improving human capabilities. While the empirical analysis presented in Part I11properly identifies the effect o f each individual intervention on a particular human development outcome, it is likely that there also exist synergies in the implementation o f different interventions. Could, for example, income and maternal nutritional knowledge be considered complements, or do they rather substitute for one another? In the former case, an integrated approach i s more effective to address child malnutrition, or alternatively, if one o f the determinants i s extremely low relative to the other, a sequenced approach would be called for. However, in the latter case, imparting maternal nutritional knowledge might suffice irrespective o f the poverty status of the household, and there i s no particular gain from an integrated approach. In the case o f child stunting there is evidence that income and community nutritional knowledge act as substitutes, suggesting that there may substantial gains from imparting nutritional knowledge, a relatively easy to implement and low cost intervention, even if people remain very poor. Yet, further investigation o f the interactions between poverty and nutritional knowledge i s necessary. There may also be a threshold below or above which determinants begin to act as complements. The existence o f synergies also has important implications for the equity-efficiency trade-off, as substantial efficiency gains may be derived from focusing a package o f interventions on some areas, which could come at the expense o f intervening inothers. 11.9 Access to information is critical to enhance people's capability to aspire and expand their choice. Wide dissemination of radios and mass civic education programs provide a cost effective and commanding medium to do so, especially when the majority o f the population is illiterate and physically isolated. Radio programs are major dialogue initiators, often empowering individuals and fostering societal change. This i s exemplified by the deeply disturbing story o f Woineshet, a 13-year old rape victim in southern Ethiopia whose father's decision to bring her case to court was prompted by his exposure while working in Addis Ababa to radio announcements and bus ads urging the prosecution o f rape cases, Enhancing people's access to information will not only require increasing people's 215 access to radios but also providing an enabling legal framework fostering open debate and supporting citizens' rights to information. 11.10 In addition to enhancing people's human capabilities and their access to (and production of) information,i.e. their agency, people's opportunity structure should also be strengthened to foster empowerment of citizens in general and women (and pastoralists)in particular. To further facilitate the ongoing transition from traditional norms to national legal frameworks discussed in Chapter 2, actions should focus on continued support to existing programs o f decentralization and support to the development of independent civil society. Particularly crucial to foster empowerment through these programs would be: (1) a continuous emphasis on capacity building at the woreda and kebele level to ensure effective use o f block grants for poverty reducing purposes; (2) the enhancement of the interface between kebele and woreda councils, and between citizens and both o f these organizational entities; (3) the increased involvement o f citizens in the formulation o f kebele plans, budgeting and monitoring; and (4) establishing functional mechanisms o f accountability, including annual performance appraisals. 11.11 To improve the position of women in Ethiopiansociety actions are recommended in the legal, social and economic spheres: 0 Further support is required to hone and deepen government strategies supporting equal legal protection o f women. This includes: (1) better aligning the penal codes and application o f existing laws to make them consonant with the word and spirit o f Article 25 o f the constitution and the National Policy on Women; (2) providing training on gender sensitization to judges, lawyers and other members of the legal profession; (3) establishing a watchdog to track changes in the law and its application; and (4) supporting legal advocacy groups and providing legal aid, women's advisory centers and shelters for abusedwomen. 0 The entrenched social norms and practices that discriminate against women in society in general and the home in particular should be addressed. The billboard actions against gender based violence undertaken by the Gurage Women and Teacher's Association are encouraging signs o f civic engagement in this context (see Picture 11.1). This includes: (1) ensuring that gender issues are appropriately addressed in all development interventions and government programs both in policy and practice; (2) organizing training events for women parliamentarians and other champions o f women's issues on communication skills, assertiveness, computer literacy, gender budgeting, planning, monitoring and evaluation for women MPs; and (3) continuing the focus on girls' education and promoting the inclusion o f gender sensitivity programs in education curricula. 0 Finally, women's participation in the formal economy should be increased by: (1) providing incentives to businesses to hire women; (2) providing business management training and follow-up support to women; and (3) continuing the expansion o f credit availability to female entrepreneurs. 216 "Stop viofeiice against wonien!" "Abduetron of girls is badpractice "Preparedby ChehaWoredaWomen It shouldbe stopped!" andTeacher's Association" 217 APPENDICES Appendix 1: Ethiopia'sAction Planto Strengthen SDPRPMonitoringand Evaluation A.1 Background The Government o f Ethiopia established a Welfare Monitoring System (WMS) in 1996 to monitor the effect o f economic policy on social outcomes. Key institutions are the Welfare MonitoringUnit inMoFED, the Central Statistical Authority and several ministries. The Welfare Monitoring Unit (WMU), which is part o f the Economic Policy and Planning Department in the Ministry o f Finance and Economic Development (MoFED) i s entrusted with the coordination o f the monitoring and evaluation system. The WMU is responsible for compiling and analyzing data collected by other institutions in order to provide performance reports on SDPRP implementation and for commissioning o f relevant research and the dissemination o f the findings. The Economic Planning and Policy Department i s responsible for drafting the SDPRF' Annual Progress Report, which lays out the annual progress made in the implementation of the SDPRP and the poverty and welfare changes associated with SDPRP policies and programs. The WMU i s responsible for ensuring that the relevant data are collected for monitoring purposes. As part o f the plans to strengthenthe monitoring and evaluation o f the SDPRP, the Welfare Monitoring Unit i s absorbing new responsibilities to spearhead the national agenda for monitoring and evaluation across government and in collaboration with a number o f non-governmental actors (academic, civil and international). The increased responsibility will be matched by an increase in its technical and financial capacity (see section on Institutional Capacity). The Central Statistical Authority (CSA) i s the main data collection authority inEthiopia. For the last fifteen years, it has carried out a complex program o f surveys and censuses, including the population and agricultural censuses, the Household Income Consumption and Expenditure Survey (HICES), the Welfare Monitoring Survey (WMS), the Health and Nutrition Survey and the Labor Force Survey. It has recently completed its second Medium Term National Statistical Program (MTNSP) (2003-2008) and i s currently in the middle o f a process o f institutional strengthening. Alongside the CSA, several ministries collect socioeconomic data, mainly from administrative sources, relevant for the monitoring o f the SDPRP implementation. Several line ministries, especially those with articulated sector programs, run their own monitoring systems. The sectors whose information systems are most relevant for the monitoring o f the SDPRP are education, health and HIV/AIDS, water, roads, agriculture, food security and vulnerability, public sector management and capacity building, private sector development, and macroeconomic and finance. The level o f development o f sector-specific information systems differ significantly from sector to sector. As a result, some sectors are able to deliver facility-level data (e.g. education), while others can only make data available by region (e.g. health). 219 The monitoring and evaluation action plan has been developed as part o f the preparation o f the Poverty Reduction Support Credit (PRSC) o f the World Bank, incollaboration with other donors, to support the implementation and monitoring o f the SDPRP. A strong monitoring and evaluation system will deliver accurate and timely information on the achievement o f the agreed set o f prior actions and targets that will trigger the release o f annual PRSC budget support funds by donors. A.2 Objective The objective of the monitoring and evaluation system is to provide the government with reliable mechanisms to measure the efficiency o f government and the effectiveness o f public policies inachieving the objectives stated inits SDPRP. A.3 Design The M&E system for the SDPRP takes into account the multi-sectoral approach to achieving national growth and poverty reduction objectives, as well as the newly implemented decentralized structure o f intergovernmentalresponsibilities. It i s designed to: a) Monitor input and process indicators across levels o f government (e.g. public expenditure, adoption o f reforms) as a measure o f implementation; b) Monitor output indicators (e.g. education, health, infrastructure) at various levels o f aggregation (household, woreda, national) as a measure o f institutional efficiency; c) Monitor developmental outcomes and final objectives to track overall progress; d) Relate performance to indicators o f reform processes for decentralization, capacity building and civil service reform to provide information on the effectiveness o f the reform process inimproving outcomes; e) Evaluate impact to determine the effectiveness o f key government policies and programs inreaching desiredobjectives. A.4 Components Resultsframework and reporting mechanisms The SDPRP has a results framework matrix, which forms the basis for intergovernmental agreements o f major outputs and outcomes to be achieved, and for the design of the M&E system. However the quality o f the SDPRP matrix will need to be strengthened in the following areas: a) Strengthen the analytical linkages between government policies and programs, allocated inputs, expected outputs, and desired developmental outcomes needed to achieve the overall objectives o f the government program; b) Improve the definition o f indicators including baseline values and medium-term targets; c) Specify sources o f data and responsibility for monitoring. 220 The current SDPRP reporting mechanisms are weak. The government plans to strengthen reporting on SDPRP progress, and regularize dialogue with donors. Specifically: a) Strengthenthe structure and content o f the SDPRP Annual Progress Report, ensuring close linkages with results-based monitoring framework; b) Establish regular consultative processes with donors and civil society, including a series o f monitoring and evaluation workshops. An initial workshop held on May 21, 2004 brought together a cross section o f government agencies (WMU, CSA and sector ministries) responsible for monitoring and evaluation implementation with several non-governmental stakeholders for the govemment to present its monitoring and evaluation strategy and obtain overall endorsement and feedback on key areas of implementation. Follow up events will provide a venue for consultation and in-depth discussion on key implementation challenges. Improve data quality and availability The government proposal for "Strengthening data collection, analysis and dissemination on poverty monitoring and the Millennium Development Goals", which has received funding support from the Development Assistance Group (DAG), develops a comprehensive strategy to improve the quality o f poverty data and analysis and enhance their use and dissemination. The proposal focuses on the activities by the Central Statistical Authority and the Welfare Monitoring Unit that will (a) create the necessary qualitative and quantitative evidence on welfare outcomes at low levels o f aggregation, and (b) enhance the broad and fast dissemination and usage of data to all stakeholders. Central Statistical Authority. The CSA proposes to create the necessary data on welfare outcomes, via new household surveys conducted at low levels o f aggregation (including the Household Income and Consumption Expenditure Survey for 2003/04, bi-annual Core Welfare Indicators Questionnaire surveys), and a more effective organization o f the available data in the form o f a Socio-Economic Database. The CSA would also improve usage o f data collectedby the CSA via the development o f a far reaching dissemination strategy. The Ethiopian Socioeconomic Database. The aim of the database would be to ensure the availability o f all