Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized An Assessment of Drivers, Constraints and Opportunities Overcoming Poverty and Inequality in South Africa OVERCOMING POVERTY AND INEQUALITY IN SOUTH AFRICA An Assessment of Drivers, Constraints and Opportunities March 2018 © 2018 International Bank for Reconstruction and RIGHTS AND PERMISSIONS Development / The World Bank The material in this work is subject to copyright. Because 1818 H Street NW The World Bank encourages dissemination of its knowledge, Washington DC 20433 this work may be reproduced, in whole or in part, for Telephone: 202-473-1000 noncommercial purposes as long as full attribution to this work is given. Internet: www.worldbank.org Any queries on rights and licenses, including subsidiary This work is a product of the staff of The World Bank in rights, should be addressed to World Bank Publications, close collaboration with the National Planning Commission The World Bank Group, 1818 H Street NW, Washington, Secretariat at the Department of Planning, Monitoring DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@ and Evaluation, and Statistics South Africa. The findings, worldbank.org. interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, Front and back cover: Kya Sands/Bloubosrand, its Board of Executive Directors, or the governments they Johannesburg, South Africa @Johnny Miller / Unequal represent. The World Bank does not guarantee the accuracy Scenes of the data included in this work. The boundaries, colors, The report was edited by William B. Hurlbut and designed denominations, and other information shown on any map and typeset by Shereno Printers, Gauteng, South Africa. in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. An Assessment of Drivers, Constraints and Opportunities iii ABBREVIATIONS AND ACRONYMS AfDB African Development Bank HOI Human Opportunity Index AsgiSA Accelerated and Shared Growth Initiative for IES Income and Expenditure Survey South ILO International Labor Organization B-BBEE Broad-Based Black Economic Empowerment ISRDP Integrated Sustainable Rural Development BCEA Basic Conditions of Employment Act Program BRICS Brazil, Russia, India, China and South Africa LBPL Lower bound poverty line CGE Computable General Equilibrium LCS Living Conditions Survey CPI Consumer Price Indexes LRA Labour Relations Act CSG Child Support Grant MPI Multidimensional Poverty Index CSP Community, social, and public NDP National Development Plan DAFF Department of Agriculture, Forestry and NIDS National Income Dynamics Study Fisheries NMW National minimum wage DG Disability Grant OAG Old Age Grant DPME Department of Planning, Monitoring and QLFS Quarterly Labour Force Survey Evaluation RIF Recentered Influence Functions EPWP Expanded Public Works Programme SAMPI South African Multidimensional Poverty ETI Employment Tax Incentive Index FPL Food Poverty Line SASSA South Africa Social Security Agency GDP Gross domestic product SD Sectoral Determination (of wages) GEAR Growth, Employment and Redistribution SMME Small, micro, and medium enterprises GHS General Household Survey Stats SA Statistics South Africa GIC Growth Incidence Curves TES Temporary employment services GNI Gross National Income UBPL Upper bound poverty line HFIAS Household Food Insecurity Access Scale UNDP United Nations Development Programme HIV Human Immuno-Deficiency Virus WDI World Development Indicators iv Overcoming Poverty and Inequality in South Africa CONTENTS Contents iv i. Access to basic services and utilities 20 Figures v ii. Housing conditions, access to education, 22 health, and assets Tables ix iii. Food security and malnutrition 24 Boxes ix iv. The South African Multidimensional Poverty 28 Foreword x Index Preface xii v. Changes in multidimensional poverty at the 29 Acknowledgements xiii national level Executive Summary xiv vi. Multidimensional Poverty Index, headcount 31 CHAPTER 1: INTRODUCTION 1 and intensity: spatial variation CHAPTER 2: EVOLUTION, DIMENSIONS AND 6 vii. Multidimensional deprivation 33 DYNAMICS OF POVERTY IN SOUTH AFRICA E. Economic mobility: transitioning from chronic 33 A. Despite progress in reducing poverty since 7 poverty to middle class 1994, poverty rates remain high for an upper i. Poverty transitions, chronic poverty, and 34 middle-income country characteristics i. Trends in national poverty 7 ii. The scope of social classes in South Africa 35 ii. International poverty trends 11 iii. The profile of social classes and factors 36 B. Who are the poor? 13 associated with escaping chronic poverty C. Where do the poor live? 16 CHAPTER 3: SOUTH AFRICA IS ONE OF THE 42 MOST UNEQUAL COUNTRIES IN THE WORLD i. Variation in poverty across provinces 16 A. Consumption inequality is very high and has 43 ii. Variation in poverty across municipalities 17 increased since the end of apartheid D. Notable progress has been made in reducing 20 B. High level of inequality of opportunity 45 multidimensional poverty since the end of apartheid in 1994 An Assessment of Drivers, Constraints and Opportunities v i. Extent of inequality of opportunity 45 G. Unions appear to distort labor supply but offer 91 substantially higher wages ii. Human Opportunity Index in South Africa 46 H. High reservation wages and very high wage 94 C. Wage inequality is very high and is 49 disparities compounded by heavy polarization between two extremes I. Labor factors affecting transitions into and out 95 of poverty—result of panel analysis i. Trends and causes of wage inequality 49 CHAPTER 6: GOOD JOBS ARE THE KEY TO 99 D. Wealth inequality is very high, even higher 51 FUTURE REDUCTIONS IN POVERTY AND than income inequality INEQUALITY E. Low intergenerational mobility is an obstacle 53 A. Projecting poverty reduction through 2030 99 to inequality reduction B. Policy interventions to gain further poverty 101 F. South Africa lags its peers on inclusiveness of 56 and inequality reduction consumption growth C. Distributional impact of labor market policies 104 i. Incidence of growth 56 and legal institutional changes in recent years G. Inequality slows down poverty reduction 58 i. The Labour Relations Amendment Act of 2014, 104 CHAPTER 4: DRIVERS OF POVERTY AND 61 labor brokering, temporary employment INEQUALITY IN SOUTH AFRICA services A. What drives changes in poverty in South 61 ii. The Employment Tax Incentive 105 Africa? iii. Expected poverty impact of national 106 B. What drives changes in inequality and 66 minimum wage legislation intergenerational mobility in South Africa? D. Future policy measures that could help reduce 109 i. Drivers of inequality of consumption 66 poverty and inequality ii. What drives intergenerational mobility? 67 REFERENCES 114 C. Achieving a more equitable society through 69 efficient social protection FIGURES CHAPTER 5: LABOR MARKET DYNAMICS AND 76 POVERTY Figure 1: Long-term trends in inequality, xv A. Dynamics and challenges in labor market 76 comparison to other countries outcomes Figure 2: Growth incidence of consumption, xv B. Explaining labor market participation and 81 expenditures by percentile, 2006–2015 employment Figure 3: Shared prosperity indicator in selected xvi C. Structural mismatch between labor demand 84 countries (2007–2014) and labor supply for unskilled workers Figure 4: Real monthly wage by percentile, xvii D. Racial and demographic factors defining 86 average annualized percentage change, 1994– employment 2014 E. Geographical segregation and role of internal 87 Figure 5: Real wage inequality, 1995–2014 xvii migration Figure 6: Overall changes in national poverty rates, xx F. Diminished role of small, medium, and micro 88 lower-bound poverty lines enterprises in employment generation Figure 7: Long-term trends in $1.9/day xx international poverty rates vi Overcoming Poverty and Inequality in South Africa Figure 8: Overall changes in international poverty xx Figure 26: Poverty headcount ratio by individual 14 rates, comparison to other upper middle-income characteristics countries Figure 27: Age-gender pyramid and poverty, 2015 15 Figure 9: Changes in the proportion of the xxi Figure 28: Poverty headcount ratio by province 16 population with access to selected basic services Figure 29: Regional poverty decomposition, 2006 16 Figure 10: The proportion of the population with xxi to 2015 access to electricity, comparison to other countries, 2014 Figure 30: Poverty incidence at the municipality 18 level Figure 11: The proportion of the population with xxi access to an improved water source, comparison Figure 31: Poverty density at the municipality level 18 to other countries, 2015 Figure 32: Comparison of municipality poverty 19 Figure 12: The proportion of the population with xxi rates, 1996 and 2011 access to improved sanitation facilities, comparison Figure 33: Dispersion and range in municipality 19 to other countries, 2015 poverty rates, 1996 and 2011 Figure 13: Poverty headcount ratio by xxii Figure 34: Changes in the proportion of the 21 characteristics of head of household population with access to selected basic services Figure 14: Poverty headcount ratio by individual xxiii Figure 35: The proportion of the population with 21 characteristics access to electricity, comparison to other countries, Figure 15: Poverty incidence at the municipality xxiv 2014 level Figure 36: The proportion of the population with 21 Figure 16: Multidimensional poverty headcount xxiv access to an improved water source, comparison ratio at the municipality level, the 20 poorest to other countries, 2015 municipalities Figure 37: The proportion of the population with 21 Figure 17: Real GDP growth decomposition 2 access to improved sanitation facilities, comparison to other countries, 2015 Figure 18: Economic structure of South Africa 2 (share of GDP, supply side) Figure 38: The proportion of the population with 22 access to electricity, by decile, 2015 Figure 19: Average labor productivity 2 decomposition (contributions to labor productivity Figure 39: The proportion of the population with 22 growth) access to an improved water source, by decile, 2015 Figure 20: Overall changes in poverty rates 7 Figure 40: The proportion of the population with 22 Figure 21: Long-term trends in US$1.9/day 12 access to improved sanitation facilities, by decile, international poverty rates 2015 Figure 22: Overall changes in US$1.9/day 12 Figure 41: Overcrowding headcount rate, by 24 international poverty rates decile, 2015 Figure 23: Overall changes in international 12 Figure 42: The proportion of the population older 24 poverty rates, comparison to other countries than 25 with primary school education, by decile, Figure 24: Overall changes in international 12 2015 poverty rates, comparison to other upper middle- Figure 43: The proportion of the population for 24 income countries whom distance to nearest hospital is at least 20 Figure 25: Poverty headcount ratio by 13 kilometers, by decile, 2015 characteristics of head of household Figure 44: Asset ownership, by decile, 2015 24 An Assessment of Drivers, Constraints and Opportunities vii Figure 45: Food security index by household 25 Figure 69: Human Opportunity Index and D-index 47 characteristics of inequality of opportunity, 2015 Figure 46: Food insecurity index by quintiles of 26 Figure 70: Change in the HOI andf decomposition 48 asset index (percent) of changes, 2002-15 Figure 47: Gender disaggregated stunting rates in 26 Figure 71: Contribution circumstances to D-index, 48 children under five 2015 Figure 48: Contribution of weighted indicators to 30 Figure 72: Wage inequality 49 SAMPI at national level Figure 73: Average wages by gourps 50 Figure 49: Multidimensional poverty measures at 31 Figure 74: Group share in the sample 50 provincial level Figure 75: Real monthly wage by percentile, 50 Figure 50: Poorest and richest districts and local 32 average annualized percentage change 1994-2014 municipalities in South Africa in 2016 Figure 76: Real wage inequality, 1995-2014 50 Figure 51: Multidimensional poverty headcount 32 ratio at the municipality level Figure 77: Households wealth inequality, Gini 52 coefficients across countries Figure 52: Deprivations affecting the poor in 2015 33 Figure 78: The share of household wealth held by 52 Figure 53: Poverty duration, 2008–2015 35 the percentiles in the distribution Figure 54: Income source by duration in poverty 35 Figure 79: Composition of wealth by income 52 Figure 55: Class sizes, 2008–2014/15 36 group Figure 56: Income by sources, classes 36 Figure 80: Correlates of households’ income and 52 wealth, coefficients from regression analysis Figure 57: Geographic distribution of South 37 Africa’s five social classes, 2008–2014/15 Figure 81: The relationship between 54 intergenerational mobility and inequality Figure 58: Pockets of high propensity to poverty 38 in South Africa, 2014/15 Figure 82: Intergenerational elasticities at various 55 percentiles of father’s income Figure 59: Racial composition of South Africa’s five 39 social classes, 2008 and 2014/15 Figure 83: Growth incidence curves, national 56 Figure 60: South Africa’s five social classes in the 40 Figure 84: Growth incidence curves 2006–2015, 57 labor market, 2008–2014/15 urban and rural Figure 61: Long-term trends in inequality, 43 Figure 85: Shared prosperity indicator in selected 58 comparison to other countries countries (2007–2014) Figure 62: Polarization indexes across countries 43 Figure 86: Decomposing changes in the poverty 59 headcount ratio into growth and redistribution Figure 63: Growth incidence of consumption 44 expenditures by percentile, 2006 to 2015 Figure 87: Decomposing changes in poverty into 60 growth and redistribution, 2006–2015, poverty gap Figure 64: Consumption shares over time 44 and squared poverty gap Figure 65: Changes in income shares by source 45 Figure 88: Contribution to poverty reduction by 63 Figure 66: Income shares over time 45 income sources over 2006–2015 Figure 67: Inequality of opportunity, cross-country 46 Figure 89: Endowments and Returns. The 64 estimates contribution of demographics, location of Figure 68: Decomposition of the inequality of 46 residence, education, access to services and labor opportunity into constituent factors to consumption growth, in %, LCS 2004/05– 2014/15 viii Overcoming Poverty and Inequality in South Africa Figure 90: Causes of welfare changes, 2006–2015, 65 Figure 111: The large firm premiums 91 in percent Figure 112: Trade union membership of formal 92 Figure 91: Factor wise contribution to inequality 66 sector employees by public and private sector (Theil-L Measure) status, selected years Figure 92: Decomposition of inequality by 67 Figure 113: Percentile distribution of log wages by 92 contributing factors union status and public/non-public sector status, 2014 Figure 93: Inequality by income sources 67 Figure 114: Union restrict supply but raise wages 93 Figure 94: Spending on social assistance as 71 percent of GDP Figure 115: Average wages and transfers 94 Figure 95: Real expenditure on social grants, 72 Figure 116: Returns from Mincer regression 94 2005/06–2015/16 Figure 117: Marginal effects for transitioning into 96 Figure 96: Social assistance coverage rates across 72 poverty quintiles Figure 118: Moving out of poverty: contributing 101 Figure 97: Simulated poverty reduction associated 73 factors with social assistance programs Figure 119: Change in poverty due to 102 Figure 98: Simulated inequality reduction 74 employment generation associated with social assistance programs Figure 120: Change in the Gini coefficient due to 102 Figure 99: Key labor market trends 2000–2016 77 employment generation Figure 100: Labor force participation rates, 77 Figure 121: Changes in simulated poverty rates 103 unemployment, and dependency ratios, by due to increase in total wages, all economy and country (selected years) beneficiaries Figure 101: Trends in South African employment 78 Figure 122: Percent reduction in poverty rates 103 following 10 percent wages growth Figure 102: Sectoral gross value-added and 80 employment growth, 2000–2016 Figure 123: ETI eligible and supported jobs by 106 sector Figure 103: Growth of employment shares by 80 sector and skills level, percent share: 1995–2015 Figure 124: Earning bands by sector (2015 rand) 107 Figure 104: Composition of employment by 80 Figure 125: Ratio of NMW to lowest and highest 107 sector and skills level, percent share: 2015 SD wages Figure 105: Determinants of labor force 81 Figure 126: First order effect: impact of projected 109 participation outcome, marginal effects for minimum wage legislation on poverty and selected years inequality Figure 106: Probability of services sector 83 Figure 127: First order effect: impact of projected 109 employment, individual effects: 1994–2015 minimum wage legislation on income, by decile Figure 107: Skill mismatch 86 Figure 128: Elasticity of poverty to consumption 111 growth, 2014/15 Figure 108: A gender gap holds except for low- 87 skill jobs Figure 129: Elasticity of poverty to consumption 111 growth, 2005–15 Figure 109: Urban wage differentials and formal 88 sector wages Figure 130: Potential impact of selected NDP 112 reforms on GDP growth Figure 110: Employment probabilities, comparing 90 small and large firms Figure 131: Projected impact of the policies on 112 poverty and social indicators An Assessment of Drivers, Constraints and Opportunities ix TABLES Box 9: Elements of the South African social security 70 framework Table 1: Inflation-adjusted poverty lines, 2006–2017 8 Box 10: What does it take for an individual to 82 (per person per month in South African Rands) obtain a job in the fast-growing services sector? Table 2: Changes in the depth and severity of 10 Box 11: Policy, legal, and institutional changes 104 poverty Box 12: Application national minimum wage 107 Table 3: SAMPI dimensions, indicators, and 29 Box 13: Growth to poverty elasticity in South Africa 111 deprivation cut-off points Table 4: Multidimensional poverty at national level 30 Table 5: Poverty transition matrices for South Africa, 34 2008-2014/15 (pooled 4 waves panel) Table 6: Frequencies of transition across income 55 quintiles (multiple imputation estimates) Table 7: Summary of regression results—upward 68 mobility Table 8: Elements of the South African social 70 security framework Table 9: Determinants of labor force participation 85 and employment transitions Table 10: Projected poverty and inequality rates— 100 baseline scenario Table 11: Amendments to the Labour Relations Act 104 BOXES Box 1: The methodology of poverty measurement 8 in South Africa Box 2: Estimating poverty at the municipality level 19 Box 3: Construction of an asset index and the 25 Household Food Insecurity Access Scale Box 4: The Alkire-Foster method 28 Box 5: Estimating chronic and transient poverty 34 Box 6: Defining the scope of middle class in South 36 Africa Box 7: Intergenerational mobility in South Africa 53 Box 8: Three methods for decomposing changes in 62 poverty x Overcoming Poverty and Inequality in South Africa FOREWORD This report is an analysis of South Africa’s progress in Government is committed to eliminating poverty, and fiscal reducing poverty and inequality since 1994, with 2006 policy is one critical lever that expresses this commitment. to 2015 as a reference period. Its aim is to understand The equitable share formula used to determine transfers the dynamics of poverty and inequality in the country, to to provincial and local spheres of government contains a identify the drivers of progress for the purpose of further poverty component as a redistributive measure. The ‘social policy actions in this area. wage’ has been used as a redistributive mechanism of the government budget deliberately aimed at improving the Reducing poverty and inequality is the overriding concern lives of the poor and reducing their cost of living. This has of South Africa’s development policies and programs, from been achieved through, among others, free primary health the onset of our democracy in 1994 in the Reconstruction care; no-fee paying schools; old age and child support and Development Programme (RDP) to the current grants; housing; and free basic services (water, electricity National Development Plan: Vision 2030 (NDP). The guiding and sanitation) to poor households. Although these principle, as captured in the NDP, is that “no political policies and interventions have resulted in notable gains democracy can survive and flourish if the mass of our people in poverty reduction since 1994, the country continues remain in poverty, without land, without tangible prospects to face the challenge of high poverty, high inequality and for a better life. Attacking poverty and deprivation must be high unemployment. The persistence of these challenges the first priority of a democratic government”. The NDP posits calls for a rigorous assessment of the drivers, constraints that to raise the living standards to the minimum required and opportunities for poverty and inequality reduction in level will involve various mechanisms, such as increasing South Africa. employment, incomes, productivity as well as through social protection and quality public services. The measure The report shows that, overall, poverty levels are lower of success of government’s development policies will be today compared to 1994. Relatively high and consistent when the lives and opportunities of poorest South Africans economic growth following the end of apartheid in 1994 are transformed for the better. up to around 2011 supported poverty reduction in South Africa, although economic growth prospects have been An Assessment of Drivers, Constraints and Opportunities xi slowing in recent years. The economy is currently not The report highlights the growing importance of generating sufficient jobs, and the unemployment rate was education (skills) and labor market outcomes in supporting 27.7 percent in the third quarter of 2017. Youth and unskilled the country’s poverty and inequality reduction agenda. workers bear the brunt of the problem as employers seek Creating more jobs in an inclusive manner is thus important skilled workers, and the youth unemployment rate was for the realization of the NDP’s vision of eliminating poverty 38,6 percent. As a result, poverty rates increased between and reducing inequality. 2011 and 2015. This experience is a reminder of the reality We hope that the report makes a valuable contribution that the country’s socio-economic challenges are deep, to this quest for effective strategies against poverty and structural and long-term. This report is therefore timely as inequality in the country, as part of national development we, as a country, continue to grapple with these challenges planning and poverty monitoring activities, and building and seek pathways to sustainable solutions, guided by the on existing work and knowledge. We would like to express NDP. our gratitude to the National Planning Commission While the long-term trend indicates progress in reducing Secretariat at the Department of Planning, Monitoring and poverty, inequality has remained stubbornly high. The Evaluation, Statistics South Africa, and the World Bank for report reveals South Africa as one of the most unequal their collaborative efforts in undertaking this study. countries in the world, with consumption inequality having increased since 1994. Wealth inequality is high and has been rising over time. A polarized labor market results in Dr Nkosazana C. Dlamini-Zuma, MP high wage inequality. Intergenerational mobility is relatively Minister in the Presidency: Planning, Monitoring and low and serves as a barrier to inequality reduction. Evaluation xii Overcoming Poverty and Inequality in South Africa PREFACE I am pleased to present the Overcoming Poverty and South Africa. South Africa has a dual economy where on the Inequality in South Africa: An Assessment of Drivers, one hand is a small high-skilled, high-productivity economy Constraints and Opportunities. This study was prepared and on the other hand, a large low-skilled, low-productivity jointly by the National Planning Commission Secretariat one. This assessment argues that it is this duality that has in at the Department of Planning, Monitoring and Evaluation part resulted in high wage inequality that has been steadily (DPME), Statistics South Africa, and the World Bank. It goes rising reflecting a highly polarized labor market. to the heart of South Africa’s major challenges of poverty This study reveals that labor market incomes are the largest and inequality which, together with unemployment, are contributor to inequality in South Africa, contributing more identified in the National Development Plan (NDP) as the than 90 percent of the overall Gini coefficient between triple challenge that is to be overcome by 2030. In this 2006 and 2015. If also finds that the nature of inequality regard, this report is also aligned to the World Bank Group’s has changed with the role of skills and labor market twin goals of assisting countries in their efforts to end factors having grown in importance in explaining poverty extreme poverty by 2030 and promote shared prosperity. and inequality while that of gender and race, though still The Government of South Africa, supported by economic important, has declined presenting an opportunity for gains made since 1994, has made significant progress policy to influence poverty and inequality outcomes. It in reducing poverty, improving access to basic services, shows that access to higher levels of education and stable education, health care, social protection, and economic labor market income are key determinants for households opportunities which have helped in reversing some of the to achieve economic stability in South Africa. Social adverse effects of a system of segregation under apartheid. protection remains important in reducing extreme poverty, However, this progress is being undermined by the but the fiscal space for further expansion is limited. country’s recent low economic growth prospects. The report identifies unlocking the full potential of labor The triple challenge of high poverty, high inequality, and markets and promoting inclusive growth through skills high unemployment persists. Poverty remains high for an creation among possible areas of intervention that will upper middle-income country with more than half (55 accelerate poverty and inequality reduction. It also argues percent) of the population of South Africa being poor at the that interventions that simultaneously stimulate growth national upper bound poverty line of ZAR 992 per person and reduce inequalities are likely to have much more per month in 2015 prices. In addition, with a consumption impact than interventions that only stimulate growth or per capita Gini coefficient of 0.63 in 2015, South Africa is one only reduce inequalities. of the most unequal countries in the world. Furthermore, As the country grapples with the triple challenges, it is my unemployment reached 25.1 percent of the workforce in hope that this evidence-based analysis will enhance our 2015 and was 27.7 percent in the third quarter of 2017. understanding of the drivers of inequality and barriers to This makes overcoming these challenges very complex, its reduction and that it will add to the ongoing public exacerbated by an environment of low growth which has debates on policies that are suitable and effective to tackle not generated sufficient jobs. poverty, inequality and unemployment in South Africa. This study offers a comprehensive assessment of the extent and causes of poverty and inequality in South Africa. The last such assessment was done in 1998. The Overcoming Paul Noumba Um Poverty and Inequality in South Africa: An Assessment of Country Director for South Africa Drivers, Constraints and Opportunities report focuses on the World Bank role of labor markets in reducing poverty and inequality in An Assessment of Drivers, Constraints and Opportunities xiii ACKNOWLEDGEMENTS This report was prepared by the World Bank jointly with The report benefited from comments and feedback from the National Planning Commission Secretariat at the participants at various stakeholder consultation workshops. Department of Planning, Monitoring and Evaluation (DPME) Three workshops were held at the inception stage to and the Poverty and Inequality Statistics Unit at Statistics present and get feedback on the Discussion Note: the first South Africa. The World Bank team comprised Victor Sulla was with government officials and organized by DPME; (co-task team leader), Precious Zikhali (co-task team leader) the second was organized by Fiona Tregenna and held Nga Thi Viet Nguyen (Poverty and Equity Global Practice), at the University of Johannesburg, and the last was held Sebastien Dessus (Program Leader, AFSC1), Marek Hanusch at the University of Cape Town, organized by Julian May (Macroeconomics, Trade and Investment Global Practice), and Murray Leibbrandt. The team would like to thank the and Kanishka Kacker (Consultant). The core team from organizers and participants for their invaluable comments the DPME comprised Mthokozisi Tshuma, Lusanda Batala, and insights. and Ziphezinhle Mzobe who made the collaboration a A working group, set up and coordinated by the DPME, success through efficient coordination, planning of various served as a platform for technical and policy guidance initiatives and technical expertise provided on the content to the study. The team would like to express gratitude to of the study. The core team from Statistics South Africa the following government departments that nominated comprised the entire Poverty and Inequality Statistics Unit officials to be part of this working group: the Department team. of Social Development, National Treasury, Economic The following consultants and World Bank staff produced Development Department, Department of Trade and technical background papers to the report: Carel van Aardt, Industry, Department of Higher Education and Training, Zaakhir Asmal, Bernadene de Clercq, Haroon Bhorat, Arden as well as the Department of Agriculture, Fisheries and Jeremy Finn, Coretta Jonah, Safia Khan, Murray Leibbrandt, Forestry. Throughout the process, the working group Indira Bongisa Lekezwa, Kezia Lilenstein, Julian May, Cecil provided invaluable insights at various stages of the study. Mlatsheni, Morné Oosthuizen, Dan Pavelesku, Ericka Rascon, The peer reviewers for the report were Thomas Farole (Lead Jamele Rigolini, Simone Schotte, Johann van Tonder, Economist), Emmanuel Skoufias (Lead Economist), and Kirsten van der Zee, and Rocco Zizzamia. Special thanks to Nobuo Yoshida (Lead Economist). Constructive comments Julian May who provided invaluable contributions at the and suggestions were provided by Rob Swinkels, Emmanuel inception stage of the work by co-authoring a Discussion Noubissie, Rose Mungai, John Gabriel Goddard, Arden Note that was used to kick-start consultations with relevant Jeremy Finn, Zandile Ratshitanga, and Jamele Rigolini. stakeholders to inform the scope and focus of the report. The report benefited from insights from the South Africa The team would like to express their gratitude to the Systematic Country Diagnostics (SCD) work led by Marek management of all the three institutions for their support Hanusch. and leadership throughout the study. The support of DPME Logistical assistance in the preparation of this report was management, Tshediso Matona, Kefiloe Masiteng, and ably provided by Santosh Kumar Sahoo, Mokgabo Molibeli, Khulekani Mathe (who has since left DPME but was present and Siele Shifferaw Ketema. Communications support was at the start of the project) is greatly appreciated. Within the provided by Zandile Ratshitanga from the World Bank side World Bank, the report was undertaken under the guidance and the Communications team from DPME. Last but not and leadership of Andrew Dabalen (Practice Manager), least, the team would like to thank everyone at DPME, Pierella Paci (Practice Manager), Paul Noumba Um (Country Statistics South Africa, and the World Bank who contributed Director), and Sebastien Dessus (Program Leader). to making this a truly collaborative effort. Thank you. xiv Overcoming Poverty and Inequality in South Africa EXECUTIVE SUMMARY For more than two decades, South Africa has sought Despite extremely high and rising unemployment, skilled to address poverty and inequality with a wide range of labor can be difficult to find in most skilled and professional initiatives, including the use of fiscal policy to support segments largely due to the poor state of the public redistributive measures. The social wage – which refers to education system. Yet education has a strong influence the government’s investment in education, health services, on the probability of labor market participation. Fourth, social development including social assistance to vulnerable location matters for labor market outcomes, with people households and individuals as well as contributory social in urban areas having better prospects of getting a job and security, public transport, housing, and local amenities as a higher probability of getting a formal job, but there are a redistributive measure – has played a notable role in the no significant differences across provinces. Location has government’s efforts to reduce poverty and inequality. implications on the travel costs which tend to be a burden These efforts can be traced back to the 1993 Reconstruction for getting jobs. The unemployed, and especially the youth, and Development Program, the first prescription of the post- tend to lack resources and mobility for a job search or ability apartheid era, which identified the reduction of poverty as to relocate as jobs could be located far. In some cases, a central goal. Other policies have continued that effort and underdeveloped transport, high cost of commuting and the most recent of these, the National Development Plan crime makes job search more difficult and raise associated 2030: Our Future—Make It Work (2012), seeks to eliminate expenses and reservation wages. Fifth, labor market poverty and reduce inequality and identifies the triple institutions and a rigid regulatory environment are shown challenge of high poverty, inequality, and unemployment to contribute to high levels of unemployment and wage as a major challenge for the country. The persistence of disparities. Sixth, Small Micro and Medium Enterprises these challenges, more than two decades after the end (SMMEs) have been struggling to advance inclusive growth of apartheid, calls for a comprehensive assessment of the and development as envisaged in the country’s NDP: the extent and causes of poverty and inequality, with attention share of SMMEs has been falling over time as well as the to trends, drivers, dynamics, policy, impact, and monitoring. proportion of employees working in this sector. All these challenges slow the ability of labor markets to accelerate High unemployment remains the key challenge for poverty and inequality reduction. Overcoming these South Africa and the country struggles to generate challenges is critical given that unemployment has an sufficient jobs. The labor market is characterized by adverse impact on poverty and inequality. Unemployment several challenges. These include, among others, first, high rates tend to be higher among the poor. Similarly, labor force level of unemployment which reached 25.1 percent of the participation is lower in poor than non-poor households. workforce in 2015 and 27.7 percent in the third quarter of 2017 associated with slow job creation as economic growth This report documents the progress South Africa has slowed in recent years. Second, racial and gender disparities made in reducing poverty and inequality since 1994, are still predominant in South Africa’s labor market, an with a focus on the period between 2006 and 2015. It enduring legacy of apartheid. Race still affects the ability aims to enhance understanding of the drivers of inequality to find a job, as well as the wages received once employed. and barriers to its reduction in South Africa, with a focus Although an increased number of women participate in on the role of labor markets. It also identifies possible areas South Africa’s economy, female participants find it harder of intervention that will accelerate poverty and inequality to find a job, and earn less than men when they do. Third, reduction. The focus on labor markets is important given there is strong evidence of structural mismatch between the persistently high unemployment in South Africa and labor demand and labor supply for unskilled workers. the consequent impact that has on poverty and inequality. An Assessment of Drivers, Constraints and Opportunities xv BY ANY MEASURE, SOUTH AFRICA IS ONE Conditions Survey 2014/15 found that the country had a OF THE MOST UNEQUAL COUNTRIES IN THE Gini coefficient of 0.63 in 2015, the highest in the world WORLD and an increase since 1994 (Figure 1). Further analysis of consumption expenditure trends provides evidence that Consumption expenditure data show that South the very poor—those in the bottom 10 percent—grew at a Africa is one of the most unequal countries in the slower pace than the rest of the population between 2006 world, and that inequality has increased since the and 2015 (Figure 2). end of apartheid in 1994.1 Analysis of the distribution of consumption expenditure per capita in the recent Living 1 It is important to note the differences in the Gini coefficients present- ed in this report and those presented in Statistics South Africa (2017). While both estimates are based on the same data, Stats SA uses dif- ferent welfare aggregates for poverty and inequality estimates. The per capita welfare measure used for poverty measurement includes all food items while for non-food items, large-sized, or “lumpy, durable goods” are excluded to reduce their biasing factor in the monthly es- timates. For inequality measurement, total consumption expenditure (including components that are excluded in the welfare aggregate used for poverty measurement), in per capita terms, is used. This re- port uses the same per capita welfare aggregate for both poverty and inequality measurement, and it is the one that excludes some com- ponents of consumption. This allows for comparison across countries, as most countries tend to use the same per capita welfare aggregate for poverty and inequality estimates. Figure 1: Long-term trends in inequality, comparison Figure 2: Growth incidence of consumption, to other countries expenditures by percentile, 2006–2015 Source: South Africa: authors’ calculations based on the Income Source: Authors’ calculations based on the Income and Expenditure and Expenditure Surveys for 2005/06 and 2010/11 and the Living Surveys for 2005/06 and 2010/11 and the Living Conditions Surveys for Conditions Surveys for 2008/09 and 2014/15 and WDI for 1996. WDI for 2008/09 and 2014/15. the rest of the countries and regional estimates.. xvi Overcoming Poverty and Inequality in South Africa South Africa also lags its peers on the inclusiveness deceleration of 1.4 percent for the period between 2011 of consumption growth. Inclusiveness in this case is and 2015. This does not compare well with the median examined by comparing the rate of consumption growth for the world (3.9 percent) or, in the later period, with Sub- for the bottom 40 percent of the population to that of Saharan Africa (Figure 3). South Africa’s BRICS partners—in comparator countries as well as Sub-Saharan Africa and the this case Brazil, Russia, and China—fare better than South World. The result: the bottom 40 percent had consumption Africa in terms of inclusiveness of growth. growth of 3.5 percent between 2006 and 2011, with a Figure 3: Shared prosperity indicator in selected countries (2007–2014) Source: Authors’ calculations based on WDI http://www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity Wealth inequality is also high and has been growing than poor households. Ownership of financial assets over time. The net wealth inequality is even higher features prominently among the factors that influence than consumption inequality in South Africa, although wealth inequality. For the poor, financial assets represent 36 there is strong correlation between levels of inequality percent of total assets compared to 75 percent for the rich. in consumption and wealth, with wealth remaining an Moreover, those with lower incomes and young to middle- important source of long-run inequality. Analysis of wealth age groups have high rates of indebtedness. This prevents inequality based on data from four rounds of wealth many segments of the population from participating in surveys carried out by UNISA between 2008 and 2015 asset accumulation and wealth building. Race and human suggests that the top percentile of households had 70.9 capital (education) have very high returns for wealth percent of the wealth and the bottom 60 percent had 7.0 generation, even higher than in the case of income or percent—richer households are almost 10 times wealthier consumption inequality. An Assessment of Drivers, Constraints and Opportunities xvii The labor market is effectively split into two extreme to workers living in developed economies, the wages of job types. At one extreme is a small number of people those at the lower end of the distribution are comparable with highly paid jobs in largely formal sectors and larger to those seen among the poorest countries. enterprises, at the other extreme is most of the population, The persistence of high wage gaps is associated with who work in jobs that are often informal and pay less the skills premiums and differences between unskilled, well. The highly paid jobs are highly sticky: once people semi-skilled, and high-skilled workers. With wages find these jobs they are unlikely to give them up. The less rising for skilled workers, the stagnation of wages for