quantitative evidence relevant to poverty monitoring and evaluation in an integrated manner. The CSA will develop an integrated electronic database o f all available survey data and other data, and all relevant documentation in a user-friendly and relevant way, both for intemal purposes within the CSA and other government departments, as well as for all other stakeholders. Welfare Monitoring Unit. To ensure the availability o f the necessary evidence from quantitative and qualitative sources on poverty and welfare outcomes and inputs, the WMU would: (1) coordinate the collection o f participatory and qualitative data on poverty via a participatory poverty assessment (PPA); (2) construct a poverty map o f Ethiopia, including poverty estimates at the woreda level, using advanced techniques exploiting the census and HICES data; and (3) establish an IntegratedAdministrative Management Information System, 221 combining administrative, budgetary and socio-economic data to monitor progress in SDPRP implementation and other purposes. Integrated Administrative Management Information System. The aim o f an integrated M I S i s to collect and organize all relevant data, including administrative data, budgetary data and all relevant socio-economic data, in a user-friendly way for policy planning and monitoring, at levels o f disaggregation relevant for decentralized policy making. The availability o f an integrated data management system will facilitate timely and opportune delivery and processing o f information by the WMU for tracking progress in the implementation o f the SDPRP. The integrated data management system will include administrative information from all relevant sectors at the woreda, regional and national levels to afford the government with vertical and horizontal monitoring capacity, strengthen its ability to identify problem areas and low performers, and improve its capability to affect success. The design o f the overall system and installation inMoFED is being financed under the DAG proposal. The Bank support, via its Institutional Development Fund grant, will attend to the institutional development in key sector ministries including education, health, HIVIAIDS, water, roads, agriculture, food security, public sector reform and capacity building, private sector, and macroeconomic and budget sectors to: a) Determine indicators for each sector MIS requiredfor SDPRP monitoring; b) Design and implement modules o fthe I-MIS for each sector; c) Design reporting mechanisms from local to central government; d) Design action plans for addressing capacity building requirements to improve data flows from local governments (woreda) to central ministries. Developing capacity to implement the SDPRPmedium-term research agenda Poor analysis o f effectiveness o fpublic policy was identified inthe government action plan as one o f the reasons why efforts on the data collection and monitoring do not translate in better and more results-based decision making. As part o f its harmonization strategy, the government decided to develop a medium-term research agenda in consultation with interested partners and with the support o f the Ethiopian research community. The objective o f the agenda i s to establish the effectiveness o f policies inpriority areas to deliver expected results and provide early feedback to redress policy in case o f failure. It will include policy analysis and impact evaluation studies geared towards validating the effectiveness o f key government policies and programs inreaching desired objectives. The Bank, through an IDF grant, will support the strengthening o f MoFED's capacity for monitoring and evaluation analysis through technical assistance and training on methods and tools for monitoring and evaluation. Impact analysis o f alternative policies and programs will focus on key SDPRP areas including agriculture, water, and decentralization. The capacity building efforts will take place in close collaboration with international and national researchers in government, research institutes, and universities to benefit from world class 222 expertise. Partnerships with Ethiopian universities and research institutes will be established to enable staff o f the MoFED to benefit from analytical expertise and to strengthen links between data generation, policy analysis, and policy formulation. The following activities will be required: a) Design and implement mechanisms for consultation and feedback with the research community. Develop roster o f analysts and researchers, and website; b) Develop medium-termresearch agenda through consultative process with research and donor communities; c) Build analytical capacity in MoFED, through training and technical assistance, and fundpriority evaluation analysis; d) Undertake analysis o f key SDPRP policies. This is expected to help the government improve its analysis of the effectiveness of policies and programs indelivering results, feed this information into the budgetprocess to strengthen the overall effectiveness o f public spending, and improve the prospects for attaining MDG targets. Strengthen the environmentfor performance-based decision making International experience shows that it i s very difficult to operationalize the incorporation o f monitoring and evaluation results in the planning and budgeting processes. For this reason, the government will pay special attention to developing clear mechanisms to link planning, budgetingand monitoring and evaluationprocesses. Activities will include: a) identify key ministries and organizations that are or will be involved in monitoring performance, their function and interaction; b) develop and implement training plan on monitoring and evaluation tools and best practices; c) develop mechanisms to strongly link planning and budgetingprocesses to monitoring and evaluation results; d) Design and implement mechanisms to widely disseminate monitoring and evaluation results to program managers, policy-makers, communities and other civil society organizations. Participatory monitoring The government has not beenvery receptive to the implementation o f a policy for open access to public information. As an interim measure, the government drafted a policy for data access which addresses the immediate concerns regarding access to primary data sources-required to enhance policy analysis and improve data quality through producerhser linkages. On the participatory monitoring side, civil society made some progress to reorganize under the recently established umbrella group called Poverty Action Network (PAN). The network has brought together a range o f civil society organizations and narrowed down some areas 223 where these organizations think they can contribute to monitoring implementation o f the SDPRP, namely, expenditure tracking and monitoring the quality o f public service delivery. The government i s moving consciously in its relations with civil society, waiting for P A N to demonstrate its ability to deliver. Strengthen institutiona1capacity Strengthening SDPRP monitoring and evaluation will require the Welfare Monitoring Unit playing a large and critical role inmanaging and spearheading the monitoring and evaluation agenda o f the government, one that cannot be accomplished without a very significant strengthening o f its capacity. New responsibilities include: managing the integrated data management system; leading the implementation o f government wide monitoring and evaluation action plan across levels and sectors o f government, ensure monitoring and evaluation standards, and support monitoring and evaluation needs assessment and capacity building; heading the SDPRP annual review process, coordinating the sector reviews, and drafting the SDPRP Annual Progress Report; coordinating the national research agenda; liaising with non-governmental institutions; developing protocols for horizontal (sector to sector) sharing o f knowledge and information within the government; and liaising with other organizations (academic, international and non governmental) to strengthenthe ability o f the government to leverage outside resources. The government has already approved the expansion o f WMU staff, and outside resources have already beenallocated to expand technical capacity and the quality o f human resources. 224 Appendix 2: SupplementaryFigures . Figure A.l.l: Ethiopia- Percent of Total Territory Exposed to Malaria ETHIOPIA--- Percent of Total Territory Exposed to NlalnrL LEGEND , 225 Figure A.1.2: Ethiopia-Percent of populationvulnerableto malaria. ETHIOPIA Parcant of Population V'ulntlrable to Wabria I,, *I- .,'./r. II',.,.0 Figure A.2.1: Educationalattainment by gender ( O h ) . 80 l3.L 1 70 60 50 40 30 20 10 0 None Completed Primary Completed Secondary Source: FDRE 2001 OMen Women ~ ~ 226 Source: CSA and ORC Macro, 2001 llMen Women 227 Appendix 3: SupplementaryTables Table A.l.l: Ethiopia, Priceindicesin 1999 (at 1995/96constantprices) Months and prices Jul-99 Jan-00 Feb-00 Average price Addis Ababa General 108.7 102.8 105 106 Food 112.6 100.7 103.5 107 Non-food 104.8 105.3 106.7 105 Rural areas General 115.5 107.3 108.6 112 Food 123.2 105.6 108.7 115 Non-food 104.6 109.7 108.5 107 Urban areas General 119 115.5 115.7 117 Food 125.8 111.8 114.1 119 Non-food 110.5 120.1 117.1 115 228 Table A.1.2: Ethiopia: Poverty lines per reporting area, 1995-1999 Reporting area 1995 1999 Food Upper Foodpoverty Lower poverty Upper poverty Lower poverty line line poverty line line poverty line line 11 RuralTigray 712.83 920.50 1143.99 749.08 895.67 983.72 12 RuralAfar 599.37 838.07 851.43 688.37 864.51 903.63 13 RuralNorth& South Gondar 661.32 796.49 896.60 760.17 889.86 948.36 *14 Rural East, West Gojam & Agawi 526.67 711.44 822.64 661.21 772.79 865.78 15 Rural North Wollo, Oromiya Zone 693.51 831.56 899.13 736.14 861.47 904.65 16 RuralSouth Wollo, Oromiya Zone & North Shewa 619.56 760.76 898.83 704.30 878.44 979.80 17 Rural East & West Wellega 590.04 765.55 934.02 649.93 803.89 891.99 18 Rural Illubabor & Jimma 608.82 763.00 985.24 631.21 738.42 875.12 19 RuralNorth & West Shewa 556.41 720.52 1002.85 694.39 874.20 992.21 21 UrbanMekele 915.55 1291.68 2023.42 1063.99 1434.36 2722.41 22 UrbanBahir Dar 818.75 1092.67 1275.97 885.72 1201.45 1687.94 23 UrbanGonder 877.66 1184.94 1437.76 1007.63 1301.67 1973.40 24 UrbanDessie 917.98 1189.76 1593.67 1009.36 1266.79 1589.97 25 UrbanJimma 716.07 943.65 1219.71 918.85 1286.37 1801.49 26 Urban DebreZeit 767.83 1030.28 1450.46 967.09 1306.37 2203.72 27 UrbanNazareth 862.75 1146.80 1509.08 1014.03 1428.30 2221.89 28 UrbanHarar 963.75 1182.43 1572.17 1127.41 1459.38 2077.50 29 UrbanAddis Ababa 937.33 1226.02 1548.75 998.57 1297.72 1845.75 110 Rural East Shewa, Arsi, Bale & Borena 648.67 818.04 1051.55 625.00 774.40 951.17 111 Rural East & West Hararghe 619.41 733.73 901.76 547.20 620.68 668.67 112 Rural Somalie 483.29 558.52 679.81 663.95 818.98 927.30 113 Rural Benishangul-Gumuz 747.19 938.02 1082.79 624.54 791.OO 887.52 114 RuralYem, Keficho, Maji, Shekicho & Bench 654.37 837.80 963.53 547.24 677.56 756.12 115 Rural North & South Omo, Derashe & Konso 373.06 479.72 624.96 562.35 699.12 754.29 116 Rural Hadiya, Kambata & Gurage 701.32 895.26 1013.57 591.57 716.85 867.77 117 Rural Sidama, Gedeo, Burji & Amaro 888.27 1159.48 1338.70 581.40 700.13 854.65 118 RuralGambela 908.14 1129.46 1301.06 628.24 811.67 897.03 119 RuralHarari 751.21 879.10 1201.83 696.55 872.47 994.30 120 RuralAddis Ababa 904.73 1126.75 1363.09 880.51 1120.01 1438.87 121 RuralDireDawa 739.45 880.02 1173.87 688.49 800.35 876.88 210 UrbanDireDawa 1135.49 1446.12 1828.44 1143.09 1361.39 1570.33 Other Urban Centers 691.89 927.41 1180.11 890.50 1192.97 1793.96 229 Table A.1.3: Ethiopia,Growthinconsumptionby expenditure decile, 1995-1999 National Rural Urban Growth rates of consumption Expenditure Decile 1995 1999 1995 1999 1995 1999 National Rural Urban 1 617.85 623.45 622.46 625.11 594.06 613.06 1.01 1.oo 1.03 2 848.65 868.52 845.07 872.13 878.88 844.39 1.02 1.03 0.96 3 1013.60 1023.15 1003.20 1022.50 1104.04 1028.30 1.01 1.02 0.93 4 1171.17 1175.88 1152.63 1172.91 1315.24 1194.32 1.00 1.02 0.91 5 1328.80 1332.98 1303.37 1327.24 1537.32 1377.76 1.00 1.02 0.90 6 1507.89 1510.58 1471.28 1500.63 1782.35 1593.39 1.00 1.02 0.89 7 1713.51 1715.73 1665.45 1698.37 2088.66 1879.13 1.00 1.02 0.90 8 1984.38 1968.59 1912.50 1937.19 2511.20 2332.84 0.99 1.01 0.93 9 2400.74 2391.27 2283.28 2315.92 3181.47 3059.03 1.00 1.01 0.96 10 4090.46 4026.65 3736.71 3672.82 5580.36 5978.02 0.98 0.98 1.07 230 Table A.1.4: Ethiopia: Reportingadult equivalent total food consumptionand share of food, 1995-1999 Real Household Real Per adult Share of food Reporting area expenditures per adult equivalent food expenditures intotal equivalent expenditures expenditures 1995 1999 1995 1999 1995 1999 11 Rural Tigray 1412.78 1409.92 740.47 894.51 0.70 0.80 12 Rural Afar 2038.56 1770.12 833.41 726.34 0.66 0.75 13 RuralNorth & South Gondar 1263.97 1629.24 665.56 904.11 0.76 0.79 14 Rural East, West Gojam & Agawi 1493.43 1937.76 570.54 873.57 0.70 0.78 15 RuralNorthWollo, Wag Hamra 1211.14 1430.13 703.94 831.44 0.79 0.80 16 Rural South Wollo, Oromiya Zone &North Shewa 1483.28 1501.79 671.47 795.06 0.72 0.77 17 Rural East &West Wellega 1732.88 1809.68 641.84 885.70 0.67 0.74 18 RuralIllubabor & Jimma 1893.37 1501.42 701.77 794.64 0.68 0.78 19 RuralNorth & West Shewa 1965.05 1928.80 716.47 837.64 0.64 0.71 110 Rural East Shewa, Arsi, Bale & Borena 1664.43 1599.80 744.07 781.02 0.68 0.70 111 Rural East & West Hararghe 2087.75 1631.33 1026.72 912.32 0.76 0.80 112 RuralSomalie 2597.24 2313.29 924.58 902.57 0.71 0.75 113 RuralBenishangul - Gumuz 1296.66 1346.97 725.63 752.30 0.72 0.70 114 Rural Yem, Keficho, Maji, Shekicho & Bench 1492.85 1514.92 695.81 757.75 0.69 0.72 115 Rural North and South Omo, Derashe & Konso 1708.02 2058.95 444.55 658.72 0.69 0.77 116 Rural Hadiya, Kambata & Gurage 1319.93 1197.25 660.18 653.52 0.69 0.74 117 Rural Sidama, Gedeo, Burji & Amaro 1257.75 1106.91 748.98 826.24 0.67 0.76 118 Rural Gambela 1464.31 1021.63 994.46 735.92 0.73 0.71 119 RuralHarari 2615.70 1901.38 1553.51 1191.26 0.74 0.75 120 Rural Addis Ababa 1543.28 1520.95 923.32 976.38 0.68 0.65 121 RuralDireDawa 1595.91 1573.94 1023.01 1071.50 0.82 0.82 21 UrbanMekele 1761.01 1983.25 906.77 1003.24 0.56 0.58 22 UrbanBahir Dar 1568.59 2266.89 833.15 1002.36 0.66 0.55 23 UrbanGonder 1512.21 2308.48 850.36 1132.86 0.65 0.60 24 Urban Dessie 1098.15 1595.56 661.82 975.99 0.66 0.64 25 Urban Jimma 1913.42 1771.96 819.49 727.89 0.60 0.57 26 UrbanDebre Zeit 1597.68 1953.89 719.13 805.32 0.62 0.55 27 Urban Nazareth 1666.41 1806.38 882.09 794.53 0.61 0.53 28 UrbanHarar 1853.16 1780.16 1210.92 1127.74 0.70 0.63 29 Urban Addis Ababa 2036.89 2030.22 1180.94 1024.04 0.63 0.59 210 Urban Dire Dawa 1386.88 1417.28 1111.43 1239.90 0.70 0.74 211 Other UrbanCenters 2151.30 2003.75 989.38 831.79 0.64 0.60 231 Table A.2.1: Prevalenceof female circumcision ~~~ ~~ ~~ ~~~~~~ Background Percentageof women Percentagewho Characteristics circumcised support practice Number Age 15-19 70.7 53.4 3,710 20-24 78.3 57.0 2,860 25-29 81.4 58.5 2,585 30-34 86.1 65.2 1,841 35-39 83.6 63.6 1,716 40-44 85.8 66.3 1,392 45-49 86.8 66.7 1,264 Residence Urban 79.8 31.0 2,79 1 Rural 79.9 66.1 12,576 Region Tigray 35.7 25.3 969 Affar 98.6 76.5 178 Amhara 79.7 60.3 3,820 Oromiya 89.8 69.6 5,937 Somali 99.7 77.3 175 Benishangul-Gumuz 73.7 53.8 160 SNNP 73.5 59.8 3,285 Gambela 42.9 26.8 40 Harari 94.3 51.3 41 Addis Ababa 79.8 16.2 684 Dire Dawa 95.1 45.5 79 Education No education 80.4 67.0 11,551 Primary 78.4 48.5 2,425 ~ Secondaryandhigher 78.2 18.6 1,391 I Employment Not employed 79.5 59.1 5,630 Employedfor cash 84.4 56.1 3,852 62.7 59.7 15,367 J 232 Table A.2.2: Elements of regional and local government Kebele: WoredaiMunicipality 1Region Assembly All residents make upthe N / A kebele association. Association members meet annually - inan assembly. Council Cabinet Cabinet members are gove unent employees. (Also Cabinets are composed Cabinets are composed o f elected (councilors) and referred to as o f councilors. co-opted members (from ministryoffices). Co- `executive Chairpersons have fill- opted members commonly include members o f the committee ') time posts, and other administration, judiciary and security. The cabinet posts tend to be chairpersodchiefexecutive i s often the woreda or part-time. region administrator. Responsible for Most cabinets have between 7 to 10 members, preparing kebele plans mostly full-time. (with sector offices), and Cabinets (through committees with sector offices' submitting these to the support) prepare budgets and plans for approval by woreda. council. Responsible for ensuring At the woreda level, set and collect local taxes, collection o f land and maintain rural tracks, water points, and income taxes, and administrative infrastructure, administer schools and organizing in-kind health services, and manage agricultural contributions to development. development activities. Appoints judges to the "social courts," charged with resolving local disputes. Sector Civil servants commonly All ministries have woreda andregional offices. departments found are development agents and health post Source: WorldBank Country Office in Ethiopia 2002; INTRAC 2004a 233 Table A.2.3: Focus grouprankingsof top five most important institutionsinpeople's lives, ruralsites (Dessie Zuria Woreda, Amhara Region) Institution/Focus Group Site A Site B Site C Farmers ChurchiMosque 2 3 Idir Peasant Association (Kebele) 4 4 Training Center Health Clinic 5 Neighborshelatives 1 Kire 3 Market 1 Work 2 WidowedWomen (elderly) Forest 4 Market 5 Idir 1 Hospital Peasant Association (Kebele) 3 Kire 1 Mosque 2 2 Sheiks 3 Dubarti (Praying) 4 Youth Dua (Praying) Traditional Medicine Health Clinic 3 Peasant Association (Kebele) 4 Police Station Kire 2 Idir 1 Mosque 2 Tsebel (holy water) 5 Source: Rahmato and Kidanu 1999 234 Table A.2.4: Focusgroup rankingsof top five most important institutionsinpeople's lives, urbansites Institution/Focus Group Aada Liben Woreda, Dessie Zuria Addis Ababa, Kebele Kebele 1I, Woreda, Kebele II, 23 Oromiya Region Amhara Region Unemployed Kebele 1 3 Police Station 2 5 4 courts 3 ChurchiMosque 4 2 2 Idir 5 1 Tsiwa (religious gathering) 4 HealthServices 1 Tsebel (Place near church for washing 3 with holy water) Neighbors 5 Housewives Idir 1 Kebele Administration Tsebel 4 Church 3 Mahiber Neighborsirelatives Hospital Elderly 5 Youth Kebele 1 2 Church 2 1 Hospital 3 4 Municipality 4 Market 5 2 Relatives 3 Governmentforest 1 GrainMill 3 Idir 5 Source: Rahmato and Kidanu 1999 235 Table A.4.1: Estimated effects of radio ownership across location 1995-1999l)'*) RURAL URBAN Private endowments Human CapitaVDemographics Size of household -0.0984 -47.15 -0.1057 -44.61 Dependencyratio -0.0113 -2.9 -0.0528 -9.29 Ratio of Femalesinhouse 0.1529 8.85 0.2381 10.46 Femaleheadof house 0.0022 0.23 -0.0905 -7.93 Age -0.0052 -3.95 -0.0215 -1 1.32 Age squared 0.0000 2.89 0.0002 8.91 Grade obtained by adultmales 0.0183 11.65 0.0086 6.46 Grade obtained by adult females 0.0153 6.54 0.0183 12.32 No of adults completed post secondary 0.1925 3.76 0.1538 12.64 Physical capital Own plough 0.0905 9.91 0.1337 6.04 Own farm animal 0.0592 6.47 0.0644 3.97 Own transport animal 0.1199 14.44 0.0126 0.53 Own bicycle 0.1374 2.7 0.1056 3.75 Own radio 0.1747 14.89 0.2065 17.54 own tv3 -0.3342 -2.55 0.3531 20.45 EA mean radio ownership 0.3992 7.91 0.1315 3.8 No toilet in household -0.0845 -5.97 -0.1500 -11.97 Livelihoods Obtain some income from coffee (=I) 0.0332 3.35 -0.0471 -2.54 Obtain some income from chat (=1) 0.0852 5.27 0.1159 2.91 Share in income from agriculture 0.0428 1.77 0.0445 1.57 Share of incomefrom wages 0.0538 1.59 -0.0303 -2.17 Share of income from other sources -0.5420 -15.9 -0.0127 -0.61 Number of livelihood strategies engagedin -0.0815 -10.13 -0.0558 -6.35 Public endowments Electricity as source of household energy 0.0650 1.74 0.0964 5.98 Distanceto food market (km) -0.0004 -0.55 0.0113 4.18 Distanceto water (km) -0.0024 -4.03 -0.0087 -2.64 Distanceto Health (km) 0.0017 2.93 -0.0026 -1.35 Distanceto transport services(km) -0.0016 -6 0.0008 1.26 Year (=1 if year==1999) 0.0737 8.06 -0.1190 -10.55 ..-.--. 179.02. 8.2644 ...._163.89 _______I__ .- 14984 12097 R-squared 0.4599 0.43 ') In this model, we pool the observations for both the 1995 and 1999 samples. Pooled woreda Fixed Effect Model, similar to model in Table 4.10, column I,but with radio andTV ownership disaggregated and inclusion of enumeration _ _ - area-levelaverage ownership of radiosto capture externalityeffects. 2, Coefficientson woreda dummies not included. 3, The negative sign on the retums to TV in rural areas is likely due to an outlier. There were only 20 householdswith TV in rural areas. 236 Table A.4.2: Estimated effects of radio ownership across locationin 1999, with additional controls for wealth effect 1999l"*) Rural Urban Log of per adult equivalentexpenditures Coef, t Coef. t Private endowments HumanCapitalDemographics Size o f household -0.1046 -37.74 -0.1342 -43.62 Dependency ratio -0.0180 -3.59 -0.0322 -4.48 Ratio o f females inhouse 0.1601 7.41 0.1697 6.1 Female head of house 0.0092 0.78 -0.1110 -7.68 Age o f household head 0.0000 0.03 -0.0130 -5.51 Age Squared 0.0000 -1.14 0.0001 3.54 Grade obtained by adult males 0.0149 7.78 0.0103 6.24 Grade obtained by adult females 0.0134 4.62 0.0141 7.97 No o f adults completed post seconds# 0.2621 4.58 0.1299 9.89 Physical capital Ownplough 0.0680 6.05 0.0858 3.18 Own Cattle or Sheep 0.0443 3.39 0.0353 2.16 Own transport animal 0.1188 11.5 0.0944 3.31 Wall dummy (1 if stoneibrick wall) 0.0266 1.01 0.1355 5.79 Roof dummy (1 if corrugated iron) 0.1291 9.32 0.0431 1.75 Number o f Rooms 0.0242 3.56 0.0659 14.39 Ownbike 0.2076 3.13 0.1368 4.26 Ownradio 0.1632 11.76 0.1429 9.81 OwnTV -0.0323 -0.04 0.2918 14.66 Ea level radio 0.4065 6.47 0.0986 2.14 No toilet inhousehold -0.0847 -5.03 -0.1359 -8.49 Livelihoods Obtain some income from coffee (1=yes) 0.0417 3.03 -0.0150 -0.55 Obtain some income from chat (l=yes) 0.0875 4.26 -0.0963 -13 2 Share inincome from agriculture 0.0183 0.49 0.0530 1.33 Share o f income from wages -0.0323 -0.65 -0.0545 -3.05 Share o f income from other sources -0.5993 -13.42 -0.0452 -1.97 Number o f livelihood strategies engaged in -0.0898 -9.8 -0.0382 -3.51 Public endowments Electricity as source o f household energy -0.0011 -0.02 0.1510 7.23 Distance to food market (km) -0.0023 -1.77 0.0027 0.96 Distance to water (km) 0.0052 2.46 -0.0048 -1.31 Distance to Health (km) 0.0031 3.78 -0.0026 -1.19 Distance to transport services (km) 0.0001 0.33 0.0042 2.61 Constant 7.7045 132.44 7.7737 116.66 Number o f observations 8091 7664 R-sauared 0.5258 0.4834 ') In this model, we add additional controls for household wealth such as the material used for the wall and roof o f the house. This information was only available for 1999. A woreda Fixed Effect Model i s used (similar to model in Table 4.10, column 1, but with radio and TV ownership disaggregated and inclusion o f enumeration area-level average ownership o f radios to capture externality effects and dummy variables for the quality o f roof and walls o f the house. 2, Coefficients on woreda dummies not included. 237 Table A.6.1: Profile of food aid distribution across market position #ofhhs %ofhhs Average Mean consecutive Market Position Degree o f food aid per food per food Net Sales grain receipts years o f `) aid aid (Birr) receipts receipts receipts food aid (Kgs) receipts Net Buyers N o FoodAid 1,903 82 -552.03 0.00 1.52 Some Food Aid 262 11 1189.43 34.47 1.80 Significant Food Aid 161 7 1891.78 144.99 2.39 Autarkic N o FoodAid 244 80 0.00 0.00 1.20 Some FoodAid 35 11 0.00 32.84 1.98 Significant FoodAid 27 9 0.00 185.68 2.57 Net Sellers N o FoodAid 1,289 89 534.06 0.00 1.39 Some Food Aid 113 8 317.89 33.41 1.87 Significant Food Aid 53 4 697.50 144.24 2.41 ')These categories for food aid receipts were defined as follows: For all households with positive food aid, the mean food aid amount was the cut-off such that households more than the mean were classified as "Significant Food Aid" recipients and those getting positive amounts less than the mean as "Some Food Aid" recipients. 238 Table A.8.1: Estimated child, household and communitydeterminantsof child height for age (pooled sample) Without Short & long - Short run Woreda Height for age z-scores (children 3-60 months) prices runprice price effects fixed effects 1) effects2' (1) (2) (3) (4) Child characteristics Sex (l=male) -0.124 -0.124 -0.124 -0.118 (5.30) (5.36) (5.35) (5.34) child age (months) -0.049 -0.048 -0.048 -0.049 (13.97) (13.84) (13.85) (14.50) child age squared 0.001 0.001 0.001 0.001 (13.03) (12.92) (12.94) (13.64) child i s twin (l=yes) -0.372 -0.382 -0.387 -0.383 (2.10) (2.12) (2.15) (2.29) Householdcharacteristics Number o f adult males (16-65 years old) 0.038 0.040 0.041 0.055 (1.80) (1.96) (1.97) (2.83) Number o f adult females (16-65 years old) 0.042 0.039 0.035 0.034 (1.78) (1.65) (1.48) (1.53) Numbero f elderly (>65 years old) 0.032 0.025 0.025 0.026 (0.66) (0.54) (0.54) (0.57) female headed household (l=yes) 0.012 0.003 0.003 0.045 (0.20) (0.05) (0.06) (0.84) highest grade completed by most educated 0.031 0.033 0.034 0.030 female adult (4.38) (4.74) (4.93) (4.79) post secondary education (l=yes) most 0.230 0.208 0.221 0.214 educated fem. adult (1.81) (1.59) (1.70) (152) info on female adult education (1-0; O=yes) 0.139 0.158 0.154 0.163 (1-00) (1.18) (1.15) (1.28) Highest grade completed by most educated 0.016 0.010 0.009 0.008 male adult (3.15) (2.02) (1.83) (1.6) Post secondary education (l=yes) most 0.246 0.268 0.276 0.327 educated male adult (3.01) (3.23) (3.34) (3.78) info on male adult education (l=no; O=yes) 0.153 0.142 0.142 0.127 (2.27) (2.19) (2.18) (2.06) Log real household expenditure per adult 0.179 0.164 0.161 0.193 equivalent3' (2.44) (2.27) (2.21) (2.81) Community characteristics Sanitation, health, communication infrastructure Non-self proportion hhsicluster who drink 0.209 0.297 0.393 0.295 water from own tap (1.34) (1.94) (2.60) (1.77) Non-self proportion hhsicluster with flush 0.285 0.308 0.320 0.242 toilets (1.19) (1.28) (1.35) (1.OO) Distance to nearest health center (15km) -0.001 -0.001 0.001 0.013 (spline) (0.07) (0.06) (0.08) (1.13) Distance to nearest health center (>5km) 0.003 0.003 0.001 -0.013 239 Without Short & long Short run Woreda Height for age z-scores (children 3-60 months) prices runprice price effects fixed effects 1) effects') (1) (2) (3) (4) (spline) (0.22) (0.29) (0.14) (1.07) Non-selfproportion hhsicluster who own radio 0.246 -0.007 -0.068 -0.109 (2.24) (0.07) (0.62) (0.92) Non-selfproportionhhdcluster who own TV 0.181 0.281 0.192 0.296 (0.73) (1.12) (0.76) (1.19) Food andfuel prices maize 0.284 0.316 (1.94) (1.92) Teff -0.436 -0.262 (5.27) (2.56) Sorghum 0.074 -0.083 (1.06) (0.61) Oil 0.018 0.036 (0.98) (1.73) Beef 0.037 -0.015 (2.91) (0.95) Sheep 0.000 0.001 (0.3 1) (1.12) Goat -0.004 -0.002 (1.25) (0.50) Unpasteurizedmilk 0.069 0.102 (1.50) (2.01) Sugar -0.012 -0.046 (0.35) (1.04) Kerosene -0.128 -0.07 1 (2.15) (0.97) Charcoal -0.269 -0.106 (3.79) (1.25) Geographical location and time rural(1=yes) -0.019 -0.065 -0.190 -0.200 (0.27) (0.64) (1.45) (2.27) enset producing zones (l=yes) 0.399 0.135 (7.44) (1.16) Year 1996 (1=yes) -0.450 -0.354 -0.467 -0.458 (13.01) (5.38) (5.97) (14.40) Year 1997 (l=yes) -0.497 -0.35 1 -0.408 -0.502 (11.47) (4.97) (4.79) (13.21) Constant -2.887 -2.407 -2.05 1 -4.042 (5.34) (4.52) (3.33) (7.26) Observations 45751 45751 45751 45751 R-squared 0.04 0.06 0.06 0.11 Absolute value of t-statistics inparentheses. ')Regional dummies are not shown. *)Woreda dummies are not shown. 