semi- well-paying jobs are more fluid, more likely to employ skilled workers fuels the increase in wage inequality. In fact, new entrants into the labor market, and more likely to workers in the middle of the distribution have witnessed an witness exits from employment. The wages between the erosion in the growth of their wages over time, relative to two extremes are highly unequal (Figure 5): those with the rest of the workforce in the labor market. This is related highly paid jobs earn nearly five times the average wage to the shrinkage of semi-skilled employment and their in low skilled jobs, yet they constitute less than a fifth of returns which points to the existence of a “missing middle” the total working population. Thus, while a segment of in the labor market, as evident in Figure 4. the population enjoys wages that are on average equal Figure 4: Real monthly wage by percentile, average Figure 5: Real wage inequality, 1995–2014 annualized percentage change, 1994–2014 Source: Post-Apartheid Labour Market Series, Authors’ calculations. xviii Overcoming Poverty and Inequality in South Africa Inequality of opportunity, measured by the influence slows the growth of the middle class, who made up about of race, parents’ education, parents’ occupation, 20 percent of the population between 2008 and 2015. place of birth, and gender influence opportunities, is Only 4 percent of the population can be considered elite high. In a society where there is equality of opportunity, with living standards far above the average. The middle these factors should not be relevant to reaching one’s full class consists of those who are in a better position to potential: ideally, only a person’s effort, innate talent, and maintain a non-poor standard of living even in the event choices in life would be the influencing forces. Analysis of of negative shocks. The size of the middle class in South the proportion of children with access to a basic service, Africa is considerably smaller than in other countries. For adjusted by how equitably the service is distributed example, close to 80 percent of Mauritius’ population could among groups differentiated by circumstances (via a be classified as middle class. Human Opportunity Index), shows that opportunities among children in South Africa vary widely depending LABOR MARKET INCOMES, EDUCATION, on the types of service. An estimation of the inequality GENDER, AND RACE ARE IMPORTANT DRIVERS OF INEQUALITY IN SOUTH AFRICA, THOUGH of opportunity index and its ratio to overall inequality EDUCATION AND INCOMES HAVE GROWN IN found that inequality of opportunity in South Africa is high IMPORTANCE WHILE GENDER AND RACE HAVE relative to its comparators. This is further compounded DECLINED by low intergenerational mobility, which is an obstacle to inequality reduction. Intergenerational mobility in South Labor market incomes, education, gender and race Africa is low in comparison to other countries indicating an are important drivers of inequality in South Africa, enduring link between life outcomes for a given generation though education and labor market incomes have versus those of the previous generation. grown in importance while gender and race have declined, contributing more than 90 percent of the overall SOUTH AFRICA HAS HIGH LEVELS OF CHRONIC Gini coefficient between 2006 and 2015. This is important POVERTY AND A RELATIVELY SMALL MIDDLE in the context of the high wage inequality, low labor force CLASS participation, and high unemployment that perpetuates high levels of inequality. For instance, high unemployment Nearly half of the population of South Africa is leads to relatively low levels of skill generation due to the considered chronically poor at the upper-bound national absence of high-paying jobs. This, in turn, perpetuates high poverty line of ZAR 992 per person per month (2015 levels of inequality. prices). This segment of the population is characterized by high poverty persistence. A second segment of the The importance of labor markets and education population has an above average chance of falling into factors in explaining inequality in South Africa has poverty (the transient poor). A third segment, the non- been growing. A decomposition analysis suggests poor but vulnerable, face above average risks of slipping race, education, and labor market income are the main into poverty though their basic needs are currently being contributors to the observed high level of inequality. met. These latter two groups made up 27 percent of the The inequality of opportunity in education is particularly population. Combining these two groups with the chronic influential in the transition to tertiary education, where poor suggests that for about 76 percent of the population, despite a high return, access to higher education remains poverty is a constant threat in their daily lives. limited. The influence of education on inequality raises concerns regarding low-income families that lack easy South Africa also has a high concentration of low access to credit markets and incur relatively high costs of income earners (the poor) and a few very high-income sending a child to college. This serves as a major barrier to earners (the rich or elite), but only a small number of getting sufficient levels of education to participate actively middle-income earners, resulting in a high level of in the semi-skilled and skilled labor market. income polarization. This high level of income polarization An Assessment of Drivers, Constraints and Opportunities xix While still an important factor, the impact of race very poor by squaring the poverty gap). This indicates an falls consistently across time in its contribution to improvement in the welfare of South Africans below the inequality. Notably though, some decline in the gender poverty line. Poverty is consistently higher among South bias for participation and employment is observed over Africans living in rural areas than for those in urban areas, time. Race and gender in earnings outcomes, while with the gap between rural and urban poverty rates retaining their predicted bias where African and female averaging around 40 percentage points during this period. workers earn, on average, significantly less than male and In rural areas, 65.4 percent of the population lived below white workers—does begin to decline after 2011. This is the poverty line in 2015, down 9.5 percentage points from important in that it creates an opportunity for policy to 74.9 percent in 2006. This is high compared to urban areas influence inequality outcomes. where 25.4 percent of the population were poor in 2015, following an 8.9 percentage point reduction from 34.3 SOUTH AFRICA HAS MADE PROGRESS IN percent in 2006. REDUCING POVERTY OF THE PAST TWO DECADES, BUT HIGH INEQUALITY ACTS Use of international poverty lines supports the overall AS A BRAKE ON POVERTY REDUCTION, SO positive story of declining poverty levels in post- POVERTY RATES REMAIN HIGH FOR AN UPPER apartheid South Africa but show that poverty rates MIDDLE-INCOME COUNTRY in South Africa are high for an upper middle-income country. The US$1.9 (2011 purchasing power parity, Close to 2.3 million South Africans escaped poverty exchange rates) poverty rate fell from 33.8 percent in 1996 between 2006 and 2015, as the poverty rate, measured to 18.8 percent in 2015 (Figure 7). Despite this long-term at the national lower-bound poverty line of ZAR 758 per progress, South Africa’s US$1.9 a day poverty rate is higher person per month (April 2017 prices), fell from 51 to 40 than that of many other upper middle-income countries percent during this period (Figure 6). Not only have the and higher than that of several countries with a per capita poverty rates fallen since the end of apartheid, poverty Gross National Income (GNI) less than that of South Africa became less deep (based on the poverty gap, a measure that (Figure 8). Further, it is higher than that of many other upper is calculated as the mean difference between consumption middle-income countries. For instance, at 18.8 percent, expenditure of each household and the poverty line) and South Africa’s US$1.9 poverty rate is higher than that of less unequal (based on the squared poverty gap which two of its BRICS partners, Russia (0.0 percent) and China (1.9 builds on the poverty gap and gives more weight to the percent). xx Overcoming Poverty and Inequality in South Africa Figure 6: Overall changes in national Figure 7: Long-term trends in $1.9/ Figure 8: Overall changes in poverty rates, lower-bound poverty day international poverty rates international poverty rates, lines comparison to other upper middle- income countries 80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 Urban Rural Total 2006 34.3 74.9 51.0 2009 31.5 74.9 47.6 2011 23.1 58.5 36.4 2015 25.4 65.4 40.0 2006 2009 2011 2015 Source: Authors’ calculations based on the Income Source: PovCalNet for 1993 to 2001 and authors’ Source: PovCalNet. and Expenditure Survey for 2005/06 and 2010/11 calculations for 2006 to 2015 based on the and the Living Conditions Survey for 2008/09 and Income and Expenditure Survey for 2005/06 and 2014/15. 2010/11 and the Living Conditions Survey for 2008/09 and 2014/15. The trajectory of poverty reduction was reversed opportunities among children and an important success between 2011 and 2015, threatening to erode some of for the education system to build on. An increase in access the gains made since 1994. At least three million more to telecommunications, electricity, improved water and South Africans slipped into poverty during this period, with sanitation, and school infrastructure has contributed to the poverty rate increasing from 36 percent to 40 percent. improved opportunities for children in South Africa. Further, Not only did poverty rates rise between 2011 and 2015, the estimation of the South African Multidimensional Poverty level of poverty became deeper and more unequal. This Index (SAMPI) shows a notable decline in multidimensional shows the welfare of poor South Africans worsened during poverty between 2001 and 2016, driven by a decline in the this period. Calculations at the US$1.9 a day poverty line proportion of households that were multidimensionally indicate a 2.4 percentage point increase in the poverty rate poor. Unemployment, followed by education (years of from 16.4 to 18.8 percent. schooling) are consistently the top two contributors to multidimensional poverty in South Africa, highlighting Consistent with the story revealed by trends in the importance of job creation and education in reducing monetary poverty rates, notable progress has been multidimensional poverty in South Africa. Comparing made in reducing multidimensional poverty since South Africa to other countries and regions in terms of the end of apartheid in 1994. Strides have been made the proportion of the population with access to electricity, in broadening access to basic public services. As Figure improved water sources, and improved sanitation facilities 9 shows, the proportion of the population with access to (Figure 9 to Figure 12) suggests South Africa lags behind an electricity, improved water sources, and improved sanitation average upper middle-income country but performs better facilities increased steadily between 1994 and 2015. than an average country in Sub-Saharan Africa. Further Analysis of the coverage rates of a basic service adjusted and consistent with use of monetary indicators, non- by how equitably the service is distributed among groups monetary indicators, specifically the SAMPI, show that the differentiated by circumstances suggests opportunities for major reduction in multidimensional poverty took place children are equalizing regardless of birth circumstances. between 2001 and 2011, while the last five years registered For instance, near-universal access to primary education stagnation in multidimensional poverty. has been achieved, a necessary first step for equalizing An Assessment of Drivers, Constraints and Opportunities xxi Figure 9: Changes in the proportion of the population Figure 10: The proportion of the population with with access to selected basic services access to electricity, comparison to other countries, 2014 Figure 11: The proportion of the population with access to Figure 12: The proportion of the population with an improved water source, comparison to other countries, access to improved sanitation facilities, comparison to 2015 other countries, 2015 Source: World Development Indicators. Note: Values for the poverty headcount ration are the most recent available over the past five years. xxii Overcoming Poverty and Inequality in South Africa POVERTY LEVELS ARE CONSISTENTLY A higher level of education of the household head HIGHEST AMONG FEMALE-HEADED and having access to stable labor market income, HOUSEHOLDS, BLACK SOUTH AFRICANS, THE by contrast, are key determinants for households to LESS EDUCATED, THE UNEMPLOYED, LARGE achieve economic stability in South Africa. Higher levels FAMILIES, AND CHILDREN of education of the household head are strong predictors of lower vulnerability to poverty. Living in a household where Poverty levels are consistently highest among female- the head has attained some tertiary education reduces the headed households, black South Africans, and children average risk of poverty by about 30 percent compared to below the age of 15 and these groups tend to have a those living in households where the head has no schooling. higher risk of falling into poverty (Figure 13 and Figure Poverty also tends to be a more temporary phenomenon for 14). Members of female-headed households are up to 10 those with higher labor market earnings. From this we may percent more likely to slip into poverty and 2 percent less conclude that improving access to quality higher and tertiary likely to escape poverty than members of male-headed education, easing labor market access, and improving the households. Race remains a strong predictor of poverty in quantity and quality of employment opportunities would be South Africa, with black Africans being at the highest risk important prerequisites to further poverty reduction. of being poor. Large families, children, and people in rural areas are especially vulnerable to being in poverty for a long time. Figure 13: Poverty headcount ratio by characteristics of head of household Source: Authors’ calculations based on the Income and Expenditure Survey for 2005/06 and 2010/11 and the Living Conditions Survey for 2008/09 and 2014/15. An Assessment of Drivers, Constraints and Opportunities xxiii Figure 14: Poverty headcount ratio by individual characteristics Source: Authors’ calculations based on the Income and Expenditure Survey for 2005/06 and 2010/11 and the Living Conditions Survey for 2008/09 and 2014/15. GEOGRAPHY IS STILL A MARKER OF POVERTY Cape had the highest poverty rate in 2015 and recorded the lowest reduction in poverty levels. Limpopo had the Poverty has a strong spatial dimension in South Africa, highest poverty headcount ratio of 67.1 percent in 2006, a demonstration of the enduring legacy of apartheid. 71.5 percent in 2009, and 52.7 percent in 2011. Its poverty As is typical in most parts of Africa, rural areas have the rate in 2015 was 57.0 percent. Gauteng consistently has the highest poverty concentration in South Africa. In 2006, lowest poverty rate (19.0 percent in 2015). At 26.0 percent 60.3 percent of the poor were in rural areas. This decreased in 2015, KwaZulu-Natal had the largest share of the poor marginally to 59.7 percent in 2015. Eastern Cape, KwaZulu- in South Africa. This is partly due to the relatively high Natal, and Limpopo were consistently the three poorest population share in KwaZulu-Natal. provinces between 2006 and 2015. At 59.1 percent, Eastern xxiv Overcoming Poverty and Inequality in South Africa Figure 15: Poverty incidence at the municipality level 1996 2011 Source: Poverty Map calculations (map in the left) are from Alderman et al. (2002) and 2011 Poverty Map calculations (map in the right) are the author’s. Notes: Darker lines correspond to province level boundaries. Not only do poverty and inequality vary cross provinces, Natal, and North West where the highest individual poverty they vary across districts and municipalities. A poverty rates at the municipality level were found. In contrast, mapping exercise using the 2011 South African population extreme poverty was highest in the central and eastern census data reveals the existence of various pockets of parts of the country (Free State, Eastern Cape, North West, poverty at the municipality level within provinces, but also and Northern Cape) in 1996. The spatial distribution of the notable dispersion of municipality poverty rates in poverty shifted from the central areas of the country in others (Figure 15). In 2011, poverty was more prevalent in 1996 to the borders and remote areas in 2011. peripheral areas of the Eastern Cape, Limpopo, KwaZulu- Figure 16: Multidimensional poverty headcount ratio at the municipality level, the 20 poorest municipalities 2001 2016 Source: Poverty Map calculations are from Statistics South Africa. An Assessment of Drivers, Constraints and Opportunities xxv The results reveal a notable divide in poverty levels returns to education, especially to the semi-skilled between two sets of provinces: Free State, Gauteng, occupations, are not increasing anymore. Urbanization, and Western Cape versus Eastern Cape, KwaZulu-Natal, demographic changes, and expansion in the provision of and Limpopo. This divide is a clear legacy of apartheid: services all contributed to the improvement of households’ compared to Eastern Cape, KwaZulu-Natal, and Limpopo; welfare. While having an employed household head the Free State, Gauteng, and Western Cape did not have does not necessarily translate to a lower vulnerability to high concentrations of “homelands” during apartheid. poverty, the type of employment that the head engages in, Homelands were areas set aside for black South Africans especially regarding its stability and duration, is important. along ethnic lines during apartheid. Public service delivery and infrastructure was poor in these areas. An estimation SOCIAL PROTECTION IS IMPORTANT IN of the multidimensional poverty index for South Africa SUPPORTING POVERTY AND INEQUALITY REDUCTION PARTICULARLY AMONG THE supports this spatial pattern of poverty. High levels of EXTREMELY POOR multidimensional poverty are found in areas that are predominantly rural. In terms of variation across provinces, Since the end of apartheid, the government has Eastern Cape had the highest SAMPI score, alongside progressively expanded its spending on the social Limpopo, driven by relatively high multidimensional wage, broadly defined to encompass investments in poverty headcount ratios. Considering performance areas deemed to help address poverty and inequality, among municipalities, the 20 poorest municipalities were while maintaining generally sound fiscal indicators. in the Eastern Cape, Limpopo and KwaZulu-Natal (Figure It broadened the tax base and built an efficient tax 16). Multidimensional poverty remains concentrated administration to generate the resources it needed to in previously disadvantaged areas, such as the former expand the social safety net for the poor. The country has homelands: the 10 poorest municipalities are in the an extensive transfers system that benefits a quarter of the former homelands of Eastern Cape and KwaZulu-Natal, population. Close to 17 million low-income South Africans highlighting the enduring effects of apartheid, which have access to means tested social grants. limited development in homelands. The 20 richest municipalities are mainly in the Western Cape. A strong Social assistance has proven successful in reducing correlation is found between municipality-level poverty extreme poverty. In 2015, government social transfers rates in 1996 and 2011: the higher the poverty rate a are estimated to have reduced the poverty headcount municipality had in 1996, the more likely it was to also have rate by 7.9 percent and the poverty gap by 29.5 percent. higher poverty rate in 2011. This suggests spatial patterns This is explained by very high rates of coverage among the of poverty have not changed much over time. poorest members of society, with coverage rates among the bottom 60 percent far above average coverage rates Labor market incomes were an important source of of other upper middle-income countries. The grants had poverty reduction between 2006 and 2015. When an impressive impact on poverty. Studies found that the decomposing change in poverty between 2006 and 2015 grants are used in many households to improve health by income sources, labor market income is shown to be the and education outcomes, resulting in long-term impact on largest contributor to improving people’s lives at national poverty reduction. At the same time the negative impact of level, and in urban settings, but less