3,Predicted;householdassets and land ownership are the identifying instruments. Source: Christiaensen and Alderman 2004 240 Table A.8.2: Child malnutritionalleviating potentialof different policy interventions Prevalence o f child stunting (%) Original observations Predicted observations after intervention after intervention Intervention base Direct % total % base direct % total % effect change effect change effect change effect change Pooledsample(model 2, table 3) Income andformal schooling 1) annual per adult equivalent income growth 63.2 61.6 -3 88.2 84.9 -4 o f 2.5 % for 15 yrs 2) at least one female adultlhousehold 63.2 59.1 -6 58.7 -7 88.2 78.0 -12 77.3 -12 educated up to primary level 3) at least one male adultihousehold educated 63.2 61.8 -2 61.7 -2 88.2 86.1 -2 85.6 -3 up to primary level 4) joint intervention(1) & (2) 63.2 57.5 -9 88.2 73.3 -17 Cerealprices 5) cereal price increase by 25 % 63.2 65.1 3 88.2 91.7 4 6) cereal price increase by 25 % 63.2 64.0 1 87.3 89.8 3 (model3,table3) Nutritional knowledge 7) increase inproportion o f right judgments 54 52.3 -3 59.7 55.7 -7 by 25 % points 8) increase inproportion o f rightjudgments 54 5 1.2 -5 59.7 52.4 -12 by 50 % points 9) joint intervention(4) & (7) 54 46.4 -14 59.7 41.0 -31 Source: Christiaensen and Alderman 2004 241 Table A.9.1: Diarrhea incidence and care-seeking Poorest Richest Below poverty Above poverty Average Quintile Quintile line line Prevalence o f diarrhea children < 5 years National average 25.4 19.3 24.1 21.0 22.6 Regional variation Tigray 18.3 13.2 18.3 17.2 17.7 Afar 23.1 19.5 16.2 16.4 16.3 Amhara 16.5 14.8 19.2 18.0 18.8 Oromiya 27.3 24.2 26.0 24.6 25.4 Somali 11.5 25.5 20.1 18.2 19.3 Benishangul-Gumuz 32.1 22.9 30.3 22.2 26.9 SNNP 29.4 27.3 30.4 28.4 29.6 Gambela 30.1 25.8 27.9 25.5 26.8 Harari 27.9 20.0 25.8 22.9 23.6 Addis Ababa 12.5 12.8 12.8 Dire Dawa 27.7 17.0 32.1 19.2 21.3 N o treatment sought 79.1 56.8 79.8 67.2 74.2 Treatment sought Public Sector Hospital 0.7 9.8 0.8 4.8 2.6 Health Center 2.6 9.3 2.4 6.2 4.1 Health Post 0.4 0.3 0.3 1.o 0.7 Community Health Worker 0.0 0.0 0.1 0.0 0.1 Other Public Sector including Health 7.7 9.1 7.0 8.7 7.8 Station Private Sector Private DoctoriHospital 0.9 5.3 0.8 3.1 1.9 Pharmacyishop 4.2 6.0 4.9 5.7 5.3 Other Private Sector 4.6 3.5 3.7 3.2 3.5 Treatment given ORS 7.8 33.2 8.5 19.72 13.1 RHF at home 1.4 9.4 3.131 9.082 5.5 Home remedyiothers 7.8 3.9 6.69 7.997 7.2 Others (Injectiodpillsisyrup) 21.2 18.0 20.28 18.38 19.5 None o f the above 61.9 35.5 61.4 44.82 54.7 Knowledge o f ORS Never heard of ORS 39.6 13.1 42.3 21.7 33.5 Used ORS 2.5 9.3 3.0 6.1 4.3 Heard of ORS 57.9 77.6 54.7 72.3 62.2 Source: World Bank 2004b 242 Table A.9.2: Burden of HIV/AIDS inAfrican countries') Number ofPeopleLiving Adults prevalence, 15 to 49 Orphans (0-14 with HIViAIDS years years old) AIDS deaths,2o01 SouthAfrica 5,000,000 20.1 660,000 360,000 Nigeria 3,500,000 5.8 1,000,000 170,000 Kenya 2,500,000 15.0 890,000 190,000 Zimbabwe 2,300,000 33.7 780,000 200,000 Ethiopia 2,100,000 6.4 990,000 160,000 Tanzania 1,500,000 7.8 810,000 140,000 DRC 1,300,000 4.9 930,000 120,000 Zambia 1,200,000 21.5 570,000 120,000 Mozambique 1,100,000 13.0 420,000 60,000 Cameroon 920,000 11.8 210,000 53,000 China 850,000 0.1 76,000 30,000 Malawi 850,000 15.0 470,000 80,000 Cote d'Ivoire 770,000 9.7 420,000 75,000 Sub-SaharanAfrica 28,500,000 9.0 11,000,000 2,200,000 Morerecent figures estimate the number ofpeople livingwith HIViAIDs inEthiopia at 1S00,OOO (Federal Democratic Republic o fEthiopia,2004) Source: World Bank 2004b 243 Table A.9.3: Summary statistics of variables included in the child mortality regression (rural specification) Variable Number Observations Mean Std. Dev. Min Max Under 5 Child Mortality Rate 5915 0.22 0.42 0 1 Females 5915 0.48 0.50 0 1 Female HeadedHousehold 5915 0.15 0.36 0 1 Mother'sAge 5915 33.73 6.87 18 49 Mother'sAge Squared 5915 1185.04 480.92 324 2401 Mother'sPrimary School Completion 5915 0.0826712 0.28 0 1 Mother'sHeight 5877 1567.37 65.98 465 1867 Mother'sWeight 5881 485.23 65.30 295 1574 Mother'sBMI 5871 1972.81 230.35 1295 4226 WealthQuintile 3 5915 0.33 0.47 0 1 WealthQuintile 4 5915 0.12 0.33 0 1 WealthQuintile 5 5915 0.11 0.3 1 0 1 No Water Access 5773 0.3824701 0.49 0 1 PipedWater 5763 0.06 0.23 0 1 OpenWater 5773 0.4432704 0.50 0 1 Covered Water 5773 0.1158843 0.32 0 1 No Sanitation 5772 0.9052322 0.29 0 1 Ethnicity: Orthodox 5915 0.3967878 0.49 0 1 Ethnicity: Muslim 5915 0.4104818 0.49 0 1 Peri-Urban 5910 0.0267343 0.16 0 1 Tigray 5915 0.1159763 0.32 0 1 Affar 5915 0.09 0.29 0 1 Amhara 5915 0.15655 11 0.36 0 1 Oromiya 5915 0.1952663 0.40 0 1 Benishangul 5915 0.0772612 0.27 0 1 SEP 5915 0.20 0.40 0 1 Gambela 5915 0.0385461 0.19 0 1 Harari 5915 0.04 0.20 0 1 Dire Dawa 5915 0.0226543 0.15 n 1 244 Table A.10.1: Sample statistics for education regressions-rural and urban regressions Variable Urban Rural Female Male Female Male Currently enrolled 0.82 0.85 0.24 0.34 Child age 8 0.12 0.12 0.15 0.15 Child age 9 0.12 0.12 0.13 0.13 Child age 10 0.14 0.14 0.14 0.13 Child age 11 0.10 0.10 0.09 0.10 Child age 12 0.15 0.16 0.14 0.15 Child age 13 0.13 0.12 0.10 0.10 Child age 14 0.14 0.13 0.10 0.11 Female headed household 0.38 0.35 0.19 0.18 Age of household head 44.4 44.8 45.8 46.1 Single head o f household 0.21 0.20 0.10 0.09 Years of schooling: household head 4.45 4.19 0.73 0.69 Years of schooling: other adult 4.06 3.80 0.39 0.35 Log o f per capita household expenditures 7.16 7.09 6.82 6.79 0.23 0.24 EA - Average male literacy 0.77 0.77 0.33 0.33 EA Average female literacy - 0.52 0.53 0.09 0.09 Distance primary school 1-2 km 0.42 0.40 0.27 0.26 Distance primary school 3-4 km 0.05 0.05 0.25 0.26 Distance primary school 5-6 km 0.01 0.01 0.16 0.16 Distance primary school 7-12 km 0.00 0.00 0.11 0.11 Distance primary school > 13 km 0.00 0.00 0.02 0.03 Distance secondary school 2.49 2.57 23.54 23.58 Distance food market 1.31 1.22 6.83 6.80 Distance health clinic 1.42 1.45 8.89 9.07 Distance post office 2.81 2.88 24.30 24.27 Numbero f schools per capita 0.26 0.27 0.25 0.25 Population density 126.65 129.57 123.58 125.24 Student-teacher ratio 59.01 59.32 60.3 1 60.65 Percent female teachers 29.00 28.72 26.46 26.22 Percent teachers with certification 0.91 0.91 0.89 0.89 Tigray 0.13 0.12 0.07 0.07 Amhara 0.29 0.29 0.22 0.22 Oromiya 0.39 0.38 0.32 0.32 Benishangul 0.04 0.04 0.05 0.05 SNNPR 0.11 0.12 0.31 0.31 Year 2000 0.59 0.60 0.67 0.67 245 Table A.10.2: Educationoutcome regression results, rural and Probability child o f primary school Probability o f completing 5th age (7-14) is currently enrolled in grade for children aged 12 to school3' 144' Independent Variables Probability Probability Probability Probability Complete Complete Current Enrolled Current Enrolled Grade 5 Grade 5 Rural (1) Urban (2) Rural (3) Urban (4) Dummy:Child age 8 0.11 0.077 (5.87) ** (4.61)** Dummy: Childage 9 0.229 0.12 (11.58)** (7.71)** Dummy:Childage 10 0.309 0.144 (15.38)** (9.48)** Dummy: Child age 11 0.304 0.133 (14.08)** (8.25)** Dummy:Child age 12 0.333 0.148 (16.75)** (9.86)* * Dummy: Child age 13 0.342 0.119 0.011 0.095 (15.94)** (7.34)** -0.87 (3.80) ** Dummy:Child age 14 0.355 0.125 0.008 0.215 (16.44)** (7.84)" * -0.62 (8.76)* * Female child -0.116 -0.043 -0.075 -0.098 (14.21)** (4.06)** (7.05)** (4.78)* * Female-headed household 0.004 0.014 0.007 -0.075 -0.21 -0.78 -0.27 (2.18)* Age household head -0,001 0.001 0 0.005 -1.79 (2.80)** -0.2 (4.91)** Single household head -0.046 -0.003 -0.016 0.037 -1.79 -0.11 -0.48 -0.86 Schooling o f household head 0.011 0.008 -0.004 0.008 (4.92)" * (5.66)* * -1.21 (2$98) ** Schooling o f non-head adult 0.011 0.004 0.005 0.012 (3.38)* * (2.28)* -1.03 (4.20)** Log o f HHper capita expenditures 0.056 0.025 -0.002 -0.001 (6.12)* * (2.30)* -0.17 -0.07 Rain-fall shock at plot level -0.042 -0.029 (2 96) ** I -1.58 EA-level average literacy rate o f males 0.098 0.155 0.062 0.138 (3.67)* * (3.41)** (1.7)" -1.54 EA-level average literacy rate o f females 0.443 0.202 0.154 0.106 (9.67)* * (4.5 1)** (2.21)* -1.23 Primary school is within 2 km -0.035 -0.02 -0.01 -0.008 (3.00) ** -1.74 -0.61 -0.35 Primary school is within 3-4 km -0.054 -0.025 0.019 -0.141 (4.60) ** -0.97 -1.11 (2.96) ** Primary school i s within 5-6 la -0.098 -0.019 -0.003 -0.193 (7.57)* * -0.33 -0.16 (2.02)" 246 Probability child o f primary school Probability o f completing 5th age (7-14) i s currently enrolled in grade for children aged 12 to school3' 144' IndependentVariables Probability Probability Probability Probability Complete Complete Current Enrolled Current Enrolled Grade 5 Grade 5 Rural (1) Urban (2) Rural (3) Urban (4) Primary school is within 7-12 km -0.149 -0.026 0 (10.49) ** -1.33 Primary school i s more than 13 km away -0.176 -0.035 -0.035 -0.352 (6.89)** -0.21 -1.07 -1.76 Distance to secondary school -0.001 -0.001 -0.001 -0.005 (2.05)* -1.18 -1.49 (2.62)* * Distance to food market 0.001 0.001 0.001 0.004 -1.07 -0.49 -0.62 -1.62 Distance to health clinic -0.001 -0.003 0 -0.002 -1.39 -1.13 -0.34 -0.32 Distance to post-office 0 0 0 0.001 -0.72 -0.36 -0.92 -0.81 Number o f schools inthe area 0.509 0.099 0.162 0.154 (12.55)** -1.38 (2.61)** -1.1 Population density 0 0 0 0 -0.83 (2.45)* -1.04 -1.4 Student-teacher ratio 0.001 -0.002 0 0 (5.53)** (4.65)* * -0.09 -0.16 Fraction o f female teachers 0 0.005 -0.002 0.001 -0.15 (5.13)** (2.62) ** -0.79 Fractions o f teachers with certificates 0.03 0.151 -0.026 0.011 -0.74 -1.71 -0.46 -0.06 Tigray 0.067 0.07 0.038 0.261 -1.62 -1.07 -0.63 -1.76 Amhara 0.089 0.115 0.027 0.211 (2.24)" (1.96)* -0.45 -1.4 Oromiya 0.053 0,091 0.072 0.262 -1.4 -1.15 -1.21 -1.72 Benishangul 0.032 0.082 0.053 0.072 -0.89 -1.69 -0.93 -0.63 SNPR 0.067 0.115 0.057 0.159 -1.76 -1.95 -0.98 -1.14 Year 2000 Dummy 0.144 0.089 0.11 0.046 (15.33)* * (5.65)* * (8.73)** -1.55 Constant 0.07 -0.397 -0.56 -1.46 Number o f Observations 11763 4979 4171 2135 Coefficients are reported as marginal probabilities. 2, Asignificant at 10 percent; * significant at 5 percent; ** significant at 1 percent. 3, Absolute value of z statistics inparentheses. 4, Robust t statistics inparentheses. 247 Table A.10.3: Educationenrollmentregressions-rural and urban in Ethiopia."*' Rural Urban Variable Girls Boys Girls Boys Child age 8 0.114 0.103 0.07 0.08 (4.59)* ** (3.68)* ** (2.88)* ** (4.09)* ** Child age 9 0.214 0.24 0.115 0.116 (7.94)* ** (8.35)" ** (5.05)*** (6.43)*** Child age 10 0.282 0.323 0.155 0.122 (10.25)*** (11.22)*** (7.16)*** (6.78)" ** Child age 11 0.249 0.352 0.15 0.11 (8.4 1)*** (11.47)*** (6.58)* ** (5.75) *** Child age 12 0.258 0.38 0.148 0.13 (9.49) *** (13.57)*** (6.87) *** (7.27) *** Child age 13 0.252 0.408 0.103 0.121 (8.59)* ** (13.68)*** (4.34) *** (6.39)** * Child age 14 0.294 0.391 0.113 0.118 (9.86)* ** (12.90)*** (4.88)* ** (6.21)*** Femaleheadof household 0.015 0.015 -0.017 0.041 -0.59 -0.52 -0.72 (1.83)* Age of headof household 0 -0.001 0.002 0.001 -0.57 (1.76)" (2.19)** -1.51 Single headof household -0.006 -0.087 0.022 -0.053 -0.2 (2.37)* * -0.75 - 1.54 Schooling of householdhead 0.016 0.004 0.008 0.008 (5.88)*** -1.14 (4.17)*** (4.55)** * Other adult schooling 0.012 0.006 0.002 0.007 (3.08)*** -1.18 -0.76 (3.33)*** Log of per capita household expenditures 0.051 0.065 -0.007 0.041 (4.40)*** (4.71)*** -0.46 (2.97) *** Rain damage -0.053 -0.02 (2.87)"** -0.94 EA- Average male adult literacy rate 0.055 0.17 0.045 0.206 (1.65)* (4.21)*** -0.7 (3.61)*** EA- Average female adult literacy rate 0.454 0.374 0.24 0.149 (8.26)*** (5.4 1)*** (3.71)*** (2.67)* ** Distance Primary School 1-2 km -0.037 -0.033 -0.022 -0.018 (2.68)*** (1.82)* -1.39 -1.24 Distance Primary School 3-4 km -0.067 -0.046 -0.057 -0.007 (4.80)*** (2.54)** -1.51 -0.22 Distance Primary School 5-6 km -0.104 -0.093 -0.071 0.034 (6.86)*** (4.55)*** -0.88 -0.48 Distance Primary School 7-12 km -0.122 -0.177 (7.07)*** (7.75)* ** Distance Primary School > 13 km -0.16 -0.179 -0.131 0.064 (4.56)*** (4.43)*** -0.6 -0.43 Distance to secondary school 0 -0.001 -0.002 -0.001 -1.17 -1.32 -1.41 -0.4 248 Variable Rural Urban Girls Boys Girls Boys Distance to food market 0.003 -0.001 0.002 0 (2.45)"" -0.43 -0.72 -0.05 Distance to health clinic -0.002 0 0 -0.007 (2.58)* ** -0.18 -0.04 (1.go)* Distance to post office -0.001 0 -0.001 0 (1.70)* -0.32 -0.4 1 -0.28 Number o f schools per capita 0.363 0.706 0.179 0.044 (7.07)*** (10.42)*** (1.69)* -0.5 Population density 0 0 0 0 (2.02)* * -1.27 -0.43 (2.76)* ** Student-teacher ratio 0.001 0.001 -0.001 -0.002 (4.23)*** (3.59)* ** (2.75)*** (3.79)* ** Percent female teachers 0 0 0.005 0.004 -0.1 -0.61 (3.79) *** (3.78)*** Percent teachers with certification 0.022 0.106 0.098 0.125 -0.43 (1.66)* -0.83 -1.18 Tigray 0.09 0.08 1 0.132 0.003 (1.69)" -1.29 (1.69)* -0.03 Amhara 0.105 0.118 0.166 0.064 (2.08)** (1.91)* (2.36)** -0.8 Oromiya -0.022 0.176 0.184 0.029 -0.5 (2.93)*** -1.62 -0.3 Benishangul -0.065 0.212 0.104 0.071 (1.79)" (3.41)*** -1.52 -1.23 SNPR -0.001 0.189 0.149 0.085 -0.02 (3.07) *** (1.87)* -1.15 Harari 0.038 0.272 0.126 -0.028 -0.71 (3.85)*** -1.64 -0.3 Year 2000 dummy 0.14 0.157 0.103 0.069 (11.76)*** (11.08)*** (4.70)*** (3.41)*** Number of Observations 5808 6299 2739 2629 'IAllCoefficients arereportedasmarginalprobabilities. 