so in rural areas. grants on employment is very small. Improvement in skills and education were instrumental for poverty reduction in South Africa, although returns to Social transfers kept inequality from rising in South education have been decreasing in recent years. In other Africa. The analysis suggests that income inequality was words, the overall population has attained more education stagnant in recent years. However, without social assistance since 2006, and that helped reduce poverty. However, the Gini coefficient would have been 10.5 percent higher, a significant and unprecedented impact on inequality. xxvi Overcoming Poverty and Inequality in South Africa On average, in upper middle-income countries, the Gini The analysis in this report highlights the importance coefficient is reduced by 1.7 percent by social transfers, of job creation and skills improvement to reducing while the reduction is 0.7 percent in Sub-Saharan Africa and poverty and inequality in South Africa. The study 1.6 percent in Latin American countries. The South African underscores the importance of growing the economy in social assistance system is thus very effective at keeping an inclusive manner that generates much-needed jobs inequality in check. to achieve further reductions in poverty and inequality. South Africa’s polarized economy, coupled with its skills Poverty reduction in the later part of the 2000s is constraint, hurts the poor and keeps inequality high. The strongly associated with expansion of social grants, lack of competitiveness from low productivity undermines but further expansion of social grants in the future job growth, thus excluding many from labor markets. South is fiscally unsustainable. Further expansion of social Africa has two segments that do not seem to integrate—a grants in a time of low economic growth and slowdown small high skill, high-productivity segment and a large low- in tax revenues poses a challenge to fiscal sustainability. skilled, low-productivity segment. The overall goal of economic policy could be to keep the current social protection system while seeking to drive Interventions that simultaneously stimulate growth growth by addressing labor market issues, skills gaps, and and reduce inequalities are likely to have much more job creation. impact than interventions that only stimulate growth or only reduce inequalities. Analysis of current policy ACCELERATING THE REDUCTION OF POVERTY interventions, such as the employment tax incentive and AND INEQUALITY WILL REQUIRE UNLOCKING the national minimum wage, suggests that their effect THE FULL POTENTIAL OF LABOR MARKETS on inequality, and thus poverty, is very modest. Creating AND PROMOTING INCLUSIVE GROWTH good jobs for the poor will have a much larger impact THROUGH SKILLS CREATION on inequality and poverty. The social impact of reforms The prospects for eliminating poverty by 2030, currently envisaged to boost growth would be significantly the goal of the government’s current policy, will amplified with reforms to equip poor to reap growth depend on gross domestic product (GDP) growth and opportunities, through the acquisition of skills. Such inequality reduction, the former being affected by the reforms would also further strengthen the social compact, level of access the poorest groups have to economic with a likely positive effect on investment. Nonetheless, opportunities, as well as by fiscal redistribution. South recognizing the time needed to increase the economic Africa has low growth-to-poverty elasticities due to its participation of the poor—whole generations—such a extremely high level of inequality. The extent of poverty package of reforms would still need to maintain social reduction therefore depends on both economic growth and assistance to the poor and vulnerable. Higher fiscal revenue inequality reduction. Sluggish growth with improvements from accelerated growth would provide the fiscal space to in access to education among the poor is anticipated to do so. slightly reduce inequality and poverty in the coming years. Poverty rates (at the lower-bound national poverty line) are projected to decrease from 40 percent of the population in 2015 to 33 percent in 2030 despite slow growth, as inequality would decline with a Gini coefficient dropping from 62.8 in 2017 to 59.5 in 2030. An Assessment of Drivers, Constraints and Opportunities 1 CHAPTER 1 INTRODUCTION Through implementing its 2012 National Development The initiatives have been supported by sound Plan (NDP), South Africa aims to eliminate poverty and institutions and economic gains since 1994. South reduce inequality by 2030. That plan builds on previous Africa is an upper middle-income economy with a generally post-apartheid policy documents for which reduction of stable macroeconomic system, diversified economy, poverty and inequality have been anchors, including the relatively low taxes and tariffs, well controlled fiscal deficit, 1994 Reconstruction and Development Program (RDP), and relaxed exchange rates. The end of apartheid in 1994 the 1996 Growth, Employment, and Redistribution (GEAR), resulted in major adjustments in the economy that helped and the 2006 Accelerated and Shared Growth Initiative for to support growth. With the end of sanctions by the South Africa (AsgiSA). Initiatives taken under those policies international community South Africa was reintegrated have sought to address the country’s triple challenges of into the global trading system and benefited from capital high poverty, high inequality, and high unemployment. The reallocation and new investment from abroad. Labor initiatives include, for example, the use of fiscal policy as a markets opened to the entire South African population as tool to effect redistribution. Specifically, transfers to different the race-based jobs reservation policy ended. The financial spheres of government are based on poverty considerations. sector also opened to more South Africans, allowing them In addition, the social wage—government investments in to access credit to build assets or finance consumption. education, health services, social development, as well as The economy grew by an average 2.9 percent between social assistance to vulnerable households and individuals 1994 and 2000 (Figure 17), supported by labor expansion and contributory social security, public transport, housing, and capital reallocation. It accelerated to an average 4.2 and local amenities—has played a notable role in these percent between 2001 and 2008, supported by significant efforts. The social wage accounts for close to 60 percent of investment, household borrowing and growing wages government spending. supporting private consumption, and buoyant commodity prices (commodities account for about 60 percent of South African exports). The average growth fell to 1.6 percent between 2009 and 2016. 2 Overcoming Poverty and Inequality in South Africa Figure 17: Real GDP growth decomposition Source: South African Reserve Bank; Bloomberg and World Bank staff calculations. Expenditure side decomposition. Figure 18: Economic structure of South Africa Figure 19: Average labor productivity decomposition (share of GDP, supply side) (contributions to labor productivity growth) Source: South African Reserve Bank; Bloomberg and World Bank staff Source: The value-added shares are from World Development Indicators calculations. (WDI), share of employment for agriculture, services, and industry is obtained using estimates from the International Labor Organization (ILO), as in Senkal (2017). An Assessment of Drivers, Constraints and Opportunities 3 Since 1994, the economy has undergone structural Very low economic growth in recent years is the transformation with a decline in primary sectors and main challenge for the government’s far-reaching expansion of tertiary sectors. From the supply side, development plan. The global financial crisis hit South growth has been driven by the services sector, which Africa hard, and economic progress has stalled since then. is made up of trade, transportation, finance, and social Growth between 2009 and 2017 averaged only 1.6 percent, services, and accounted for 70 percent of gross domestic gradually declining from an intermittent high in 2011 to only product (GDP) in 2016, up from 60 percent in 1994 (Figure 0.3 percent in 2016/17. Further, low quality of education, 18). Both the primary and secondary sectors have been high HIV/AIDs prevalence, and poor government service losing GDP share. Agriculture, including forestry and delivery to remote and poor communities compromise fisheries, fell from 3 percent in 1994 to 2 percent in 2016—a efforts to reduce unemployment, poverty, and inequality. small share by regional standards, owing in no small part to High unemployment remains the key challenge as the the relatively high level of development and sophistication country struggles to generate sufficient jobs. Overall, of the South African economy. Industry, comprising since 1994, a growing economy created many jobs in mining, manufacturing, utilities, and construction, fell South Africa—but not enough to significantly reduce from 37 percent of GDP in 1994 to 28 percent in 2016. unemployment. Although the NDP envisions the creation To a significant degree, the South African economy had of 11 million jobs between 2011 and 2030, this is unlikely been built on mining, but the sector has increasingly lost to occur. To achieve the employment target of the NDP, share to services. Lack of a dynamic, job-generating, and the economy would need to create about 600,000 jobs a competitive manufacturing sector remains a significant year, but the economy has barely been managing to create growth challenge. half of that. Net job creation between 1993 and 2015 was South Africa’s structural transformation is well 2.7 million in the private sector (formal and informal) and advanced, but factors of production are not always 470,000 in the public sector, almost exclusively created after allocated to their most productive use. The services 2005. Most private sector jobs were created in the services sector is already the largest sector in the economy and the sector, with agriculture and manufacturing shedding engine of growth. Labor productivity since 1994 has mostly jobs—not least because of increasing capital intensity been driven by productivity gains within sectors with in those sectors. Between 2011 and 2015, an average relatively little reallocation of factors of production across of 589,000 workers entered the labor force every year, of sectors (Figure 19). Rigidities and frictions in the economy, which only 424,000 found employment; 165,000 became including relatively inflexible labor and capital markets—to unemployed and 20,000 left the labor force discouraged a significant extent due to muted competition—constrain from being able to find any work. Thus, despite significant the efficient reallocation of factors and both capital and job creation, the pace of employment growth was too labor do not always allocate factors of production to their slow for the pool of unemployed workers and new labor most productive use.2 Such inefficiency is one reason for market entrants. Unemployment hit a 14-year high of 27.7 the poor performance of total factor productivity. Since percent in the first quarter of 2017. High unemployment the global financial crisis, total factor productivity has been is increasingly putting pressure on South Africa’s social declining, costing an estimated 0.6 percentage points of contract as a job is the main way out of poverty and toward forgone GDP growth every year.3 a more prosperous life. The target of the South African government is to cut unemployment by at least half, to a maximum of 2 For capital reallocation, see 10th South Africa Economic Update (2017) 14 percent, in 2020. However, it is not evident that this and for labor reallocation see “Firm level dynamics, job flows and pro- target can be met given the modest gains in employment ductivity: South Africa 2009–2014,” mimeo, World Bank and National Treasury of South Africa. in the recent past. Further, it is worrying that employment 3 10th South Africa Economic Update (2017). continues to have a gendered and generational distribution. 4 Overcoming Poverty and Inequality in South Africa Women have far worse employment prospects than men— The persistence of these challenges, 24 years after the around 37 percent of working age women are employed end of apartheid, calls for a comprehensive assessment compared to 50 percent of men. Youth (15–24 years old) of the extent and causes of poverty and inequality in unemployment is also high, consistently around 50 percent, South Africa with attention to trends, dynamics, policy, and much higher than all of the older age groups. Finally, impact, and monitoring. This is especially pertinent given employment does not necessarily reduce poverty: about that the last comprehensive national poverty and inequality 35 percent of those who are employed are in households assessment was published in 1998 (May et al. 1998). It was living below any of the poverty lines. commissioned by the government of South Africa with assistance from the United Kingdom’s Department for Companies doing or planning to do business with International Development, United Nations Development the South African government must comply with Program (UNDP), World Bank, and the Dutch government. Broad-Based Black Economic Empowerment  (B-BBEE) policies. B-BBEE policies aim to redress past imbalances and The purpose of this report is to document South Africa’s broaden the economic access to members of historically progress in reducing poverty from 1994 to date. It aims disadvantaged communities, and this way, facilitate to contribute to the realization of South Africa’s national socioeconomic transformation. The goal is to increase the targets of eliminating extreme poverty and reducing number of black South Africans that either own or manage inequality by 2030. The specific objectives are as follows: companies. Under the policies, companies gain credits • To enhance understanding of the barriers to and if they have a certain percentage of black ownership and engines of reducing poverty and inequality in South participate in the supply chain with companies complying Africa in recent years. with the policies. Since 2012, the government has announced plans to tighten labor and foreign ownership • To critically assess the role of labor markets in reducing laws and mandated industrial localization. Sectors of poverty and inequality in South Africa. concern have included the extractive industries, security • Based on the results from the analysis, to identify services, and agriculture. It remains uncertain in which possible areas of intervention that will accelerate the direction government will go to address rigidities in labor reduction of poverty and inequality. regulations in the face of popular discontent around Pursuing these objectives enables the report to unemployment, poverty, and inequality. contribute to policy dialogue toward the attainment The need for more inclusive growth has been of the NDP vision. It also offers insights into how the recognized in South Africa. Growth is less likely to be challenge of unemployment can be tackled through the sustainable with high and increasing inequalities; and creation of more and better jobs as well as by improving high inequality can hamper growth. While the use of the the employability of the labor force. The focus on the fiscal system (specifically the social wage) is central to labor markets is justified given the challenge of high the reduction of poverty and inequality, the current low unemployment and the impact that has on poverty and level of growth and accompanying deterioration in the inequality. Unemployment rates tend to be higher among fiscal balance raise questions about the sustainability of the poor. Similarly, labor force participation is lower in poor the social wage–oriented interventions against poverty than non-poor households. and inequality. As the World Bank (2014) notes, although The report draws on several technical background fiscal policy goes a long way toward reducing poverty and papers produced by local and international inequality, both remain high. The challenge, therefore, is to researchers. It also builds on substantial existing work and make growth more inclusive by finding ways to boost the a knowledge base that includes large sample surveys, panel incomes of poor. data, detailed evaluations, and impact assessments, as well as qualitative studies undertaken by Statistics South Africa An Assessment of Drivers, Constraints and Opportunities 5 (Stats SA) and independent South African and international The report is organized as follows: Chapter 2 discusses researchers. the trends in monetary and non-monetary poverty since the end of apartheid, with a focus on 2006–2015. It also The report adds value to existing work in six ways. provides a profile of the poor and their location. The extent First, it adopts a policy focus that is missing in the bulk and determinants of transitions into and out of poverty of existing literature on poverty and inequality in South is also discussed. The chapter documents South Africa’s Africa. Second, given the global reach of the World Bank, progress in reducing poverty since 1994, though poverty the report benchmarks against and brings experiences rates remain high for an upper middle-income country and from other countries in similar circumstances. Third, the trajectory of poverty reduction was reversed between focusing on inequality in addition to poverty (including 2011 and 2015. Chapter 3 presents different dimensions the bottom 40 percent, middle classes, and vulnerable of inequality and documents the unusually high level of groups) brings a new perspective. Fourth, by using panel inequality in South Africa. Chapter 4 examines the drivers data sources, particularly the four waves of the National of poverty reduction and inequality largely through Income Dynamics Study (NIDS), the report is able to frame decomposition analysis. Labor market incomes emerge the whole discussion dynamically. Fifth, the use of non- as a large contributing factor. Chapter 5 details the link monetary (beyond income/consumption) indicators of between labor market dynamics and poverty in South poverty that are relevant to South Africa ensures that the Africa. Chapter 6 concludes by synthetizing the policy analysis will inform policy dialogue. Sixth, the recently implications of the preceding chapters and identifying completed Living Conditions Survey 2015 by Stats SA possible areas of intervention that would accelerate the creates an opportunity to provide a comprehensive and reduction of poverty and inequality. up-to-date analysis of poverty and inequality. 6 Overcoming Poverty and Inequality in South Africa CHAPTER 2 EVOLUTION, DIMENSIONS AND DYNAMICS OF POVERTY IN SOUTH AFRICA Poverty levels in South Africa have fallen since 2006. In 1996, the poorest provinces. Poverty is persistent in South Africa and 33.8 percent of South Africans lived below US$1.9 a day. This the economy is highly polarized as evident in a relatively small fell to 25.5 percent in 2006 and to 18.8 percent in 2015. Using middle class and high levels of chronic poverty. Almost half of the national lower bound poverty line of R647 per person per the population is considered chronically poor at the national month in 2015 prices, 51.0 percent of the population was poor upper bound poverty line, of R992 per person per month in in 2006 and 40.0 percent in 2015. However, as the chapter 2015 prices. A higher level of education of the household head documents, although the overall trend indicates progress and access to stable labor market income are key determinants toward poverty reduction between 1996 and 2015, between for households to achieving economic stability in South Africa. 2011 and 2015 poverty rates rose from 36.4 percent to 40.0 This chapter discusses the trends in monetary and non- percent at the national lower bound poverty line. Consistent monetary poverty since the end of apartheid, with a focus with this, non-monetary indicators of poverty indicate notable on the period between 2006 and 2015. International progress in reducing multidimensional poverty after 1994, but poverty lines are used to compare South Africa to peers in it has stagnated in recent years. A profile of the poor shows terms of income levels. The chapter also profiles the poor a typical poor household is rural and headed by a single, based on individual and household characteristics as well as economically inactive female black South African. Rural areas their geographic distribution. The extent and determinants remain the regions of highest poverty concentration and the of transitions into and out of poverty are also discussed. Eastern Cape, KwaZulu-Natal, and Limpopo are consistently An Assessment of Drivers, Constraints and Opportunities 7 A. DESPITE PROGRESS IN REDUCING the population living below a specific poverty line (Figure POVERTY SINCE 1994, POVERTY RATES 20). The number of South Africans living below the food REMAIN HIGH FOR AN UPPER MIDDLE- poverty line (FPL) fell from 28 percent in 2006 to 25 percent INCOME COUNTRY in 2015. The corresponding decline was from 51.0 to 40.0 percent at the lower bound poverty line (LBPL), while it i. Trends in national poverty declined from 66.6 percent to 55.5 percent at the upper bound poverty line (UBPL). In absolute terms, around 2.3 South Africa recorded a decrease in consumption million South Africans escaped poverty at the LBPL and 1.2 poverty rates between 2006 and 2015, regardless million at the UBPL. However, around 343,000 more South of the poverty measure used.4 All measures indicate Africans were poor based on the FPL in 2015 than in 2006 a decline of at least 3 percentage points in the national (see Box 1 for an explanation of poverty measurement poverty headcount ratio, which captures the proportion of methodology in South Africa). 4 Henceforth, 2006 refers to 2005/06, 2009 to 2008/09, 2011 to 2010/11, and 2015 to 2014/15 survey years. Figure 20: Overall changes in poverty rates Source: Authors’ calculations based on the Income and Expenditure Surveys for 2005/06 and 2010/11 and the Living Conditions Surveys for 2008/09 and 2014/15. 8 Overcoming Poverty and Inequality in South Africa Box 1: The methodology of poverty measurement in South Africa In South Africa, absolute poverty is measured by comparing per capita household consumption expenditure to a specified national poverty line. All food items are included in the welfare indicator while non-food items, large-sized, or “lumpy, dura- ble goods” are excluded to reduce their biasing factor in the monthly estimates. To get the welfare indicator, all household consumption expenditures are annualized and then adjusted according to household size. The surveys used for this welfare measurement are typically the Income and Expenditure Surveys (IES) and the LCS which are administered by Stat SA and collect detailed information on household expenditures. In addition, the surveys collect information on household expen- ditures, education, demographics, income, and as of 2015, labor market status. The households sampled in each wave are meant to be nationally and regionally representative. Poverty lines are determined using a cost-of-basic-needs (CBN) approach.5 In 2012, Statistics South Africa (Stats SA) pub- lished a suite of three national poverty lines to be used for poverty measurement. These have since been used in most official studies of poverty. The three poverty lines are the food poverty line (FPL), the lower bound poverty line (LBPL), and the upper bound poverty line (UBPL). The FPL is the level of consumption below which individuals are unable to purchase sufficient food to provide them with an adequate diet. It is determined in two stages. First, a food reference basket is con- structed. Second, the basket is costed to determine the level of the FPL. This line is also considered the extreme poverty line. The LBPL and UBPL lines are computed by including an allowance for non-food consumption. To determine the level of the LBPL, the average expenditure on non-food items by households whose total expenditure is close to the FPL is added to the FPL. Thus, the LBPL is based on households that sacrifice some of their basic food requirements to meet their non-food needs. The UBPL, on the other hand, is computed by adding the average expenditure on non-food items by households whose food expenditure is very close to the food line as the reference group. For these households, in addition to the basic food requirements that are measured by the FPL, there are certain basic non-food items that they need. Individuals can purchase both adequate food and non-food items at the UBPL. The three poverty lines are updated periodically using the Consumer Price Indexes (CPIs). The mechanism used to update the poverty lines is described in Stats SA (2008: 23) Table 1: Inflation-adjusted poverty lines, 2006–2017 (per person per month in South African Rands) Year Food poverty line Lower bound poverty line Upper bound poverty line 2006 219 370 575 2007 237 396 613 2008 274 447 682 2009 318 456 709 2010 320 466 733 2011 335 501 779 2012 366 541 834 2013 386 572 883 2014 417 613 942 2015 441 647 992 2016 498 714 1,077 2017 531 758 1,138 Source: Stats SA (2017). Note: All values are linked to March prices, except for 2015, 2016, and 2017 which are linked to April prices. 5 5 See Stats SA (2008 and 2015) for the history and technical discussion of poverty lines in sSouth Africa An Assessment of Drivers, Constraints and Opportunities 9 Three areas of improvement in the way poverty is measured in South Africa are noted. First, the way in which non-food items are selected for inclusion in the welfare indicator could be further improved in line with international best practices and reflected in Deaton and Zaidi (2002). Second, the value of the consumption flow from durable goods needs to be more comprehensively included in the welfare indicator. Third, the introduction of adjustments for regional differences in prices (spatial deflation) in addition to the intra-year temporal deflation to compute a real welfare indicator is recommended. The welfare of South Africans below the poverty It builds on the poverty gap and gives more weight to the line improved between 2006 and 2015. A reduction is very poor by squaring the poverty gap. It reflects the degree revealed in two alternative measures of poverty that focus of inequality among the poor themselves.7 The squared more on the poor and capture the depth and severity of poverty gap declined from 12.2 to 9.1 percent between poverty: the poverty gap and poverty severity. The depth 2006 and 2015, suggesting reduced severity of poverty. of poverty is a measure of intensity and is calculated as Overall, these two measures suggest poverty became less the mean difference between household consumption deep and less unequal between 2006 and 2015. expenditure and the poverty line.6 It is expressed as a percentage of the poverty line. Measured at the LBPL, Table 2 shows that the poverty gap fell by 5.5 percentage points from 22.2 percent in 2006. This means the per capita amount of resources needed to eliminate poverty through perfectly targeted cash transfers decreased between 2006 and 2015. The squared poverty gap is an indicator of poverty severity. 7 The squared poverty gap considers not only the poverty gap but also 6 The poverty gap is the mean shortfall of the entire population from the inequality among the poor by placing more weight on house- a specified poverty line. It is measured from zero to 100, with zero holds that are further from the poverty line. A transfer from a poor to a meaning no poverty while 100 indicates zero consumption expendi- less-poor person raises the squared poverty gap while a transfer from ture for everyone and a positive poverty line. a poor to a poorer reduces it. 10 Overcoming Poverty and Inequality in South Africa Table 2: Changes in the depth and severity of poverty   Poverty gap Change: Change:   2006 2009 2011 2015   2015–2006 2015–2011 Urban 4.0 6.3 3.6 4.1 0.1 0.5 Food poverty line Rural 16.9 22.5 12.1 17.7 0.8 5.5 Total 9.3 12.3 6.8 9.0 -0.2 2.3 Urban 12.6 12.2 8.2 8.9 -3.6 0.7 Lower bound poverty line Rural 35.9 36.0 24.3 30.0 -5.9 5.8 Total 22.2 21.0 14.3 16.6 -5.5 2.4 Urban 23.8 22.2 16.5 17.5 -6.3 1.0 Upper bound poverty line Rural 52.6 52.6 40.3 45.5 -7.1 5.2 Total 35.6 33.5 25.5 27.7 -7.9 2.2   Squared poverty gap Urban 1.6 2.9 1.5 1.8 0.1 0.3 Food poverty line Rural 8.0 11.3 5.6 9.1 1.1 3.5 Total 4.2 6.0 3.0 4.5 0.2 1.4 Urban 6.1 6.3 4.0 4.4 -1.8 0.4 Lower bound poverty line Rural 20.8 20.9 12.8 17.3 -3.5 4.5 Total 12.2 11.7 7.3 9.1 -3.1 1.8 Urban 13.6 13.0 9.1 9.7 -3.9 0.6 Upper bound poverty line Rural 35.3 35.4 24.9 29.7 -5.7 4.8 Total 22.5 21.3 15.0 17.0 -5.5 1.9 Source: Authors’ calculations based on the Income and Expenditure Surveys for 2005/06 and 2010/11 and the Living Conditions Surveys for 2008/09 and 2014/15. Poverty is higher in rural than in urban areas, and the 8.9 percent. The amount of resources needed to bring the gap between rural and urban poverty rates widened consumption expenditure of the poor up to the poverty between 2006 and 2015. In rural areas, 65.4 percent lived line is higher in rural than urban areas. Similarly, inequality below the LBPL in 2015, down from 74.9 percent in 2006. among the poor is relatively larger in rural than in urban In urban areas, 25.2 percent of the population were poor, a areas: the squared poverty gap was 17.3 percent in rural drop from 34.3 percent in 2006. The gap between rural and areas, while it was 4.4 percent in urban areas in 2015 at urban poverty did not change significantly between 2006 the LBPL. The challenge around the depth and severity of and 2015: it was about 41 percentage points in 2006 and 40 poverty at the food (extreme) poverty line is shown in Table percentage points in 2015. 2: the poverty gap and the squared poverty gap increased, albeit slightly, in both rural and urban areas between 2006 Not only is the poverty headcount ratio higher in rural and 2015. areas compared to urban areas, poverty is deeper and more unequal in rural areas as well. However, at the Despite the positive trend on poverty reduction LBPL, the depth and severity of poverty fell faster in rural between 2006 and 2015, poverty rates increased than in urban areas between 2006 and 2015. The poverty between 2011 and 2015. At least 2.5 million more gap in rural areas decreased by 5.9 percentage points from South Africans slipped into poverty between 2011 and 35.9 percent in 2006 to 30.0 percent in 2015. In urban areas, 2015, despite a positive overall trend in poverty reduction a 3.6 percentage point reduction was recorded from 12.6 to between 2006 and 2015. Forty percent of the South African An Assessment of Drivers, Constraints and Opportunities 11 population lived below the LBPL in 2015, up from 36.4 ii. International poverty trends percent in 2011. In absolute terms, this translates to over 3.1 The overall positive story of declining poverty levels million more South Africans slipping into poverty between in post-apartheid South Africa is supported by the 2011 and 2015. international poverty lines. Figure 21 shows a positive Not only did poverty rates rise between 2011 and 2015, overall trend in poverty reduction at the US$1.9 (2011 the level of poverty became deeper and more unequal. purchasing power parity, PPP, exchange rates) poverty line, Measured at the LBPL, Table 2 shows that the poverty gap which fell between 1996 and 2015. Between 2006 and 2015, rose by 2.3 percentage points from 14.3 percent in 2011. the poverty headcount ratio also fell. At the US$3.1 a day This means the per capita amount of resources needed international poverty line, poverty levels fell between 2006 to eliminate poverty through perfectly targeted cash and 2015. In absolute terms, the number of poor fell by transfers increased between 2011 and 2015. The squared around 1.8 million between 2006 and 2015 at the US$1.90 poverty gap increased from 7.3 to 9.1 percent suggesting a day poverty line and by 2.4 million at the US$3.1 a day increased severity of poverty during this period. According international poverty line. to the Stats SA’s 2017 poverty trends report, the increase The use of international poverty lines also supports in the poverty levels between 2011 and 2015 is associated the story of increasing poverty rates between 2011 with “a combination of international and domestic factors and 2015. Calculations indicate a 2.4 percentage point such as low and anemic economic growth, continuing increase at the US$1.9 a day poverty line (Figure 22). high unemployment levels, lower commodity prices, At the US$3.1 a day international poverty line, the rate higher consumer prices (especially for energy and food), increased by 2 percentage points between 2011 and 2015. lower investment levels, greater household dependency A 1.7 percentage point increase is observed at the US$5.0 on credit, and policy uncertainty.” (Statistics South Africa a day poverty line. Around 1.8 million more South Africans 2017, pp 16). Rather than focus on the most recent trends, slipped into extreme poverty measured at the international this study takes a longer-term perspective with the aim of poverty line of US$1.9 a day between 2011 and 2015. understanding the causes and consequences of polices This figure rises to around 2.2 million when the US$3.1 a and sources of poverty reduction. This requires a longer- day poverty line is used, and to around 2.7 million South term perspective and makes it possible to better capture Africans at the US$5.0 a day poverty line. and explore factors and polices affecting inclusive growth and poverty in South Africa. 12 Overcoming Poverty and Inequality in South Africa Figure 21: Long-term trends in US$1.9/day international Figure 22: Overall changes in US$1.9/day poverty rates international poverty rates Source: Authors’ calculations based on the Income and Expenditure Surveys for 2005/06 and 2010/11 and the Living Conditions Surveys for 2008/09 and 2014/15. PovCalNet for the years 1993, 1996 and 2001. South Africa’s US$1.9 (PPP) a day poverty rate is is higher than that of many other upper middle-income higher than that of many other upper middle-income countries. For instance, at 18.8 percent, South Africa’s countries. In 2015, 18.8 percent of South Africans lived poverty rate is higher than that of two of its BRICS partners, below the US$1.9 a day international poverty line. This is Russia (0 percent) and China (2 percent) (Figure 23 and higher than several countries that have a lower per capita Figure 24). gross national income (GNI) than South Africa. Further, it Figure 23: Overall changes in international poverty Figure 24: Overall changes in international poverty rates, comparison to other countries rates, comparison to other upper middle-income countries Source: PovCalNet and WDI. Note: Values are the most recent available over the past five years. An Assessment of Drivers, Constraints and Opportunities 13 B. WHO ARE THE POOR? head of household, unemployment status as proxied by economic activity, and the composition of the household, The profile of the poor is presented at the LBPL. The such as its size and age structure. Figure 25 and Figure 26 demographic characteristics of households, such as family present poverty rates by each characteristic considered size, structure, and ethnicity, are important in determining in this chapter. The profiles are generated using the LBPL, the socioeconomic status of the family and its level of consistent with the focus on the LBPL in the NDP. It is poverty. Thus, the analysis in this section focuses on the important to note that these profiles do not use equivalent demographic composition of households, attainment scales, but rather are drawn from a welfare measure of education, and labor indicators. The considered (consumption per capita) that treats everyone the same and characteristics include gender, race, and education of does not account for different needs within households. Figure 25: Poverty headcount ratio by characteristics of head of household Source: Authors’ calculations based on the Income and Expenditure Surveys for 2005/06 and 2010/11 and the Living Conditions Surveys for 2008/09 and 2014/15. The calculations are done using the LBPL. Poverty is higher among individuals living in female- poverty rates of the two groups did not change over the headed households compared to those living in male- years, remaining at around 20 percentage points in each headed households across all periods analyzed. In period. 2006, 63.4 percent of female-headed households were Black South Africans consistently exhibit the highest poor compared to about 41.5 percent of households with poverty rates. In 2015, 47 percent of the households male heads. In 2015, the poverty headcount among female- headed by black South Africans were poor. This was very headed households was 51.2 percent compared to 31.4 high compared to 23 percent for those in households percent among male-headed households. The reduction in headed by a person of mixed race (colored), a little more poverty rates was not significantly different between the than one percent for the population in households headed two groups: the decline was 11 percentage points among by an Indian/Asian South African, and less than one percent female-headed households and 10 percentage points among those in households headed by white South Africans. among male-headed households. The gap between the 14 Overcoming Poverty and Inequality in South Africa Between 2006 and 2015, all ethnic groups experienced a Poverty declines with rising levels of education. In reduction in poverty rates, with black and colored South 2015, 73.1 percent of the population living in households Africans experiencing the fastest decline. Black South whose head did not have a formal education versus 2.6 Africans make up close to 80 percent of the population. percent of those living in households whose head had Despite the gains made by these two population groups attained an education beyond upper secondary school between 2006 and 2015, they registered an increase in were poor. Between 2006 and 2015, the population living poverty between 2011 and 2015. The black South African in households with heads who had completed primary group registered an increase of 3.7 percentage points while school experienced the fastest decline in poverty. Similar the colored group registered an increase of 2.5 percentage patterns are true for individuals: in 2015, 55.0 percent of points. individuals with no formal education were poor compared to 2.6 percent of those who went beyond upper secondary school. Figure 26: Poverty headcount ratio by individual characteristics Source: Authors’ calculations based on the Income and Expenditure Surveys for 2005/06 and 2010/11 and the Living Conditions Surveys for 2008/09 and 2014/15. The calculations are done using the LBPL. Participation in economic activities matters for poverty social protection transfers, which could be benefiting the reduction; the non-working or economically inactive unemployed or economically inactive. experience higher rates of poverty than those who are Considering poverty across different age groups active. The poverty rate among the economically inactive suggests poverty is highest among children below the was 46.3 percent in 2015, down from 57.7 percent in 2006. age of 15. Children up to age 5 consistently register the In comparison, the economically active registered a poverty highest poverty rates across all four periods, although falling rate of about 20.5 percent in 2015, down from 27.3 percent from 63.0 percent in 2006 to 52.6 percent in 2015. Children in 2006. The fall in the poverty rate was higher among the aged 6–14 had a poverty rate of about 50.5 percent in 2015, economically inactive (11 percentage points) compared to compared to 63.4 percent in 2006. Children up to age 14 the economically active (7 percentage points). This could constituted 30 percent of the entire population in 2015. The be a result of the poverty-reducing impact of government An Assessment of Drivers, Constraints and Opportunities 15 fastest decline in poverty was experienced by the elderly, of members to the household progressively increases the aged 65 and above, whose poverty rate fell by around 19 probability of being poor. percentage points between 2006 and 2015, possibly due to A profile of the poor shows a typical poor household government social transfers that targeted the elderly. as rural and headed by a single, economically inactive The more children a household has, the higher the female black South African. This is informed by statistical chances of being poor. Around 22.9 percent of the tests to examine the differences between poor and non- population with no child in the household was poor in poor households in 2015 to complement Figure 25 and 2015 following a decline from 36.9 percent in 2006. The Figure 26. The tests suggest that poor households are population living in households with at least three children, less likely to have heads who are employed in the formal on the other hand, had a poverty rate of 76.3 percent in sector and fewer adults employed in the formal sector. In 2015, compared to 88.9 percent in 2006. Although they terms of education, poor households have fewer heads constitute the largest proportion of the entire population, who have completed primary school, compared to non- the share of poor with no child declined by about 10 poor households. The average age of a household head percentage points between 2006 and 2015. is higher among the poor (51 years) compared to the non-poor (48 years). Poor households tend to be larger The larger the size of the household, the higher the (4.9 members) than non-poor households (2.8 members). incidence of poverty. This relationship is consistent across Poor households tend to have fewer adults than non-poor all years. For instance, in 2015, the poverty headcount households. Thus, the average number of children is higher ratio among the population of one-person households among the poor households compared to the non-poor. was 5.0 percent compared to a ratio of 67.6 percent for As expected, the profile of the bottom 40 percent of the households with at least seven members, who made up consumption distribution is very similar to that of the poor. around 31 percent of the population in 2015. Thus, addition Figure 27: Age-gender pyramid and poverty, 2015 Source: Authors’ calculations based on the Living Conditions Survey for 2014/15. 