2,* significant at 10 percent; ** significant at 5 percent;***significant at 1 percent; robust z statistics are in parentheses I 249 Table A.10.4: Primary schoolcompletionregression-rural and urbaninEthiopia."*' Variable Rural Urban Girls Boys Girls Boys Child age 13 -0.002 0.028 0.147 0.252 -0.14 - 1.44 (5.25)*** (8.69) *** Childage 14 0.01 0.009 0.287 0.355 -0.62 -0.47 (10.84)*** (12.39)*** Female head o f household -0.014 0.021 -0.019 -0.015 -0.51 -0.57 -0.66 -0.44 Age of head o f household 0 0 0.003 0.002 -0.04 -0.27 (3.83)""" (2.04)* * Single head o f household 0.003 -0.041 -0.034 0.06 -0.08 -0.95 -0.93 -1.39 Schooling o f household head -0.001 -0.007 0.004 0.009 -0.34 -1.38 -1.63 (3.43)* ** Other adult schooling 0.003 0.007 0.003 0.004 -0.55 -1.29 -1.31 -1.57 Log of per capita household expenditures 0.002 -0.006 -0.037 0.002 -0.12 -0.34 (2.23)"" -0.09 Rain damage -0.02 -0.041 -0.89 -1.47 EA-Average male adult literacy rate 0.057 0.07 0.062 0.203 -1.39 -1.33 -0.78 (2.37)** EA-Average female adult literacy rate 0.126 0.113 0.069 -0.014 (1.91)* -1.31 -0.9 1 -0.17 Distance Primary School 1-2 km -0.013 -0.017 -0.048 0.01 -0.72 -0.73 (2.55)** -0.48 Distance Primary School 3-4 km -0.006 0.033 -0.078 -0.038 -0.3 1 -1.39 (1.92)* -0.8 Distance Primary School 5-6 km -0.016 0 -0.007 0.049 -0.76 0 -0.07 -0.38 Distance Primary School 7-12 km -0.023 -0.042 -0.95 -1.35 Distance Primary School > 13 km -0.06 -0.045 0.039 -1.11 -0.74 -0.16 Distance to secondary school -0.001 0 0 0.002 (2.30)"* -0.13 -0.21 -0.81 Distance to food market 0.003 -0.001 -0.007 0 (2.05)** -0.63 -1.45 -0.02 Distance to health clinic -0.001 0 0.003 -0.005 -1.08 -0.06 -0.64 -0.75 Distance to post office 0 0 -0.001 0.002 -0.68 -0.86 -0.51 -0.92 Numberof schools per capita 0.063 0.199 0.173 0.28 -1.03 (2.58)*** -1.22 (2.09)** Population density 0 0 0 0 -0.07 -0.99 -0.61 -1.33 Student-teacher ratio 0 0 0,001 -0.001 -1.09 -0.3 1 -1.21 -0.86 250 Variable Rural Urban Girls Boys Girls Boys Percent female teachers -0.001 -0.003 -0.001 0.003 -0.69 (2.80) *** -0.73 (1.80)* Percent teachers with certification -0.094 -0.009 0.169 -0.081 -1.38 -0.11 -1.19 -0.53 Tigray 0.061 0.037 0.196 0.221 -0.81 -0.43 -1.23 -1.43 Amhara 0.03 0.017 0.223 0.312 -0.44 -0.22 -1.32 (2.OO)** Oromiya 0.047 0.116 0.236 0.26 -0.7 -1.43 -1.53 (1.75)* Benishangul 0 0.121 0.246 0.174 0 -1.52 (1.93)* -1.48 SNPR 0.022 0.107 0.164 0.239 -0.33 -1.3 -1.06 -1.64 Harari -0.016 0.174 0.287 0.111 -0.26 (1.72)" (1.90)* -0.78 Year 2000 dummy 0.104 0.115 0.084 0.043 (6.87)*** (6.04)* ** (3.27) *** -1.5 Number o f Observations 2014 2290 2927 2731 All Coefficients are reported as marginal probabilities, 2, *significant at 10 percent; ** significant at 5 percent;***significant at 1 percent; robust z statistics are in parentheses. 251 Appendix 4: The Theoretical, Empirical, and Historical Case for Agriculture Led Development The focus on agriculture as the engine o f economic growth and poverty reduction i s grounded, both theoretically and empirically, in a close examination o f the linkages between the different livelihood systems (agriculture and non-agriculture) in rural ec~nomies.~"Three types o f linkages are usually identified: (1) production linkages, (2) consumption linkages, and (3) saving and investment linkages. Production linkages can be further classified as backward and forward linkages. Backward linkages follow from the increased demand for inputs, while forward linkages arise due to increased demand for processed products or other downstream activities, which stimulate the agro-processing industry. To better identify the conditions and factors which enable large multiplier effects from raising agricultural income, we illustrate the different linkages in Figure (A41.311 Exogenous changes in policies, technologies, institutions, markets, infrastructure or capital may induce changes in productivity and prices in rural economies (see left side o f figure). To examine how these changes work their way through the system, it i s important to distinguishbetween tradable and non-tradable goods and services, with tradables being defined as goods and services that can be imported and exported to and from the area.312Productivity increases in non-tradable activities often lead to lower prices as local demand does not increase sufficiently to absorb the additional supply. Ifthe concerned goods (e.g. staple or services make up a large share o f the budget, this will increase consumers' real income (see left bottom of graph). A reduction inthe price o f tradables will yield similar results. A consumption linkage will then arise as the increased real incomes translate into increased demand for locally produced non-tradable goods and services, which in turn generates local employment opportunities and income. A virtuous circle i s established, whereby the original gains inreal consumer incomes (from price declines due to productivity gains) are multiplied through an expansion o fthe local economy. 3'0Mellor, 1995; Adelman and Morris, 1973. 311The discussionrelies heavily on Kyddet al., 2001. 3'2The tradability o f goods and services depends on: (1) the size of the area (the larger the area, the larger the proportion o f non-tradables); (2) its accessibility (the less accessible, the larger the proportion o f non-tradables); and (3) the cost differential with the outside world. Together these factors determine the cost differential between areas or the spread between "import" and "export" parity prices. Note that while the terminology i s borrowed from international trade it is equally applicable to intra-national trade. 313Given prohibitive transport and marketing costs, cereal markets in Ethiopia are in effect isolated from world markets. Moreover, teff, an indigenous grain which makes up an important part o f the daily diet, i s only produced in Ethiopia. High intra-regional marketing costs further inhibit intra-regional trade, with only 28 percent o f total farm output marketed in 1996 (Gebremeskel et al., 1998). Of the cereals marketed, nearly one- thirdwas sold directly fromproducers to consumers. 252 The size o f the multiplier effect will critically depend on three key factors.314 If the extra income i s spent on tradables (e.g. radios and TVs) as opposed to locally produced non- tradables (e.g. non-agricultural goods and services such as housing improvements and locally provided services), local employment generation will be reduced. To the extent that increased income results inmore demand for locally produced food it will help keep up food prices and strengthen the production linkages (see below). Nonetheless, reviewing the empirical evidence Bell and Hazel1 (1980) emphasize that the multiplier effect increases the greater people's propensity to consume locally produced non-food goods and services. The multiplier effect i s also reduced when local producers cannot sufficiently respond to the increased demand for non-tradables. This generates inflationary pressures and offsets the increase in real incomes. The consumer linkage effect presupposes that there i s underutilized labor and other resources in the rural sector which can be mobilized. In addition to labor or capital constraints, inelastic supply may also follow from poor market development or high transaction costs. Finally, gains from increased demand for locally produced goods will also be reduced if the capital or import content o f the production is high, or if they only provide returns to a limitedgroup o fpeople. The effects on producers of increases in non-tradable productivity are mixed. Lower prices could potentially offset the gains from productivity if demand i s inelastic. Lower prices for tradables (e.g. cash crops) have similar negative effects on producers o f tradables. Higher prices and increased productivity in the production of tradables have positive effects on the income o f their producers. The multiplier effects o f the backward production linkages are often small. The increased demand for additional inputs (e.g. fertilizer and mechanical inputs) and marketing o f additional outputs does not generate substantial employment, as these goods are typically tradable and imported from abroad, or capital intensive. Additional employment will only arise from the distribution and marketing o f these goods, which i s a labor intensive activity. Inconclusion, while production linkages are generally considered to be weaker, consumption linkages are in fact found to be quite strong, especially in closed economy settings.315 Other linkages include potential savings and investment linkages, where increased real incomes stimulate savings and investment in capital. This could reduce household vulnerability and enhance the supply elasticity and productivity o f local non-tradables. The strength o f these linkages depends on the availability o f and returns to local investment opportunities, and the extent to which the local financial markets are already integrated with the wider economy. Finally, growth in production o f tradables may improve telecommunication and transport services, following the need to handle greater volumes. It may also lead to increased provision o f external agricultural inputs (e.g. fertilizer, improved seeds) as well as economies o f scope within the household where equipment used for production o f tradables could also be used for the production o f non-tradables. Givena better understanding o f the linkages betweenprice changes and productivity increases in tradables and non-tradables following exogenous investments, policy and institutional 3 I4Kyddet al. (2001) speak o f `leakages.' Gabremadhin, 2004. 3'5 Gabremadhin, 2004. 254 interventions, and the conditions under which these spawn the largest multiplier effects and have the greatest impact on poverty, we now turn to the relative roles o f farm and non-farm production in this process. Inrural areas, two broad sources o f growth can be identified: (1) growth inthe production o ftradables (which increases local incomes directly), and (2) growth in production of non-tradables (which increases local incomes indirectly by lowering For growth in tradables to be effective in reducing poverty, it must raise earnings among a large part of the population. This implies that production must be by the poor themselves (either as hiredlaborers or through self-employment), or be widespread with high labor content (so that the poor can benefit from the consumption linkage effects generated in the local economy). Apart from primary resource extraction (mining, forestry, fishery) it i s hard to imagine many other non-farm activities which engender broad employment opportunities in economies such as Ethiopia's with limited communication infrastructure and linkages to urban or export markets. Opportunities for other non-farm employment typically only develop as links with urban areas deepen.317Moreover, the few non-farm employment opportunities available at this stage o f development often have high entry barriers, limiting both the potential gains for the poor as well as the potential for widespread adoption and thus poverty reduction through consumption linkages. Expansion of the production of agricultural tradables (e.g. cash crops, tradable food crops, horticulture and livestock) offer much more potential, with direct gains from increasedemployment and income opportunities for the poor, as well as gains through backward and forward linkages(e.g. employment opportunities inprocessing). The extent to which the poor can benefit from advances in productivity (and increased prices) in agricultural tradables depends on asset distribution, which i s quite equal in Ethiopia, and their access to complementary inputs. Promoted technologies should be scale neutral and labor intensive (e.g. fertilizer, improved seeds). Nonetheless, while progress in cash crop production technologies may offer important opportunities for poverty reduction, greater opportunities for the poor are to be expected from consumption linkages resulting from productivity increases innon-tradable food production. Growth in non-tradables can engender significant poverty reduction through consumption linkages, if it concerns goodshervices with a high average budget share. Again, it is hard to imagine any other non-tradable which qualifies apart from farm products such as staple foods (cereals in Ethiopia). There will be important direct gains through decreased food prices for all net buyers and subsistence producers. However, the greatest benefits are to be expected from the consumption linkages through increased demand for and employment generation in the non-tradable non-farm sector (together with livestock and horticulture production). For the consumption linkage to result in sizeable multiplier effects, the income elasticity for non-food non-tradables must be large, and the local supply elastic and labor intensive. While net cereal sellers could potentially lose if demand i s inelastic, it must be emphasized that the majority o f the marketed surplus is usually produced by a 3'6Migrant labor and remittances, not considered here, could be another source o f growth for the local economy, with labor export ineffect being a tradable. 