16 Overcoming Poverty and Inequality in South Africa The gendered and young face of poverty is evident Eastern Cape, KwaZulu-Natal, and Limpopo are the in the age-gender pyramid (Figure 27). Poverty is more poorest provinces. At 59 percent, Eastern Cape had the pronounced among females compared to males.8 While the highest poverty rate in 2015. Limpopo had the highest poverty incidence among the two groups is not strikingly poverty headcount ratio of 67 percent in 2006, about 72 different, especially in the lower ages, the poverty incidence percent in 2009, and 53 percent in 2011. Its poverty rate remains higher for women as age increases compared to in 2015 was 57 percent. Gauteng consistently has had the men. Further, the pyramid suggests both the population lowest poverty rate (19 percent in 2015) (Figure 28). All and poverty in South Africa have a predominantly young provinces experienced a reduction in poverty between face. This is reflected in a wide base of the population 2006 and 2015, using the LBPL. Mpumalanga recorded pyramid. the highest reduction in poverty levels, with the poverty rate falling from 60 percent to 43 percent between 2006 C. WHERE DO THE POOR LIVE? and 2015. Eastern Cape recorded the lowest reduction in poverty levels. Not only is Limpopo the poorest province i. Variation in poverty across provinces measured at the poverty headcount ratio, the depth and Rural areas have the highest poverty concentration. severity of poverty was the highest in three out of four In 2006, 60.3 percent of the poor were in rural areas. years, while it was the second highest in 2015. All provinces This decreased marginally to 59.7 percent in 2015. The except for Mpumalanga recorded an increase in poverty distribution of the population suggests the increased rural- between 2011 and 2015. This holds for all three poverty to-urban migration could be contributing to the decline in measures: poverty headcount ratio, poverty gap, and rural poverty, in addition to real reduction in poverty levels squared poverty gap. Mpumalanga is the only province observed nationally. The proportion of South Africans living that consistently recorded a decrease in poverty rates in rural areas fell from 41.0 in 2006 to 36.5 percent in 2015. across all the years. 8 This is estimated by assuming a person living in a poor household is poor. That is, household rather than individual welfare measures are used. Figure 28: Poverty headcount ratio by province Figure 29: Regional poverty decomposition, 2006 to 2015 Source: Authors’ calculations based on the Income and Expenditure Surveys for 2005/06 and 2010/11 and the Living Conditions Surveys for 2008/09 and 2014/15. Changes are calculated at the LBPL. An Assessment of Drivers, Constraints and Opportunities 17 KwaZulu-Natal drove the reduction in poverty rates of the poor did not change much between 2006 and 2015. between 2006 and 2015. Relative contributions of each KwaZulu-Natal accounts for the biggest share of the poor in province to aggregate poverty reduction between 2006 and the country, followed by Eastern Cape and then Gauteng. 2015 are reported in Figure 29. These “intra-sectoral effects” ii. Variation in poverty across municipalities are computed as the change in the poverty headcount ratio for each province between 2006 and 2015, multiplied A spatial representation of the poverty levels by its population share in 2006. About 21.5 percent of the supports the existence of pockets of poverty in some reduction in the national headcount ratio was due to gains municipalities, but also dispersion of municipality in KwaZulu-Natal, while 13.0 percent was due to poverty poverty rates in others. In 2011, extreme poverty— reduction gains in Gauteng. The contribution of Gauteng to measured at the food poverty line—was more prevalent aggregate poverty reduction is not only due to its poverty in peripheral areas of the North West, Limpopo, KwaZulu- reduction record (11 percentage point reduction between Natal, and Eastern Cape where the highest individual 2006 and 2015) but also due to the magnitude of its share poverty rates at the municipality level were found (Figure of the population (24.0 percent in 2015). 30Figure 23). Most of the 30 municipalities with the highest The aggregate contribution of shifts in population and the rates—from 28 percent to 63 percent of households living interaction effects between sectoral gains and population in extreme poverty—were in KwaZulu-Natal and Limpopo. shifts was also estimated. About 15.3 percent of the decline The 30 municipalities with the lowest household poverty in the national headcount ratio was due to population shifts rates were in Gauteng and Western Cape. between provinces. Keeping the provincial headcount In contrast, extreme poverty was highest in the ratios constant and considering only the changes in central and eastern parts of the country (Free State, provincial population shares, however, suggests poverty Eastern Cape, North West, and Northern Cape) in would have declined by only 1.7 percentage points. People 1996. Comparing the quantiles of 1996 and 2011 poverty most likely moved out of high-poverty into low-poverty maps at the municipality level, Figure 30 shows that provinces and the growth in the population of Gauteng the municipalities ranking highest for extreme poverty might reflect this. The negative interaction effect could be have faced a modest change over time in the northeast because the population was moving out of high-poverty of Northern Cape and the east of North West. Northern areas such as the Eastern Cape. Cape and Free State have seen a decrease in poverty rates. At 26.0 percent in 2015, KwaZulu-Natal had the largest Overall, Northern Cape and Free State have improved their share of the poor. This is partly due to the relatively high ranking in the poverty rate distribution at the municipality population share in KwaZulu-Natal, 19.9 percent in 2015, level. However, in general, poverty rates present a higher down from 21.0 percent in 2006. The pattern of distribution heterogeneity in the poorest quantile in 2011 than in 1996. 18 Overcoming Poverty and Inequality in South Africa Figure 30: Poverty incidence at the municipality level 1996 2011 Figure 31: Poverty density at the municipality level 1996 2011 Source: Poverty map calculations (map in the left) are from Alderman et al. (2002) and 2011 poverty map calculations (map in the right) are those of the authors. Notes: Darker lines correspond to provincial boundaries. Despite a change in the spatial distribution of the of the total number of poor people in the country. The poverty rates between 1996 and 2011, the spatial municipalities with the highest proportions of poor in both distribution of the poor did not change notably during years were in Limpopo, Gauteng, North West, and KwaZulu- this period. Figure 31 displays the poverty density in 1996 Natal. The west of Northern Cape and the south of Free and 2011. The labels at the left of each map correspond to State show a modest improvement in their poverty density the percentage of poor population living in the municipality rankings (see Box 2). An Assessment of Drivers, Constraints and Opportunities 19 Box 2: Estimating poverty at the municipality level Aggregating poverty levels at national and provincial levels is likely to understate extreme poverty within districts and thereby mask heterogeneity across subnational levels. To better understand the heterogeneities, a poverty mapping exercise was conducted. Using consumption data from the Income and Expenditure Survey (IES) 2010/11 and the geographical coverage of the Population Census 2011, a poverty map was constructed using the standard method developed by Elbers et al. (2002)—also known as the ELL (Elbers, Lanjouw, and Lanjouw 2002) method—and considering the suggestions of Tarozzi and Deaton (2009). To construct poverty estimates, detailed information on household expenditure or income are used to project welfare indicators into census records at geographical partitions not possible when using the IES. Thus, the results are expected to help inform provincial and local governments where policy implementation occurs and where information about the poor is needed. Poverty estimates were calculated for all 234 municipalities in the country. The FPL was used (R335 per person per month in March 2011). The focus on the FPL is consistent with policy emphasis on eliminating extreme poverty. There is a strong correlation between municipality- in 2011. Similarly, larger municipalities had lower poverty level poverty rates in 1996 and 2011. As presented in rates in both periods and poverty in these municipalities Figure 32, the higher the poverty rate a municipality had in fell. 1996, the more likely it was to also have higher poverty rates Figure 32: Comparison of municipality poverty rates, 1996 Figure 33: Dispersion and range in municipality and 2011 poverty rates, 1996 and 2011 Source: Authors’ calculations. 20 Overcoming Poverty and Inequality in South Africa The variation in poverty levels between municipalities often exceed their costs as reflected in levels of household is high and has been widening. Disparities in poverty expenditures on these items. Similarly, the social impacts levels across municipalities widened between 1996 of unemployment stretch beyond the observed income and 2011. As presented in Figure 33, the dispersion in loss to affecting the quality of life of concerned individuals. poverty rates between municipalities, expressed using the Considering non-money-metric measures of well-being is coefficient of variation, increased by 36.6 percent between especially important in South Africa given the government’s 1996 and 2011. In addition, the range, which measures use of the social wage—the redistributive elements of the difference between the richest and poorest municipalities, government budget that provide free basic services and was high and increased during this period. social protection—to increase access to basic services for the previously marginalized communities. D. NOTABLE PROGRESS HAS BEEN MADE IN The non-monetary indicators analyzed in this section REDUCING MULTIDIMENSIONAL POVERTY include access to basic services and utilities, education, SINCE THE END OF APARTHEID IN 1994 food security and malnutrition, and ownership of This section complements the preceding analyses by durable household assets. The choice of indicators is exploring levels and trends in non-monetary poverty influenced by availability of data and relevance to South and well-being during the period 1993–2016. The race- Africa. These indicators have been shown to improve based exclusionary policies of apartheid prevented most of livelihoods and thus are important dimensions of poverty. South Africa’s population from participating in meaningful Though comprehensive and aligned to the context of economic activities and accessing basic public services. South Africa, the indicators analyzed in this chapter is not This resulted in unequal distribution of resources, which led exhaustive. to high levels of poverty among marginalized groups. With i. Access to basic services and utilities the advent of democracy in 1994 came a strong need for transformation and redistribution of resources to address South Africa has made strides in broadening access the prevailing racial, spatial, and economic inequalities. This to basic public services since the end of apartheid. As resulted in policies such as the RDP, GEAR, AsgiSA, and is Figure 34 to Figure 37 show, the proportion of the population reflected in the current NDP as well, which advocates for with access to electricity, improved water sources, and “leaving no one behind” and aims to eradicate poverty and improved sanitation facilities increased steadily between reduce inequality by 2030. 1994 and 2015. In 2015, 93 percent of the population had access to improved water source compared to 83 percent It is important to go beyond monetary poverty in 1994. In 1994, 62 percent had access to electricity and measures, and track progress based on more this rose to 87 percent in 2014. In 2015, 66 percent of the comprehensive non-monetary dimensions that capture population had access to improved sanitation facilities, the multidimensionality of poverty. Money-metric following a 13 percentage point increase from 53 percent poverty measures have been criticized for being unable to in 1994. Comparing South Africa to other countries and capture the well-being impacts of use of services that are regions suggests that it lags average upper middle-income not transacted in markets. For example, outcomes related countries in all three basic public services, but it performs to educational attainment, health, water and sanitation, better than an average country in Sub-Saharan Africa. and food security affect well-being, yet their intrinsic values An Assessment of Drivers, Constraints and Opportunities 21 Figure 34: Changes in the proportion of the population with Figure 35: The proportion of the population with access to selected basic services access to electricity, comparison to other countries, 2014 Figure 36: The proportion of the population with access to Figure 37: The proportion of the population with an improved water source, comparison to other countries, access to improved sanitation facilities, comparison to 2015 other countries, 2015 Source: World Development Indicators. 22 Overcoming Poverty and Inequality in South Africa Access to basic public services is positively correlated of households, 54 percent had access to an improved with income, with access lowest among the poorest water source in 2015, 43 percentage points lower that the segments of the population. Figure 38 to Figure 40 proportion among the richest 10 percent. A focus on the present the proportion of the population with access to a poor shows a percentage of households with access to an selected service by per capita consumption decile, using improved water source of around 71 percent compared the LCS 2014/15 data. At 98 percent, the rates of connection to 95 percent of the non-poor. The same pattern holds for to the electricity supply among the richest decile are 20 access to an improved sanitation facility (Figure 40). These percentage points higher than the proportion among the patterns underscore poverty as a barrier to access to basic poorest decile (78 percent). Of the poor at the LBPL, 83 services and a contributor to and/or a result of resource percent had access to electricity in 2014 compared to 93 inequality. In addition, the patterns highlight the need for percent among the non-poor (Figure 38). the government to address the constraints (for example in terms of affordability or infrastructure) which limit access Access to an improved water source is uneven across by the poor. income groups (Figure 39). Of the poorest 10 percent Figure 38: The proportion of the Figure 39: The proportion of the Figure 40: The proportion of the population with access to electricity, population with access to an population with access to improved by decile, 2015 improved water source, by decile, sanitation facilities, by decile, 2015 2015 Source: Authors’ calculations based on the Living Conditions Survey for 2014/15. ii. Housing conditions, access to education, health, household is overcrowded. In 2015, about 39 percent of and assets the population was defined as being overcrowded. The poor had an overcrowding headcount rate of 60.8 percent, The poor tend to live in overcrowded housing which is high compared to 23.6 percent among the non- conditions. Living in overcrowded conditions has been poor (Figure 41). Overcrowding rates are shown to fall linked to worsening of health and education outcomes with income levels. The overcrowding rate for the bottom (see, for example, Leventhal and Newman 2010 and 10 percent was 67.9 percentage points higher than for Lund et al. 2010) and thus is a good indicator of poverty. the top 10 percent. This suggests that use of persons per The number of persons per bedroom in a dwelling unit bedroom is a reliable indicator of deprivation caused by is used here to measure overcrowding. A two persons- low consumption expenditure.9 per-bedroom standard is applied to determine whether a 9 No direction of causality is implied: the analysis focuses on correla- tions rather than causal relationships. An Assessment of Drivers, Constraints and Opportunities 23 Educational outcomes are uneven across consumption that indicated the nearest hospital was more than 20 expenditure groups, in favor of rich households. This kilometers from their dwelling unit. For the poorest decile, is revealed in Figure 42, which shows the proportion of 33.8 percent lived at least 20 kilometers away from a hospital, South Africans older than 25 that had completed primary 27 percentage points higher than the proportion among education in 2015. Among individuals in the top 10 the richest decile. Consistent with this, poor individuals consumption decile, the proportion who had completed lived farther away from a hospital compared to the non- primary school was 35.4 percentage points higher than poor. As expected, asset ownership indexes were higher the proportion for the bottom 10 percent. Of individuals among richer households (Figure 44). In 2015, the richest older than 25 among the poor 53.4 percent had completed decile had an average of 19 out of 36 asset types, which primary school compared to 72.9 percent among the non- was close to three times that of the poorest decile (details poor. of how the assets indexes were constructed are in Box 3). Household ownership of physical assets is frequently used Access to health and assets is uneven across income to examine the welfare status of households insofar as they groups. The rich have better access to hospitals than the capture material deprivation. poor. Using distance to the nearest hospital as an indicator of access, Figure 43 shows the proportion of South Africans 24 Overcoming Poverty and Inequality in South Africa Figure 41: Overcrowding headcount rate, by decile, Figure 42: The proportion of the population older 2015 than 25 with primary school education, by decile, 2015 Figure 43: The proportion of the population for whom Figure 44: Asset ownership, by decile, 2015 distance to nearest hospital is at least 20 kilometers, by decile, 2015 Source: Authors’ calculations based on Living Conditions Survey for 2014/15. iii. Food security and malnutrition in Box 3, show a modest increase in food insecurity since 2012. In addition, measures of child malnutrition based on Food insecurity, stunting, and child malnutrition remain anthropometric data show little improvement and may challenges in South Africa and have deteriorated since even have worsened in recent years. 