3'7Bryceson, 1999; Reardon, et al., 1994. 255 minority o f the farmers, who tend to be the larger and richer farmers.318 The majority o f households in rural Ethiopia are subsistence farmers and net food buyers. So are urban households. Food price decreases following technological change thus hold the promise o f substantially increasing real incomes. Nonetheless, for this strategy to be sustainable food price declines have to be gradual, and large fluctuations must be avoided as they generate disincentive effects for net sellers to sustain the adoption o f modern inputs, leading to fluctuations in output and prices which hurt the poor. This will require both a parallel increase in income and demand for food through growth in the non-food sector, reduction o f transaction costs through better development o f the market channels to cater to this increase in demand, and a better management o f food aid.319 We return to these issues in more depth below. Over the past decade, the Asian economies have often been heralded as models o f economic growth and poverty reduction. Close inspection shows that with the exception o f Singapore and Hong Kong, all the successfully transforming countries o f Asia experienced agricultural revolutions prior to industriali~ation.~~~These countries started from a primarily agrarian base with a stagnant and low productivity food sector, unstable food prices and heavy reliance on export crops, a situation much like that o f Ethiopia today. While not sufficient, a technologically driven agricultural transformation proved necessary in these countries to generate structural transformation. Overall, agricultural growth multiplier effects in different parts o f the world have been estimated to range from 1.5 to over two, implying that a US$ 1 increase inagricultural income generates an additional US$0.5 to1 inincome.321 3'8Weber et al., 1988. 319The estimatedamount of marketable surplus available for local food aid procurement was 530,000 MT, while the food aid needs o f the chronically food deficit (a population of about five million people) was estimated at 557,204 MT. Yet by March 2002, only 236,374 M T had been locally procured due to insufficient financing (Gabremadhin, 2002). 320Rosegrant and Hazell, 2000. 32'Reardon, 1998; Delgado et al., 1998. 256 Appendix 5: Price Fluctuations,Substitution, and Market Activity Per Cereal Are households selling low and buying high? Absent credit markets, liquidity constraints may force households to sell off their grains during the harvest seasonwhen prices are usually lower, and buy cereals later inthe season when their stocks are depletedand cereal prices are likely higher. Depending on the degree of seasonal price fluctuations, net cereal sellers may thus turn into net cereal buyers. While the limited average volume of cereals sold by net buyers indicates that this is unlikely to change most people's net market position, it is nonetheless useful to explore this further, as it would provide some sense o f the welfare diminishingeffects o flarge intra-annualprice fluctuations. Table (A5) 1presents the producer and retail price levels during the harvest, the post harvest and hunger season, while Table (A5)2 presents the sale and purchase transactions across net buyers and sellers in these three periods. Table (A5)l: Intra-annual cereal price fluctuation in 1995-96 Harvest season Post harvest season Hunger season Ratio retail price (Oct 95-Jan 96) (Feb - May 96) (June-Sept 96) hunger seasoniproducer price Producer Retail Producer Retail Producer Retail harvest season Maize 0.85 0.98 0.84 1.oo 1.oo 1.11 1.31 Wheat 1.41 1.69 1.32 1.68 I.37 1.78 1.26 Teff 1.55 1.78 1.42 1.76 1.44 1.83 1.18 Barley 1.17 1.50 1.22 1.53 1.22 1.59 1.36 Sorghum 1.03 1.18 0.95 1.17 I.06 1.12 1.09 Millet 1.02 1.52 1.02 1.48 1.38 1.64 1.61 Average 1.17 1.44 1.13 1.44 1.25 1.51 1.30 Source: Own calculationsfrom CSAproducer and retail market price series From Table (A5)l it can be seen that in 1995-96 the ratio o f the retail price duringthe hunger season and the producer price during the harvest seasonranged between 1.09 for sorghum and 1.61 for millet, which i s the least important cereal. The average ratio amounted to 1.30, about the ratio for maize and a bit larger than the ratio for teff, two much traded cereals. Harvest and post harvest prices were very similar on average, while hunger season prices were only five to 10 percent larger in 1995-96. Retail prices tended to be 20 to 25 percent higher than producer prices. Table (A5)2 shows that the majority o f the sales transactions among net buyers occur during the post harvest season (February-May), while purchases beginduring the harvest season with their intensity increasing thereafter.322 In other words, net buyers would stand to lose from large intra-annual price fluctuations. The majority o f the purchase transactions among net sellers happen during the hunger season, while most o f their sales transactions happen during the post harvest season. Insum, given that (1) net buyers sell primarily duringthe post harvest season, while they buy especially during the hunger season, and (2) intra-annual price fluctuations were limited in 322Note that it only concerns the observed transactions up till July and that many households anticipated to buy cereals during August and September as well. 257 1995-1996 when compared to other years, the percentage of net buyers observed in 1995- 96 most likely represents a lower bound, compared to years characterized by larger intra-annualpricefluctuations. Table (A5)2: Incidence of cereal market transactions across net cereal buyershellers between October 1995 and September 1996 Net buyer Total # of Percentage o fpurchases Total # of Percentage of sales t n x w d o n s Oct 1995- Feb -May June-July transactions Oct 1995- Feb- M a y June-July Crop Jan1996 1996 1996 Jan1996 1996 1996 maize 1693 29 37 35 1431 2 66 33 wheat 933 24 40 36 473 28 63 9 teff 558 19 47 34 340 2 87 11 barley 996 17 42 39 630 5 88 7 sorghum 1142 18 37 46 991 14 52 34 millet 163 16 60 25 45 0 7 93 average 21 44 36 9 61 31 Net seller Total # of Percentage of purchase Total # of Percentage of sales transactions transactions transactions transactions Crop Oct 1995- Feb -May June-July Oct 1995- Feb- M a y June-July Jan 1996 1996 1996 Jan1996 1996 1996 maize 643 4 34 62 692 26 59 13 wheat 368 1 4 94 368 12 52 35 teff 654 1 25 73 944 11 57 31 barley 377 0 15 84 391 20 51 29 sorghum 43 1 1 5 94 484 9 60 30 millet 87 0 95 5 125 16 65 20 average 1 30 69 16 57 26 Source: Own calculationsfrom CSA Food Security Survey, 1996 Are households sellinghigh value grains to buy lower value ones? Another component o f marketbehavior that affects the valuation o fnet sales is the mix o fcrops that arebeingbought and sold. From Table (A5)3, we note for example that teff was the highest priced cereal, closely followed by wheat, while maize was the cheapest. Are households substituting more expensive cereals (wheat and teff) for cheaper ones? Ifthis i s an important phenomenon, one would expect households who sell a particular cereal (especially the more expensive cereals) to also buy a large quantity o f other cereals. Table (A93 presents the number o f purchase transactions for each cereal conditional on the household having sold a particular cereal. 258 Table (A5)3: Number of cereal purchase transactions conditional on the sale of a cereal Percentage o f households who buy a particular cereal given that they have sold Number o f sale Maize Wheat Teff Barley Sorghum Millet transactions') Maize 900 32 41 37 34 26 19 Wheat 414 18 32 24 23 23 13 Teff 1040 17 16 16 17 17 13 Barley 470 23 23 26 28 23 33 Sorghum 648 17 20 23 23 27 17 Millet 140 5 2 3 4 4 7 Total 3612 112 134 129 129 120 102 ')Not weighted by population expansion factors. Source: Own calculations from CSA Food Security Survey There is a significant amount of substitution between cereals, with households selling one cereal to buy back the same or a different one later. About 30 percent o f those who sell maize, wheat, barley or sorghum buy it back later, suggesting that large intra-annualprice fluctuations would be harmful. However, this happened only in one to six cases among those selling teff, the most expensive but also the most sold cereal. Millet i s least marketed and usually not sold to be bought back. Contrary to our hypothesis, it i s maize, a less expensive cereal, which i s mostly sold to buy back other cereals. Similarly, but less frequently, wheat, barley and sorghum are often sold to buy back other cereals. The most often bought cereals are wheat and teff, the more expensive cereals. As indicated above, our basic finding o f an important number o f net cereal buyers is not affected by these findings, as they also hold when looking at quantities only. Table (A94 explores further the extent to which the different crops are traded and how their sales and purchases are concentrated across net buyers and sellers. Maize emerges as a much traded crop, with about 30 percent o f all net buying rural households buying maize, and 20 to 30 percent selling maize depending on the season. Millet on the other hand i s the least traded crop. Interestingly,only 10 to 14 percent (dependingon the season) o f all the net buyingrural households buy teff, while 31 to 25 percent (depending on the season) o f all net selling rural households sell teff, This suggests that teff i s an important cash crop for most net selling households, that net buying rural households tend to meet more o f their own demand for teff, and that teff i s especially consumed in urban areas. This i s consistent with the observed consumption patterns in rural and urban areas. Rural households spent on average 9.7, 6.9 and eight percent o f their cereal purchases on maize, wheat and teff respectively, while in urban areas on average 1.6, 5.6 and 11.6 percent o f cereal purchases were spent on maize, wheat and teff. Wheat on the other hand i s more widely bought among net buying households, though its production i s concentrated inthe hands o f fewer households. 259 Table (A5)4: Crop level market participation characteristics by season Net Buver Autarkic N e t Seller Market Net Market Net Market Net Season Grain Partici- Partici- SalesNumber Partici- pation m e Number pation Sales Number pation Sales % ( B 4 % (Birr) % (Birr) Maize 1,490 31.04 140.89 1049 22.91 688 22.60 259.41 Wheat 916 19.08 166.07 547 11.95 362 11.89 363.75 Oct95- May96 Teff 493 10.27 157.82 987 21.56 950 31.21 281.53 Barley 765 15.94 147.67 882 19.27 372 12.22 263.64 Sorghum 984 20.50 151.56 829 18.11 552 18.13 190.00 Millet 152 3.17 121.02 284 6.20 120 3.94 152.09 Maize 1,270 30.79 172.83 1,613 22.63 344 29.45 125.59 Wheat 726 17.60 410.19 989 13.87 110 9.42 166.17 June 96- Sept 96 Teff 565 13.70 405.58 1,573 22.06 292 25.00 128.10 Barley 546 13.24 221.31 1,328 18.63 145 12.41 170.12 Sorghum 911 22.08 416.00 1,235 17.32 219 18.75 120.28 Millet 107 2.59 121.15 391 5.48 58 4.97 139.01 260 REFERENCES Aadland, 0.2002. Sera: Traditionalism or Living Democratic Values? Case Study among the Sidama in Southern Ethiopia. in Zewdie, B. and Pausewang, S. (eds). 2002. Ethiopia; The Challenge of Democracy from Below. Ethiopia: Forum For Social Studies. Abrar, S., O., Momssey and T.,Rayner. 2004. Crop-Level Supply Response by Agro-Climatic Region inEthiopia. Journal of Agricultural Economics 55(2): 289-3 11. Administrative Committee on CoordinatiodSub-Committee on Nutrition (ACC/SCN). 2000a. 4th Report on the World Nutrition Situation. New York: United Nations, in collaborationwith the International Food Policy Research Institute, Washington D.C. Administrative Committee on CoordinatiodSub-Committee on Nutrition (ACC/SCN). 2000b. Low Birth Weight. A Report Based on the International Low Birth Weight Symposium and Workshops Held on 14-17 June 1999 at the International Centre for Diarrhoeal Disease Research in Dhaka, Bangladesh. NutritionPolicy Paper #18. Edited by J. Pojda, and L.Kelley. Geneva: ACC/SCN incollaboration with ICDDRB. Adelman, I., and C.T. Morris. 1973. Economic Growth and Social Equity in Developing Countries. Stanford: Stanford University Press. Admassie, Y., F. Guta, and A., Ayalew. 2003. Social Viability: Spatial Population Balance and Rural Viability inEthiopia. mimeo. AgCnor, Bayraktar, and El Aynaoui, 2004. Roads Out of Poverty? Assessing the Links Between Aid, Public Investment, Growth, and Poverty Reduction. Background paper prepared for World Bank Country Economic Memorandum, Ethiopia. World Bank, Washington, DC. Alderman, H., and J. R. Behrman. 2003. Estimated Economic BeneJits of Reducing Low Birth Weight in Low-Income Countries. Philadelphia, PA: Universityo f Pennsylvania. Alderman, H., J. Hoddinott, and B. Kinsey. 2003. Long Term Consequences Of Early Childhood Malnutrition. Washington, DC: World Bank and International Food Policy Research Institute, Processed. Alsop, R., and N. Heinsohn. 2004. Measuring Empowerment - Structuring Analysis and Framing Indicators. Draft Working Paper. Poverty Reduction Group, PREM, World Bank, Washington, DC. Alsop, R., N. Heinsohn, and A. Somma. 2004. Measuring Empowerment: An Analytic Framework. Poverty Reduction Group, PREM, World Bank, Washington, DC. 261 Amacher, G., L. Ersado, D. Grebner, and W. Hyde. 2004. Disease, Micro dams and Natural Resources in Tigray, Ethiopia: Impacts on Productivity and Labour Supplies.Journal of Development Studies 40(6): 122-145. Appleton, S. 2003. Regional o f National Poverty Lines? The Case o f Uganda in the 1990s. Journal of African Economies, 12(4): 598-624. Appleton, S. 1995, Gender Differences in the Returns to Schooling in Three African Countries, Centre for the Study for African Economies, Asfaw, A., et al. 2004. The Economic Costs o f Illness inLow Income Countries: The Case o f RuralEthiopia. Quarterly Journal of International Agriculture 43(3): 247-266. Barker, D.J.P. 1998, Mothers, Babies and Health in Later Lfe, 2"d ed. New York, NY: Churchill Livingstone. Barrett, C., and D. Maxwell. 2003. Food Aid After 50 Years: Recasting its Role. New York: Routledge. Barrett, C., et al. 2000. Heterogeneous Constraints, Incentives and Income Diversification Strategies inRuralAfrica. mimeo. Basta S., S. Soekirman, D. Karyadi, and N.S.Scrimshaw. 1979. Iron deficiency anaemia and the productivity o f adult males inIndonesia. American Journal of Clinical Nutrition. 32: 916-925. Behrman, J. R., H. Alderman, and J. Hoddinott. 2004. Hunger and Malnutrition. Paper prepared for Copenhagen Consensus - Challenges and Opportunities (February 2004 draft). Bell, C., andP.B.R. Hazell. 1980. Measuring the indirect effects o f an agricultural investment project on its surrounding region. American Journal of Agricultural Economics 62( 1): 75-86. Bhargava, Alok. 2005. AIDS Epidemic and the Psychological Well-Being and School Participation o f Ethiopian Orphans. Psychology, Health and Medicine. forthcoming. Bold, T., J. D e Weerdt, S. Dercon, and A. Pankhurst. 2004. Extending Insurance: Funeral Societies inEthiopia and Tanzania. mimeo. Brautigam and Knack. 2004. Foreign Aid, Institutions, and Governance in Sub-Saharan Africa. Economic Development and Cultural Change 52(2): 255-285. Bruns, B.,A. Mingat, andR. Rakotomalala.2003. Achieving Universal Primay Education by 2015: A Chancefor Every Child. Washington, D.C.: World Bank. 262 Bryceson, D.F. 1999. Sub Saharan Africa Betwixt and Between. Working Paper. African Studies Centre, University o f Leiden, Leiden,Netherlands. Central Statistical Authority, Ethiopia. 2003. Report o f Urban Bi-Annual Employment Survey, October 2003. Federal Democratic Republic o f Ethiopia, Central Statistical Authority, Addis Ababa. Central Statistical Authority, Ethiopia. 2000. Analytical Report on the 1999 National Labour Force Survey. Statistical Bulletin234. Federal Democratic Republic o f Ethiopia, Central StatisticalAuthority, Addis Ababa. Central Statistical Authority (CSA) Ethiopia, 2001. Report on the Year 2000 Welfare Monitoring Survey Vols. Iand 11. Statistical Bulletin 259. Federal Democratic Republic o f Ethiopia, Central Statistical Authority, Addis Ababa. Central Statistical Authority, Ethiopia, and ORC Macro. 2001. Ethiopia Demographic and Health Suwey 2000. Addis Ababa, Ethiopia and Calverton, MD, USA: Central Statistical Authority and ORC Macro. Chavez, A., and C. Martinez. 1984. Behavioral Measurements o f Activity in Children And Their Relation to Food Intake in A Poor Community. in E. Pollitt and P. Amante, eds., Energy Intake and Activity. New York, NY:Alan R. Liss. Christiaensen, L., and H. Alderman. 2004. Child Malnutrition in Ethiopia: Can Maternal Knowledge Augment the Role o f Income? Economic Development and Cultural Change 52(2): 287-3 12. Christiaensen, Luc, and K. Subbarao, 2005, Toward an Understanding o f Household Vulnerability inRural Kenya, Journal ofAfrican Economies, forthcoming. Christiaensen, L.,V. Hoffinann, and A. Sarris. 2004. Vulnerability among Smallholder Coffee Growers inKilimanjaro, Tanzania. mimeo. Collier, P., and D.Dollar. 2001. Can the World Cut Poverty inHalf? How Policy Reform and Effective Aid Can Meet International Development Goals. World Development 29( 11): 1787-1802 Collier, P., and J. Gunning. 1999. Explaining African Economic Performance. Journal of Economic Literature, 36( 1): 64- 111. Commander, S. 2004. Ethiopia: Labour Market Study. mimeo. Conley, D., K. Strully, and N. Bennett. 2003. A Pound of Flesh or Just Proxy? Using Twins Differences to Estimate the Effects o f Birthweight on (Literal) Life Chances. New York University, Department o f Sociology, New York processed. 263 Croppenstedt, A., and C. Muller. 2000. The Impact o f Farmers' Health andNutritional Status on Their Productivity and Efficiency: Evidence from Ethiopia. Economic Development and Cultural Change 48(3): 475-502. Cropper, M. L., et al. 2004. The Demand For A Malaria Vaccine: Evidence from Ethiopia. Journal of Development Economics 75(1): 303-318. Dasgupta, P. 1997. Nutritional Status, The Capacity For Work, And Poverty Traps. Journal of Econometrics 77: 5-37. Deaton, A., and M. Grosh, 2002. Consumption. In Grosh, M. and P. Glewwe, eds. 2000. Designing Household Survey Questionnaires for Developing. Washington, DC: The World Bank. Deininger, K., S. Jin, B.Adenew, S. Gebre-Selassie, and M.Demeke.2003. Market and Non- Market Transfers o f Land in Ethiopia: Implications for Efficiency, Equity, and Non- Farm Development. World Bank, Washington, DC. mimeo. Delgado, L.C., et al. 1998. Agricultural Growth Linkages in Sub-Saharan Africa. IFPRI Research Report. International Food Policy Research Institute, Washington D.C. Demeke, M., A. Mekonnen, A. Admassie, and D. Aredo, eds. 2001. Technological Progress in Agriculture, Proceedings of the National Workshop on Technological Progress in Ethiopian Agriculture, Nov. 29-30, Addis Ababa, Ethiopia. Organized by the Department o f Economics, Faculty o f Business and Economics, Addis Ababa University. Demeke, M., A. Said, and T. Jayne. 1997. Promoting Fertilizer Use in Ethiopia: The Implications o f Improving Grain Market Performance, Input Market Efficiency, and Farm Management. Working Paper 5. Grain Market Research Project, Ministry o f Economic Development and Cooperation, Addis Ababa. Dercon, S. 2004. Growth and shocks: evidence from rural Ethiopia. Journal of Development Economics 74(2): 309-329. Dercon, S. 2002. The Impact o f Economic Reforms on Rural Households in Ethiopia: A Study from 1989 to 1995. Poverty Dynamics inAfrica Series. World Bank, Washington DC. Dercon, S. 1995. On Market Integration and Liberalization: Method and Application to Ethiopia, Journal of Development Studies October: 112-143, Dercon, S., and P. Krishnan. 2000a. Vulnerability, Seasonality and Poverty in Ethiopia. Journal of Development Studies 36(6): 25-53. 264 Dercon, S., and P. Krishnan. 2000b. In Sickness and in Health: Risk Sharing within Households inRural Ethiopia. Journal of Political Economy 108(4): 688-727. Dercon, S., and P., Krishnan. 1996. Income Portfolios in Rural Ethiopia and Tanzania: Choices and Constraints. Journal of Development Studies 32(6): 850-875. Devarajan, Miller and Swanson. 2002. Goals for Development: History, Prospects, and Costs. Policy Research Working Paper Number 2819. World Bank, Washington, DC. Development in Practice Ltd. 2003. Review of Capacity Building Approaches for Local Government in Ethiopia. Draft Report. Government o f Ethiopia and Department for InternationalDevelopment, UK. Diao, X. et al. 2004. Growth Options and Poverty Reduction inEthiopia - An Economy Wide Model Analysis for 2004-2015. International Food Policy Research Institute: Washington DC. Easterly, W. 2002. Growth in Ethiopia: Retrospect and Prospect. Center for Global Development, Institute for Intemational Economics, Washington, DC. Ehui, S., and J. Pender. 2005. Resource Degradation, Low Agricultural Productivity and Poverty in Sub-Saharan Africa: Pathways out o f the Spiral. Agricultural Economics 31(2): 217-233. Ellis, F. 2000a. Rural Livelihoods and Diversity in Developing Countries. Oxford: Oxford University Press. Ellis, F. 2000b. The Determinants of Rural Livelihood Diversification in Developing Countries. Journal of Agricultural Economics 5 l(2): 289-302. Ethiopian Economic Association / Ethiopian Economic Policy Research Institute, 2002, Land Tenure and Agricultural Development inEthiopia, mimeo. Fafchamps, M., and A. Quisumbing. 2002. Control and Ownership o f Assets within Rural Ethiopian Households. Journal of Development Studies 38(6): 47-82. Federal Democratic Republic o f Ethiopia (FDRE). 1997. Ethiopian 1996/96 Household Income, Consumption and Expenditure Survey and Welfare Monitoring Survey. Standardized Welfare Indicators. Provisional Results. Central Statistical Authority, Addis Ababa. Federal Democratic Republic of Ethiopia (FDRE). 2001. Report on the 1999/2000 Household Income Consumption and Expenditure Survey. Statistical Bulletin. Central Statistical Authority, Addis Ababa. 265 Federal Democratic Republic o f Ethiopia (FDRE). 2002. Ethiopia: Sustainable Development and Poverty Reduction Program. Addis Ababa: Ministry of Finance and Economic Development. Food and Agricultural Organization (FAO). 1975. Organic materials as fertilizers. FA0 Soil Bulletin27. FAO, Rome. Gabremadhin, E. 2004. Can Agriculture Lead Growth in Ethiopia? The Importance o f Linkages, Markets and Tradability. Background paper prepared for the Ethiopia Country Economic Memorandum. World Bank, Washington D.C. mimeo. Gabriel, A., and M. Demeke. 2003. Endowment Profiles and Adoption o f Agricultural Technologies: Distributional Dimensions and Impact on Direct Production Entitlement. in Demeke et al., eds. 2003. Technological Progress in Ethiopian Agriculture. Proceedings o f the national workshop on technological progress in Ethiopian agriculture, Nov. 29 - 30,2001, Addis Ababa. Gebremedhin, B., and S. Swinton. 2002. Sustainable Management o f Private and Communal Lands in Northern Ethiopia. in Barrett, C.B., F. Place, and A.A. Aboud, eds. 2002. Natural Resources Management in African Agriculture: Understanding and Improving Current Practices. Wallingford, Oxon: CAB1Publishing. Gebremedhin,Berhanu, Scott M.Swinton and Y. Tilahun (1999) Effects o f stone terraces on crop yields and farm profitability: Results o f on-farm research in Tigray, Ethiopia, Journal o f Soil and Water Conservation, v54, n3, pp. 568-573. Gebremeskel D., T. S. Jayne, and J. D. Shaffer. 1998. Market structure, conduct, and performance: Constraints on performance of Ethiopia grain markets. Grain Market Research Project (GMRP), MEDaC Working Paper 8, Addis Ababa, Ethiopia. Geresu, T. 1996. Burden o f Diseasehlortality Analysis - Ethiopia. PHRD Study Report No. 3. PHRD Office, Addis Ababa. Girishankar, N., A. Alemayehu, and Y. Ahmad. 2001. Handling Hierarchy in Decentralized Settings: Governance Underpinnings o f School Performance in Tikur Inchini, West Shewa Zone, Oromia Region.Africa Region Working Paper Series No. 21. World Bank, Washington, DC. Glewwe, P. 1999: The economics of school quality investments in developing countries :an empirical study of Ghana. New York, NY: St. Martin's Press. Glewwe, P., and H. Jacoby. 1995: An Economic Analysis o f Delayed Primary School Enrollment and Childhood Malnutrition in a Low Income Country, Review of Economics and Statistics 77(1): 156-69. 266 Glewwe, P., H. Jacoby, and E. King. 2001. Early Childhood Nutrition and Academic Achievement: A Longitudinal Analysis. Journal of Public Economics 81(3): 345-368. Haas, J., S. Murdoch, J. Rivera, and R. Martorell. 1996. Early nutrition and later physical work capacity. Nutrition Reviews 54: S41-S48. Hoddinott, J.; Owens, T.; Kinsey, B. 2003. The impact o f agricultural extension on farm production in resettlement areas of Zimbabwe. Economic Development and Cultural Change 51(2): 337-358. Holden, S., S. Benin, B. Shiferaw, and J. Pender. 2003. Tree Planting for Poverty Reduction in Less-Favoured Areas of the Ethiopian Highlands. Small-scale Forest Economics, Management and Policy 2( 1): 63-80. Intrac. 2004a. Building Capacity in Ethiopia to Strengthen the Participation o f Citizens' Associations in Development: A Study o f the Organizational Associations o f Citizens. Draft paper prepared for the World Bank. Oxford, United Kingdom. Intrac. 2004b. Study on Effective Empowermentof Citizens in Ethiopia. Draft paper prepared for the World Bank. Oxford, UnitedKingdom. de Janvry, A,, and E. Sadoulet. 2002. World Poverty and the Role o f Agricultural Technology: Direct and Indirect Effects. Journal of Development Studies 38(4): 1-26. Jayne, T.S., et al. 2003. Smallholder Income and Land Distribution inAfrica: Implications for Poverty Reduction Strategies. Food Policy 28(3): 253-75. Jayne, T.S., J. Strauss, T. Yamano, and D. Molla. 2002. Targeting o f food aid in rural Ethiopia: chronic need or inertia? Journal of Development Economics 68(2): 247-288. Kakwani, N. 2004. New Global Poverty Counts. International Poverty Centre, United Nations Development Program. I n Focus September 2004: 9-11. Available: http://www.undp.org/povertycentre/newsletters/infocus4sep04eng.pdJ: Kaufmann, D., A. Kraay, and M. Mastruzzi. 