2012. All components of the Household Food Insecurity Access Scale (HFIAS), the construction of which is described An Assessment of Drivers, Constraints and Opportunities 25 Box 3: Construction of an asset index and the Household Food Insecurity Access Scale Construction of the asset index. The asset index is constructed by counting the number of asset types a household owns from a specified set of durable assets. A set of 36 assets was identified in the 2015 dataset and used in this analysis. The assists are radio; stereo/HiFi; satellite TV; television; DVD/Blu-ray player; deep freezer-free standing; refrigerator/combined fridge freezer; stove; microwave oven; dishwasher; washing machine; tumble dryer; vacuum cleaner; hot water heater; kitchen furniture; dining room furniture; bedroom furniture; lounge furniture; desktop computer; laptop/notebook/netbook; tablets; camera; cellular telephone; telephone; connection to the internet; motor vehicle; motorcycle/scooter; bicycle; canoe/boat; generator; power-driven tools; plow; tractor; grinding mill; wheelbarrow; bed (base set and mattress). For each durable asset, a dummy variable was created that takes the value of one if a household owns at least one of that item and zero otherwise. The total asset ownership index for each household was computed by adding up the dummy variables. Given that the set being analyzed comprises 36 items, the index ranges from zero (none of the items) to 36 (at least one of each item). A household owning 10 out of the 36 items, for example, gets a score of 10. Construction of the HFIAS. The General Household Survey (GHS) has seven questions related to hunger and food availability that are used to generate eight variables on food security (GHS report 2015). These questions specifically seek to establish if any member of the household has gone without food, skipped meals, eaten a smaller variety of food, or cut meal sizes. These questions also have a component that establishes the frequency of occurrence of any of those situations. For example: “For the past 12 months did any adult (18 years and above) in this household go without food?” (GHS 2015 Questionnaire, page 41). Responses are on a five-point scale from never to always. In line with the HFIAS methodology, all eight variables that measure occurrence and intensity are used. In replicating Stats SA, every affirmative answer to a food insecurity question was scored one and non-affirmative zero. The index is then generated as an additive index of the scores. Categories are then created on the following basis: a score of 0–1 reflects adequate food security, 2–6 is considered inadequate, and 7–8 is severely inadequate. This approach is applied to ensure coherence between the index created and the Stats SA index. Food insecurity is gendered and more prevalent forms of deprivation in South Africa, black South African among the black African population. Consistent with households are most likely to be food insecure followed by other forms of deprivation, women are more likely to be colored households (Figure 45). poor and go hungry compared to men. As with all other Figure 45: Food security index by household characteristics Source: Authors’ calculations based on GHSs for 2012–2015. 26 Overcoming Poverty and Inequality in South Africa Food security has a clear spatial dimension, with tribal informal (32 percent) and rural formal (37 percent) localities. areas recording the highest level of food insecurity In terms of differences by age, each year added to the age compared to urban and farm areas. These patterns are of a head of household increases the likelihood of food like those found by the South African National Health and security, but this is a quadratic relationship. Increments in Nutrition Examination Survey (SANHANES) 2012, which age eventually increase the likelihood of a household being reported that the largest percentage of participants who food insecure. experienced hunger (food insecurity) in 2012 were in urban Figure 46: Food insecurity index by quintiles of asset index Figure 47: Gender disaggregated stunting rates in (percent) children under five Source: Authors’ calculations using GHS 2012–2015. Source: Authors’ calculations using NIDS wave 4. The poor bear the brunt of food insecurity: while incomes that were significantly higher than households in most income groups experienced a decline in food the other groups. Not only was food security lower among security between 2012 and 2015, the poorest quintile the poor, inequalities in food security exist, which generally experienced the largest deterioration (Figure 46).10 favored the rich. Households in the poorest quintile recorded the highest People practicing subsistence agriculture have level of both severe and moderate food insecurity in all years. higher rates of food security. Given the declining role This decreased for each progressive quintile until it reached of agriculture in the South African economy and the its lowest level in the richest quintile in which less than low prevalence of smallholder agricultural production 10 percent of households had inadequate food security. in compared to other African countries, it is striking that The greatest increases in the food security index were in households that engaged in some form of subsistence the middle quintiles with the richest quintile experiencing agriculture were more likely to be food secure than those only modest change. A consistent pattern is revealed when that did not. This suggests that interventions that support the mean per capita monthly income of households in the the production of food would be appropriate, even if these different food insecurity bands is computed. In all years, are not a central component of government food security households with adequate food security had per capita and nutrition strategies. 10 The quintiles are based on an asset index created using Principal Com- Another aspect of food insecurity that is important, ponents Analysis on asset variables available in the GHS data. The in- dex measures the socioeconomic status of households. in addition to hunger and the quantity of foods An Assessment of Drivers, Constraints and Opportunities 27 consumed, is dietary diversity. Various studies have Multiple factors predefine malnutrition including shown that South Africa has low dietary diversity levels. poverty status, mother’s food security, mother’s own Many households consume diets that are energy dense condition, and access to health care. Investigation of and lack micronutrients that are needed for proper growth the prevalence and determinants of malnutrition among and development in children. For instance, Labadarios, children under age five found that sex and age of child, Steyn, and Nel (2011) found low dietary diversity, which employment, body mass index (BMI) of mother, age of was characterized by limited eggs, legumes, and fruits mother, height of mother, and household incomes are and vegetables rich in vitamin A. Faber, Wenhold, and significant determinants of malnutrition. The study also Laurie (2015) supported this and further highlighted the shows that a mother’s height is directly associated with association between dietary diversity and household food child malnutrition regardless of BMI or weight category. The security, with food secure households having a higher mean implication is that women who may themselves have been dietary diversity score. Further, these studies highlighted stunted are likely to give birth to children who become the spatial dimensions of food insecurity and low dietary stunted. This situation reflects the cumulative effects of diversity. The provinces with the highest prevalence of poor socioeconomic, environmental, health, and nutritional dietary diversity are Limpopo and the Eastern Cape while conditions. However, these levels and trends vary by the Western Cape has a low score. South Africans in rural economic status of households. Also, previous studies and informal urban areas tend to be the worst affected. suggest that other contributors to malnutrition include micronutrient deficiencies arising from unhealthy diets, Malnutrition is linked to the physical environment in low birth weight of children due to maternal ill health, and which people live, inadequate and unsafe water, poor the impact of repeated enteric infections arising from poor sanitation, and unsafe hygiene practices are the main sanitation conditions. causes of infections of the intestinal tract. Multivariate analysis reveals that people living in informal dwellings South Africa already has several important initiatives to are more likely to be food insecure than those living in address food insecurity and malnutrition. This includes informal houses. Further, people living in urban areas face mandatory fortification of staple foods, the provision of a significant threat of food insecurity. This confirms that food supplements for mothers and children, as well as malnutrition is linked to the physical environment in which the social protection programs such as the child support people live, especially children. Improved sanitation and grant (CSG) and school feeding program. Except for the hygiene and access to safe water can reduce the frequency CSG, poor implementation has been identified as a reason and severity of infections of children and pregnant women, these programs have not performed as well as anticipated. including diarrheal diseases. Ingestion of feces and soil For the CSG, leakages of the grant to other household contribute to the risk in polluted environments, such as members, and the small value of the grant (relative to other dense shack settlements where human overcrowding and grants and the costs of nutrition), have been identified as animals are present. possible reasons why malnutrition has not declined despite more than 12 million children having access to the grants. Stunting remains a problem, with boys and younger Other policies focus on increasing the availability of food, children at higher risk (Figure 47). Of all the forms of including those of the national Department of Agriculture, malnutrition examined, stunting remained unusually high. Forestry, and Fisheries (DAFF), such as garden projects that Additionally, stunting is more prevalent in male children are implemented by the provincial departments. These than in female children at all ages and younger children targets both rural and urban food security by supporting are at a higher risk of malnutrition than older children. High urban agriculture, community food projects, household stunting rates are a cause for concern because the higher food production, new gardens, and rehabilitating stunting rates of younger children today are likely to result abandoned projects. Further, the Department of Public in even higher stunting rates when these children become Works offers food-for-work programs for unemployed older. 28 Overcoming Poverty and Inequality in South Africa persons in addition to the Community Works Program and household experiences with respect to health, education, the Expanded Public Works Program (EPWP). By identifying and living standards. It allows for comparisons within the food insecure and including them in such income- regions, countries, and areas/provinces within countries. It generating programs the self-provisioning of food can be allows for the identification of the most deprived. enhanced. The SAMPI was chosen for its ability to provide an integrated iv. The South African Multidimensional Poverty picture that could help assess the impact of government Index programs to achieve poverty reduction wherein the index incorporates basic services, education, living standards, This section describes the non-monetary poverty levels health, and economic activities as highlighted in the NDP. in South Africa for the period 2001–2016 using the South Furthermore, its key attribute of being decomposable by African Multidimensional Poverty Index (SAMPI). The SAMPI space and population attributes makes it a powerful tool for uses the Alkire-Foster method (Box 4) and builds on the not only identifying who the poor are and where they are global Multidimensional Poverty Index (MPI) developed but also for guiding targeted policy interventions on what by the Oxford Poverty and Human Development Initiative contributes to poverty in those areas so that resources can (OPHI) and the UNDP to measure acute poverty. The be channeled properly. MPI captures severe deprivations that each person or Box 4: The Alkire-Foster method To explore the nature and extent of multidimensional poverty in South Africa, a “counting” approach developed by Alkire and Foster (2011) is used to estimate the SAMPI. The approach complements monetary measures of poverty by identifying and counting the number of overlapping deprivations experienced simultaneously by an individual or household. It is built on three premises: the selection of the dimensions and indicators of poverty; the identification of the poor based on set criteria, which involves setting cut-offs or poverty lines against which the poverty/deprivation status is determined; and the aggregation of information through a poverty index. Stats SA used four guiding principles during SAMPI construction: the Global MPI and its dimensions and indicators; the country context and issues affecting poverty; the availability of data items in censuses; and the suitability and robustness of these data after data exploration, confrontation, and consultation. Given the desire to domesticate the Global MPI to be anchored in the South African context, it was impossible to ignore the country’s massive unemployment challenge. According to the Quarterly Labour Force Survey for the fourth quarter of 2017, unemployment stood at 26.7 percent. Hence, a fourth dimension dealing with unemployment was added to the three standard dimensions already present in the Global MPI. While there is obviously a monetary element to employment, the SAMPI embraced a more social dimension in its measurement and adopted a deprivation cut-off that represented an extreme situation that is unhealthy for the social development of the household. A household is considered deprived in this dimension if all adults in the economically active age cohort (ages 15 to 64) are unemployed using the expanded definition of unemployment (which includes those defined as unemployed as well as discouraged work-seekers). If there are any adults who are not economically active, such as still in education, retired, or looking after the home, they would not be defined as unemployed and, therefore, the household would not be classified as deprived in this indicator. Therefore, someone who simply lacks a job does not necessarily qualify as deprived (even if by implication it does have a significant bearing on the money-metric poverty status of a household), but rather, this indicator aims to measure the totality of the unemployment situation in a household. Thus, the consequences of being deprived in this indicator manifests in a much more significant way that transcends the simple loss of income. Ultimately, this dynamic of no employed adults in the household seriously compromises the social fabric of the household. An Assessment of Drivers, Constraints and Opportunities 29 Thus, the SAMPI was customized to suit the context of South Africa. While the Global MPI consists of three dimensions and 10 indicators, the SAMPI comprises four dimensions and 11 indicators. As Table 3 indicates, equal weights across dimensions is assumed, along with equal weights across indicators within each indicator. The data sources for the analysis are the 2001 and 2011 Population Census data as well as the 2016 Community Survey data. Table 3: SAMPI dimensions, indicators, and deprivation cut-off points Dimension Indicator Deprivation cut-off Weight Health Child mortality If any child under age 5 has died in the past 12 months 1/4 If no household member age 15 or older has completed 5 1/8 Years of schooling Education years of schooling School attendance If any school-aged child (ages 7 to 15) is out of school 1/8 Fuel for lighting If household is using paraffin/candles/nothing/other 1/28 Fuel for heating If household is using paraffin/wood/coal/dung/other/none 1/28 Fuel for cooking If household is using paraffin/wood/coal/dung/other/none 1/28 Standard of Water access If no piped water in dwelling or on stand 1/28 living Sanitation type If no flush toilet 1/28 Dwelling type If an informal shack/traditional dwelling/caravan/tent/other 1/28 If household does not own more than one of radio, televi- 1/28 Asset ownership sion, telephone, or refrigerator and does not own a car Economic If all adults (ages 15 to 64) in the household are unem- 1/4 Unemployment activity ployed Source: Authors’ representations. v. Changes in multidimensional poverty at the stagnated between 2011 and 2016. The improvement national level between 2001 and 2011 could reflect, in part, the positive impact of redistribution programs on multidimensional South Africa recorded a notable decline in poverty. These programs include, for example, compulsory multidimensional poverty between 2001 and 2016, education for children aged 7 to 15, no-fee schools, driven by a decline in the proportion of households feeding schemes, access to free basic services for indigent that were multidimensionally poor. In 2001, 17.9 households, and social grants. The stagnation between percent of South Africans were multidimensionally poor; 2011 and 2016 is consistent with the trend in monetary this dropped to 7.0 percent in 2016 (Table 4). The major poverty headcount ratio, which showed a notable decline reduction occurred between 2001 and 2011, with the up to 2011 but an increase after 2011. multidimensional poverty headcount falling by almost 10 percentage points. Sadly, multidimensional poverty 30 Overcoming Poverty and Inequality in South Africa Table 4: Multidimensional poverty at national level Year Headcount (H) Intensity (A) SAMPI (HxA) 2001 17.9% 43.9% 0.08 2011 8.0% 42.3% 0.03 2016 7.0% 42.8% 0.03 Source: Authors’ calculations based on the Population Censuses for 2001 and 2011 as well as the Community Survey 2016. The reduction in the intensity of multidimensional in South Africa. Figure 48 shows the extent to which poverty, which measures the average proportion of each indicator contributed to multidimensional poverty indicators in which multidimensionally poor people in the three years considered. While the contribution to are deprived, has been slower compared to the multidimensional poverty of most indicators decreased reduction in the proportion of the multidimensionally between 2001 and 2016, the contribution of unemployment poor. The intensity of poverty fell marginally from 43.9 increased. The contribution of unemployment to the percent in 2001 to 42.3 percent in 2011 virtually stagnating SAMPI increased from 2001 to 2016. This underscores the at 42.8 percent in 2016. The slow reduction of intensity importance of job creation in reducing multidimensional of multidimensional poverty indicates that while the poverty in South Africa. The reduction in the contribution proportion of multidimensionally poor households fell, the of the education indicators and the living standards circumstances of the poor hardly got better. indicators, on the other hand, points to an improvement in service delivery and as well as the education profile of Unemployment dampens progress toward the country. This may be due to programs and policies such reducing multidimensional poverty in South Africa. as no-fee schools, compulsory education, and free basic Unemployment and education (years of schooling) remain services for indigent households, among initiatives. the top two contributors to multidimensional poverty Figure 48: Contribution of weighted indicators to SAMPI at national level Source: Authors’ calculations based on the Population Censuses for 2001 and 2011 as well as the Community Survey 2016. An Assessment of Drivers, Constraints and Opportunities 31 vi. Multidimensional Poverty Index, headcount and hides the worsening situation of the multidimensionally intensity: spatial variation poor. The result, coupled with the finding that Gauteng had the lowest monetary poverty in South Africa in 2015, Computation of multidimensional poverty at the suggests better performing provinces do have pockets of provincial level shows that the Eastern Cape had the intense multidimensional poverty. highest multidimensional poverty headcount ratio in 2016 at 12.7 percent, followed by Limpopo at about 11.5 All provinces experienced a steady reduction in the percent (Figure 49). The Eastern Cape also has the highest multidimensional poverty headcount ratio between MPI score, alongside Limpopo, driven by relatively high 2001 and 2016. However, the multidimensional poverty multidimensional poverty headcount ratios. Interestingly, headcount ratio in Limpopo increased from 10.1 percent at 4.6 percent, Gauteng had the least multidimensional in 2011 to about 11.5 percent in 2016. Seven out of nine poverty headcount in 2016 but has the highest intensity provinces experienced either an increase or near stagnation of multidimensional poverty. This is of policy relevance in the intensity of poverty between 2011 and 2016. Only as it supports caution around formulating policies or the Free State and Western Cape registered a clear, though interventions based only on the poverty headcount ratio. modest, reduction in intensity of multidimensional poverty. The multidimensional poverty headcount ratio in this case Figure 49: Multidimensional poverty measures at provincial level Headcount ratio Intensity MPI Source: Authors’ calculations based on the Population Censuses for 2001 and 2011 as well as the Community Survey 2016. Analysis at the district level suggests multidimensional and the 18 rural nodes11 that were selected in 2001 for poverty in 2016 was highest in the Alfred Nzo district accelerated development under the Integrated Sustainable municipality in the Eastern Cape followed by the OR Rural Development Program (ISRDP). Evaluating what has Tambo district municipality. Amathole district was the transpired in these 18 nodes in terms of poverty since 2001 third-poorest district (Figure 