2003. Governance Matters 111: Governance Indicators for 1996-2002. World Bank, Washington, DC. Knack, S. and P. Keefer. 1995. Institutions and Economic Performance: Cross-Country Tests UsingAlternative Institutional Measures. Economics and Politics 7(3): 207-227. Knight, et al. 2003. The Role o f Education in Facilitating Risk-taking and Innovation in Agriculture.Journal of Development Studies 39(6): [11-22. Kronlid, K. 2001. Household welfare and Education in Urban Ethiopia. World Institute for Development Economics Research, United Nation University. Wider discussion Paper (International) Number WDP 2001/144. 267 Kuma, T. 2002. Trends in Agricultural Production, Technology Dissemination, and Price Movements of Outputs and Inputs. Policy Forum on Agriculture Technology Diffusion and Price Policy, Ethiopian Development Research Institute (EDRI), International Food Policy Research Institute (IFPRI), and 2020 Vision Network for East Africa, Addis Ababa. Kydd, J., A. Donvard, J. Morisson, and G. Cadisch. 2001. The Role o f Agriculture in Pro- Poor Economic Growth in Sub Saharan Africa. Background paper prepared for DFID. ImperialCollege o f Science, Technology and Medicine at Wye. mimeo. Landell Mills. 2004. Evaluation o f the Water Harvesting Schemes Component o f the EC Funded Programmes IFSP 1998 and 2000 in Tigray Regional State. Final Evaluation Report. mimeo. Legovini, A. 2004. Preliminary Findings from a Study on `Measuring Women's Empowerment. 'World Bank, Addis Ababa. Lentz, E., and C. Barrett. 2004. Food aid targeting, shocks and private transfers among East Africanpastoralists. Mimeo. Li,H.,A. Stein, H.Barhhart, U.RamakrishnanandR.Martorell, 2003: Associations Between Prenatal and Postnatal Growth and Adult Body Size and Composition. American Journal of Clinical Nutrition, forthcoming. Lipton, M., and M. Ravallion. 1995. Poverty and Policy. in J. Behrman and T.N. Srinivasan (eds). 1995. Handbook of Development Economics Volume 3. Amsterdam: North- Holland. Lister, S. 2003. The Processes and Dynamics o f Pastoralist Representation in Ethiopia. Draft for Comment. Institute o f Development Studies, Sussex, UnitedKingdom. Lucas, R. 1976. Econometric Policy Evaluation: A Critique. Carnegie-Rochester Conference Series on Public Policy 1: 19-46. MacMillan, et al. 2003. Agriculture and Trade: Ethiopia Diagnostic Trade Integration Study, Vol2. Annex 8. World Bank: Washington D.C. Martorell, R. 1999. The Nature of Child Malnutrition and its Long-Term Implications. Food and Nutrition Bulletin 20: 288-292. Martorell, R. 1995. Results and Lmplications o f the INCAP Follow-up Study. Journal of Nutrition 125(Suppl): 1127s - 1138s. 268 Martorell, R., K. L. Khan, and D.G. Schroeder. 1994. Reversibility o f Stunting: Epidemiological Findings in Children from Developing Countries. European Journal of Clinical Nutrition, 48(Suppl): S45457. Martorell, R., J. Rivera, and H. Kaplowitz. 1989. Consequences of Stunting in Early Childhood for Adult Body Size inRural Guatemala. StanfordUniversity, FoodResearch Institute, Stanford, CA, mimeo. McPeak, J. and C. Barrett. 2001. Differential risk exposure and stochastic poverty traps among East-African pastoralists. American Journal of Agricultural Economics 83(3): 674-679. Mellor, J., ed. 1995. Agriculture on the Road to Industrialization. Baltimore, MD: John Hopkins University Press. Miguel,E. and M.Kremer.2005. Worms: IdentifyingImpacts on Health and Education inthe Presence o f Treatment Externalities, Econometrica (forthcoming). Milas, S., and K. El Aynaoui. 2004. Four Ethiopia's: A Regional Characterization - Assessing Ethiopia's Growth Potential and Development Obstacles. Ethiopia Country Economic Memorandum. World Bank, Washington D.C. MinistryofFinance and Economic Development (MoFED), Ethiopia. 2002. Development and Poverty Profile o f Ethiopia: Analysis Based on the 1999/00 Household Income, Consumption and Expenditure and Welfare Monitoring Surveys. Federal Democratic Republic o f Ethiopia, Ministry o f Finance and Economic Development, Welfare MonitoringUnit,Addis Ababa. Okumu, B., et al. 2002. A Bio-Economic Model o f Integrated Crop-Livestock Farming Systems: The Case o f the Ginchi Watershed in Ethiopia. inBarrett, C., F. Place, and A. Aboud, eds. 2002. Natural Resources Management in African Agriculture: Understanding and Improving Current Practices. Wallingford, UK:CAB International. Pankhurst, A., and A. Gebre. 2003. The Current Understanding of Poverty and Wellbeing in Ethiopia. Draft Report. Universityo f Bath. Pender, J., et al. 2001. Strategies for Sustainable Agricultural Development in the Ethiopian Highlands. American Journal of Agricultural Economics 83(5): 1231-1240. Pender, J., and B. Gebremedhin. 2004. Impacts o f Policies and Technologies in Dryland Agriculture: Evidence from Northern Ethiopia. in Rao, S., ed. 2004. Challenges and Strategies for Dryland Agriculture. Special Publication 32. American Society o f Agronomy and Crop Science Society of America, Madison Wisconsin. 269 Polhemus H., and L. Yohannes. 2002. Situational Analysis of the Governance Sector in the Southern Nations Nationalities and Peoples Regional State and Tigray National Regional State o f the Federal Democratic Republic of Ethiopia. Pollitt, E. 1990. Malnutritionand Infectioninthe Classroom. UNESCO (Paris) Quisumbing, Agnes, Lawrence Haddad, and Christine Pena, 2002, Are Women Overrepresented among the Poor? An Analysis of Poverty in 10 Developing Countries, Journal of Development Economics, 66-1, pp. 225-69. Rahmato, Dessalegn. 1998. The Dynamics of Rural Poverty: Case Studiesfrom a District in Southern Ethiopia. East Lansing, MI:Michigan StateUniversity Press. Rahmato, Dessalegn, and Aklilu Kidanu. 1999. "Ethiopia: Consultations with the Poor." Poverty Reduction Group, PREM, World Bank, Washington, DC. Ravallion, M. 2003. Measuring aggregate welfare in developing countries: how well do national accounts and surveys agree? The Review of Economics and Statistics 85(3): 645-652. Reardon, T., et al. 2000. Effects of Non-Farm Employment on Rural Income Inequality in Developing Countries: An Investment Perspective. Journal of Agricultural Economics 51(2): 266-288. Reardon, T. 1998. Rural Non-Farm Income in Developing Countries, The State o f Food and Agriculture. Foodand Agriculture Organization, Rome. Reardon, T., et al. 1994. I s Income Diversification Agriculture-Ledinthe West African Semi- Arid Tropics? The Nature, Causes, Effects, Distribution and Production Linkages of Off-Farm Activities. in Atsain, W.S., and A.G. Drabek, eds. Economic Policy Experience in Africa: What Have We Learnt? Nairobi, Kenya: African Economic ResearchConsortium. Roose, E.J. 1996. Land husbandry: components and strategies. FA0 Soils Bulletin 70. Food and Agricultural Organization, Rome. Roose, E., and B. Barthes. 2001. Organic Matter Management for Soil Conservation and Productivity Restoration in Africa: A Contribution from Francophone Research. Nutrient Cycling in Agroecosystems 61(1-2): 159-170. Rosegrant, M. and P. Hazell. 2000. Transforming the Rural Asian Economy: The Unfinished Revolution. Oxford: Oxford University Press. Sahn, D., and D. Stifel. 2000. Poverty Comparisons Over Time and Across Countries in Africa. WorldDevelopment 28(12): 2123-2155. 270 Sen, A. 1985. Commodities and Capabilities. Amsterdam: North-Holland, Sharp, K., S. Devereux, and Y. Amare. 2003. Destitution in the North-Eastem Highlands (Amhara Region). Instituteo f Development Studies and Save the Children, Ethiopia. Shaxson, F. 1999. New concepts and approaches to land management in the tropics with emphasis on steeplands. FA0 Soils Bulletin 75. Food and Agricultural Organization, Rome. Shimeles, A. 2004. Trends inGrowth, Income Distribution and Poverty in Ethiopia: Evidence from Household Panel Data. Draft. Department o f Economics, University o f Goteborg, Sweden. SmithK.,C. Barrett, and P. Box. 2001. Not necessarily inthe same boat: Heterogeneous risk assessment among East-African pastoralists. Journal of Development Studies 37(5). Sodhi, A,, B. Manna, and H. Wadhawan. 1999. Agricultural Extension and Food Security. World Bank Food Security Mission. mimeo. Strauss, J., and D. Thomas, 1998: Health, Nutrition, and Economic Development, Journal of Economic Literature, 36(2): 766-817. Summers, R. and A. Heston. 1991. The Penn World Table (Mark 5): an expanded set of intemational comparisons, 1950-1988. Quarterly Journal of Economics 106(2): 327-68. Svedberg, P. 1990. Undernutrition in Sub-Saharan Africa: I s There a Gender Bias?, Journal of Development Studies 26(3): 469-486. Thomas, D., and J. Strauss. 1997. Health and Wages: Evidence on Men and Women inUrban Brazil. Papers 97-05, RAND - Reprint Series. Toulmin, C., et al. 2000. Diversification o f Livelihoods: Evidence from Ethiopia and Mali. ResearchReport 47. Institute o f Development Studies, Brighton, UnitedKingdom. UNOffice for the Coordination ofHuman Affairs (UNOCHA). 2002. Ethiopia: Focus on the Afar people. IRINnews.org, Addis Ababa. (October 15, 2003). Van den Broeck, K. 2004. Determinants o f Crop Output Change in Ethiopia between 1994 and 1997. Center for Studies o f African Economies, Oxford University, mimeo. Vaughn, S., and K. Tronvoll. 2003. The Culture o f Power in Contemporary Ethiopian Political Life. Sida Studies No. 10. The Swedish International Development Cooperation Agency, Stockholm, Sweden. 271 Venvimp, P. 1999. Estimating Retums to Education in Off-Farm Activities inRural Ethiopia. Discussion Paper 99.03. Center for Economic Studies, Department o f Economics, Catholic University o f Leuven, Belgium. Wax, E. 2004. "Ethiopian Rape Victim Pits Law Against Culture." The Washington Post. Washington, DC. June 7, 2004, page Al. Weber, M.T., et al. 1988. Informing Food Security Decisions in Africa: Empirical Analysis and Policy Dialogue. American Journal of Agricultural Economics 70 (December): 1044-1052. Weir T. 1999. The Effects o f Education on Farmer Productivity in Rural Ethiopia, Center for the Study of African Economies. Working Paper Series No. WPS/99-7: 1-50. Weir, S. and J. Knight. 2000. Adoption and Diffusiono f Agricultural Innovations inEthiopia: The Role o f Education. Oxford University Center for Studies o f African Economies Working Paper 2000-5. World Bank. 2005 (forthcoming). Ethiopia: RuralEconomic Sector Work. Washington, DC. World Bank. 2004 (draft). "Making the Connection between Power and Poverty: Ethiopia Case Study." Poverty Reduction Group, PREM Network, Washington, DC. World Bank. 2004a. Education in Ethiopia: Strengthening the Foundation for Sustainable Progress. Report No. 28037-ET. Discussion Draft of Education Country Sector Report, February2004. World Bank, Washington, DC. World Bank. 2004b. Ethiopia: A Country Status Report on Health and Poverty. Draft Report No.28963-ET, September 2004. The World Bank, Africa Region Human Development,and Ministry o fHealth, Ethiopia. World Bank. 2004c. Ethiopia: A Strategy to Balance and Stimulate Growth. A Country Economic Memorandum. Poverty Reduction and Economic Management 2 (AFTP2), Africa Region, World Bank, Washington, DC. World Bank. 2004d. Ethiopia: Draft Budget Update-2004/05 FY.BackgroundNote for 2004 Joint Budget and Aid Review. A WB/DBS Joint Assessment. Draft, August 2004. World Bank. World Bank. 2004e. Ethiopia: Risk and Vulnerability Assessment. Report No. 26275-ET. World Bank, Human Resources Development Group 111. AFCO6. Africa Region. Washington, DC. World Bank. 2004f. SDStats Country Summary: Ethiopia. April 6. Social Development Department,SDStats, Washington, DC. 272 World Bank. 2004g. WorldDevelopment Indicators 2004. World Bank: Washington, D.C. World Bank. 2004h. World Development Report 2004: Making Services Work for Poor People. Washington, DC: The World Bank. World Bank, 2003a. Ethiopia: Public Expenditure Review - Volume 111: Medium-Tenn Trends and Recent Developments in Public Spending. AFTP2, Africa Region, Washington, DC. World Bank. 2003b. Memorandum of the President of the Intemational Development Association to the Executive Directors on a Country Assistance Strategy for the Federal Democratic Republic of Ethiopia. Report No. 25591-ET. Country Department 6, Africa Region, Washington, DC. World Bank. 2002a. Ethiopia: Summary Gender Profile. GenderStats. Washington, DC. World Bank. 2002b. World Development Indicators 2002. World Bank: Washington, DC. World Bank. 2001. Promotion of Small-scale Irrigation in Food Insecure Woredas O f Ethiopia. Washington, DC. World Bank. 2000. World Development Report 2000/20001: Attacking Poverty. Washington, DC: The World Bank. World Bank. 1999. Ethiopia: Poverty and Policies for the New Millennium. Report No. 19804-ET. World Bank, Africa Region, Macroeconomics 2, Washington D.C. World Bank Country Office in Ethiopia. 2002. Ethiopia: The Woreda Studies. Volume 1: The Study. Country Department 6, Africa Region, Washington, DC. Xiao, Y., and A. Gelb. 2004. Why has Ethiopia not seen a poverty reduction despite GDP growth inrecent years?DiscussionNote. World Bank. mimeo. Yamano, T., H., Alderman, and L, Christiaensen, 2005, Child Growth, Shocks and Food Aid inRuralEtiopia, American Joumal ofAgricultural Economics, 87(2): 273-88. Zala-Daget Project. 2004. Lessons leamt from 10 years of research on soil erosion and water conservation inTigray. Extension Manual. mimeo. Zewdie, B. and Pausewang, S., (eds). 2002. Ethiopia: The Challenge of Democracy from Below. Ethiopia: ForumFor Social Studies. 273