50). Important to note is the would inform efforts to accelerate poverty reduction. comparison between the poorest district municipalities 11 These areas were earmarked for accelerated development under the Integrated Sustainable Rural Development Programme (ISRDP). For details, see Statistics South Africa (2016b), “Quest for nodal develop- ment: Evidence from Census 2001 and Census 2011.” South Africa 32 Overcoming Poverty and Inequality in South Africa Figure 50: Poorest and richest districts and local municipalities in South Africa in 2016 Top and bottom 10 districts Top and bottom 20 municipalities Source: Poverty Map calculations are from Statistics South Africa. Multidimensional poverty is revealed to be higher Most municipalities in the 20 poorest local in rural areas compared to urban areas. Focusing on municipalities in 2016 were in the Eastern Cape, the 20 poorest districts in 2016 shows that the majority Limpopo, and KwaZulu-Natal. Fifteen of the 20 poorest (15 out of 20) are in rural nodes (Figure 50). Fourteen of municipalities are in the Eastern Cape; four (Msinga, the 18 rural nodes are on the list of 20 poorest districts. uMhlabuyalingana, Maphumulo, and Mzumbe) are in Dr. Ruth Segomotsi Mompati, Vhembe, Ngaka Modiri KwaZulu-Natal; and the remaining municipality, Mutale, is Molema, Ilembe, Waterberg, and Bojanala Platinum district in Limpopo. It should also be noted that all the 10 poorest municipalities are worse off compared to other areas municipalities are in the former homelands of Eastern Cape that were selected for accelerated development, such as and KwaZulu-Natal, highlighting the enduring legacy of Central Karoo, Thabo Mofutsanyane, uMgungundlovu, and apartheid. The richest 20 municipalities consist mainly of Ehlanzeni district municipalities. This suggests that the list municipalities in the Western Cape (15 out of 20). These of areas earmarked for accelerated development needs to patterns are illustrated in the SAMPI maps for 2001, 2011 be reconsidered. and 2016, which suggest areas that were disadvantaged under apartheid still have the highest multidimensional poverty levels (Figure 51). Figure 51: Multidimensional poverty headcount ratio at the municipality level 2001 2011 2016 Source: Poverty Map calculations are from Statistics South Africa. An Assessment of Drivers, Constraints and Opportunities 33 vii. Multidimensional deprivation conditions, incomplete primary school, lack of access to an improved sanitation facility or improved drinking water At least 4 percent of the monetarily poor were affected source, and others (Figure 52). by an additional deprivation in 2015. Among those depravations were lack of assets, overcrowded housing Figure 52: Deprivations affecting the poor in 2015 Source: Authors’ calculations based on Living Conditions Survey for 2014/15. The monetary poor are simultaneously deprived in spans—at present—6 years between 2008 and 2014/15. multiple dimensions. In 2015, 3.7 percent of the monetarily The survey is held every two years, for a total of four waves poor lived in overcrowded housing and had no connection from 2008 to 2014/15. NIDS collects information on four to electricity supply. The share of monetarily poor with no modules: income, expenditure, assets, and debts. Data on access to improved water and sanitation facilities was 5.2 income and expenditure was collected in all four waves, percent. The proportion of the monetarily poor that were while wealth (defined as assets less debts) information was food insecure and asset deprived was 4.2 percent. The collected only in waves 2 and 4. NIDS holds two advantages highest proportion of simultaneously deprived households compared to IES. One, it has more detailed labor market was 5.7 percent for households that were monetarily poor, information in addition to labor market status (whether lived more than 20 kilometers from the nearest hospital, a respondent is employed, unemployed, or inactive) as it and had not completed primary school. collects information on the sector and occupation if the respondent is employed. Two, NIDS collects information E. ECONOMIC MOBILITY: TRANSITIONING on the education and work status of parents. These FROM CHRONIC POVERTY TO MIDDLE characteristics have been shown to be extremely influential CLASS in determining equity of opportunity. They thus form a key part of the empirical analysis. NIDS data is used in this section. NIDS is a multi-year dataset aimed at gathering information over a panel of This section analyzes NIDS data to provide a dynamic households in South Africa. Implemented by the South perspective on the experience of poverty in South Africa, African Labor and Development Research Unit at the aiming to deepen understanding of the extent and University of Cape Town School of Economics, this survey the determinants of transitions into and out of poverty. 34 Overcoming Poverty and Inequality in South Africa Transition matrices provide a basic understanding of escape poverty in the next wave. Moreover, about a quarter the degree of economic mobility, duration of poverty of those with a per capita expenditure above the UBPL in a spells, and intertemporal consumption averages that given wave fell into poverty in the next wave. decompose standard poverty measures into chronic and Table 5: Poverty transition matrices for South Africa, transient components. Subsequently, a model of poverty 2008-2014/15 (pooled 4 waves panel) transitions is used to examine the individual and household t (destination) characteristics associated with observed mobility patterns. Five social classes are defined based on their probability of Poor Non-Poor falling into poverty: chronic poor, transient poor, vulnerable, t-1 (origin) Poor 82.7 17.3 middle class, and elite. Finally, the section profiles the relative size, growth, racial composition, and other demographic Non-Poor 24.8 75.2 characteristics of the classes, as well as their geographic location, labor market resources, and mobility patterns. Source: Authors’ calculations using NIDS waves 1 to 4 pooled panel of wave-to-wave transitions (weights corrected for panel attrition). The analysis uses the UBPL—set at ZAR992 per person per Calculations done using the UBPL. month in 2015 prices—as it is deemed more realistic in the Chronic poverty is the dominant contributor to total context of the focus on social classes. poverty, accounting for more than 80 percent of the i. Poverty transitions, chronic poverty, and upper bound poverty rate. Applying the two approaches characteristics presented in Box 5 to South Africa suggests that between 80 to 90 percent of the poor, using the UBPL, can be Poverty is persistent in South Africa. Table 5 presents classified as chronically poor. That is, for a large share of the four sets of poverty transition matrices using the UBPL for population, poverty is a permanent state. The share of the the period 2008–2014/15, based on the pooled sample of transient poor tended to be highest in 2010/11, when— wave-to-wave transitions. A sizable proportion of those likely due to the global economic crisis—some households living below the UBPL in a given wave of the survey did not were temporarily pushed below the poverty line. Box 5: Estimating chronic and transient poverty Two approaches have been used in the literature to decompose poverty at one time into a long-run chronic component and a short-run transient component. The components approach, developed by Jalan and Ravallion (1998), calculates the permanent component of household income (or consumption expenditure) by taking the intertemporal average. The chronically poor are then identified as those for whom this component falls below the poverty line. The spells approach, accounts more explicitly for the time spent in poverty by counting the number of poverty spells experienced over a given number of periods and defining a duration cut-off above which households are classified as chronically poor (Bane and Ellwood 1986, Calvo and Dercon 2009, Foster 2009). Seventy-eight percent of South Africans were in Another 39 percent fell into poverty at least once during poverty at least once during the 2008–2014/15 period. the 2008–2014/15 period (Figure 53). Figure 54 shows Thirty-nine percent of the South African population, 21.7 that the chronically poor tend to be dependent more on million people, were poor in all periods of the analysis. government social grants and less on labor market incomes. An Assessment of Drivers, Constraints and Opportunities 35 Figure 53: Poverty duration, 2008–2015 Figure 54: Income source by duration in poverty Source: Authors’ calculations based on NIDS data, 2008-2014/5. Notes: Calculations based on the UBPL. ii. The scope of social classes in South Africa are currently being met but who face above-average risks of slipping into poverty; the middle class, who are in a The concept of a middle class has been broadly discussed in better position to maintain a non-poor standard of living socioeconomic literature and policy debates in South Africa even in the event of negative shocks; and the elite, whose and abroad. Empirical evidence suggests that countries living standards situate them far above the average. with a larger share and faster growth in the middle class are associated with better reforms and governance. As people Only one in four South Africans can be considered gain middle-class status, they tend to accumulate savings stably middle class or elite, whereas the other three and acquire secondary and tertiary education investments are either poor or face an elevated risk of falling into in the future. Members of the middle class are likely to poverty (Figure 55). The size of the middle class is thus support accountable government and the rule of law. This considerably smaller, and growth has been more sluggish group acquires higher levels of education, consumes high- than suggested by other studies (Box 6). Moreover, about quality goods and services, and fosters economic stability. 14 percent of the population is in the vulnerable group. Faster growth and poverty reduction is associated with the That is, a substantial share of the non-poor still faced a appearance and growth of the middle class. considerable risk of falling into poverty. Among the poor, about 80 percent could be considered chronically poor This section defines and analyzes the middle class in (accounting for half of the South African population), South Africa based on the four waves of the NIDS survey. whereas the remaining 20 percent of the poor (accounting A conceptual framework is described in the background for 13 percent of the population) could be classified as note12. The framework proposes a multilayered class model transient poor. At 20 percent of the population, the share that differentiates five social classes: the chronic poor, of the middle class in South Africa is relatively small. For characterized by high poverty persistence; the transient example, close to 80 percent of the population of Mauritius poor, who have above-average chances of escaping could be classified as middle class. poverty; the non-poor but vulnerable, whose basic needs 12 Schotte S, Zizzamia, R. and Leibbrandt M. (2017). Assessing the Extent and Nature of Chronic Poverty, Poverty Dynamics, and Vulnerability to Poverty in South Africa, forthcoming. 36 Overcoming Poverty and Inequality in South Africa Box 6: Defining the scope of middle class in South Africa In face of the ambitious hopes placed on the middle class as torchbearers of both democracy and long-term economic growth, it is little wonder that upbeat stories about a rapidly expanding new middle class in Africa (AfDB 2011) have been excitedly embraced by the business community, policymakers, and the media (Giesbert and Schotte 2016). The conceptual contribution consists of proposing a class schema with particular relevance for the emerging and developing country context marked by high economic insecurity. The method is based on López-Calva and Ortiz-Juarez 2014. Following Cappellari and Jenkins (2002, 2004, 2008), the analysis uses a multivariate regression model that explicitly allows for possible feedback effects from past poverty experiences and accounts for potential endogeneity of initial conditions, unobserved heterogeneity, and non-random panel attrition—four factors insufficiently addressed in existing studies when estimating poverty risks. Details of the methodology are presented in the background paper to this report. Households were classified as being poor versus non-poor using the UBPL set at R992 (in January 2015 prices) per person per month, equivalent to about US$5.5 a day (in 2011 PPP). The multivariate model of poverty transitions is fitted to four waves of panel data from the NIDS covering the period 2008–2014/15. Figure 55: Class sizes, 2008–2014/15 Figure 56: Income by sources, classes Source: Authors’ calculations using NIDS waves 1–4 pooled sample (post-stratified weights corrected for panel attrition). iii. The profile of social classes and factors associated derive more than half of their income from government with escaping chronic poverty social grants (Figure 56). By comparison, social grants make up one-fourth of the income of the transient poor and This section profiles the five social classes and identifies one-fifth of the income of the vulnerable. In comparison, factors associated with the probability of escaping chronic 7 percent of total household income of the middle class poverty.13 is derived from grants. Those who remain stably out of The relative importance of social grants in the lives of poverty rely heavily on labor income. the poor remains significant. Specifically, the chronic poor The chronic poor are deprived in multiple dimensions. 13 The factors associated with escaping chronic poverty are assessed us- Unsurprisingly, those who are poor in multiple periods are ing panel regression. The dependent variable of the regression is the probability that the household will get out of chronic poverty and not also relatively more deprived in their access to basic goods fall back into poverty afterwards. The explanatory variables included and services. Only 21 percent of the chronic poor had demographic characteristics, location, labor, education and skills, and changes in these variables. access to electricity, flowing water, a flushable toilet, and An Assessment of Drivers, Constraints and Opportunities 37 formal housing, while close to 77 percent of the never poor and the second smallest middle class (after Limpopo). At the had access to all of those assets. same time, KwaZulu-Natal has the fourth largest elite (after Gauteng, the Western Cape, and Mpumalanga), indicating Social classes have a strong geographical split in South a substantial degree of local social inequality. Chronic Africa (Figure 57). Nine in 10 of South Africans who did not poverty is lowest in the Western Cape and in Gauteng— experience poverty between 2008 and 2014/15 were urban which also have the strongest middle class and elite. While dwellers as compared to four out of 10 among the chronic vulnerability is substantial in all provinces, including those poor. Of the transient poor, 17 percent resided in traditional with low levels of chronic poverty, a negative relationship areas. Similarly, about 27 percent of the vulnerable lived in between the extent of chronic and transient poverty across traditional areas, compared to 5 percent of the middle class. the provinces is observed. KwaZulu-Natal has the highest incidence of chronic poverty Figure 57: Geographic distribution of South Africa’s five social classes, 2008–2014/15 Source: Authors’ calculations using NIDS waves 1–4 pooled sample (post-stratified weights). The provinces with the highest average propensity closely to the geographic location of former homelands. It is to poverty are KwaZulu-Natal, Eastern Cape, and apparent that the apartheid legacy is still most strongly felt Limpopo (Figure 58, panel a). These provinces contain in these severely underdeveloped traditional areas, which most of the former Bantustans. This is clearer in Figure 58 remain poorly integrated into the South African economy. (panel b), where the darkest areas of the map correspond 38 Overcoming Poverty and Inequality in South Africa Figure 58: Pockets of high propensity to poverty in South Africa, 2014/15 a. Probability of remaining poor or falling into poverty, b. Probability of remaining poor or falling into poverty, at the province level at the district level Source: Authors’ calculations using NIDS waves 1–4 pooled sample (post-stratified weights). Notes: In panel b, only districts with at least 400 respondents in NIDS wave 4 were kept. At 42 years of age on average, household heads in black and colored South Africans. These two groups also the vulnerable class tend to be younger than those in constitute most of the transient poor and the vulnerable. the other classes, which may be associated with a less However, colored South Africans seem to be more heavily stable position in the labor market. At 50 years of age, concentrated among the transient poor (though this household heads tend to be the oldest among those living lower chance to be persistently poor was not statistically in chronic poverty. This may link to formation of larger, significant in the regression results) and the stable middle intergenerational households, where adult children or class, facing lower risks of downward mobility. Although grandchildren co-reside with (grand)parents receiving old black South Africans also constitute the largest proportion age pensions (see Klasen and Woolard, 2009). Furthermore, of the middle class—with a growing trend in recent years seven out of 10 chronically poor individuals live in as illustrated in Figure 59—their share among the two top households with a female head, compared to five to six groups remains far from demographic retrospectivity. That out of 10 among the transient poor and vulnerable classes, is, while black South Africans make up about 80 percent of and three out of 10 among the middle class and elite. This the total population, in 2014/15 they made up just above reflects the higher incidence of poverty and vulnerability to 50 percent of the middle class. On the other hand, while poverty among single mothers in South Africa. whites constitute a mere 10 percent of the South African Race is a strong predictor of poverty, and the population, almost one in three members of the middle chronically poor group is almost exclusively made up of class and two in three members of the elite are white. An Assessment of Drivers, Constraints and Opportunities 39 Figure 59: Racial composition of South Africa’s five social classes, 2008 and 2014/15 Source: Authors’ calculations using NIDS waves 1–4 pooled sample (post-stratified weights). There is a strong relationship between the educational inactive or unemployed. Among the transient poor and attainment of household heads and the incidence and the vulnerable, about 50 percent are employed. This figure persistence of poverty (similar patterns are observed rises substantially for the middle class and elite. About 80 at individual education levels). Given that higher levels percent of the household heads in these two classes are of education tend to be accompanied by a lower risk of economically active and the employment rate is above 75 poverty, heads of chronically poor households are the percent. Overall, employment of any household member least educated, with no more than five years of primary raises significantly the probability that the household will education, while the transient poor and the vulnerable escape extreme poverty, and getting a skilled job further tend to have some secondary education. A household significantly increases the probability. Those who have head in the middle class generally has completed remained out of poverty live in households with heads who secondary schooling, while those in the elite tend to have are more likely to actively participate in the labor market, some tertiary education. Of those who did not experience and of those who participate, a substantially larger share a single poverty spell between 2008 and 2014/15, 93 are employed. percent lived in households with a head who attained at The employed can be categorized into five types of least secondary schooling. Of those, two-thirds had either economic activity: subsistence agriculture (accounting completed secondary education or even attained or for a marginal share of total employment in South completed tertiary education. Particularly, having attained Africa), casual work, self-employment, employees some tertiary education appears to be correlated with with temporary or time-limited work contract, and lower consumption volatility and poverty risks. employees with a permanent work contract. Precarious The classes clearly differ in their access to the labor forms of work, including casual employment and market: the more disadvantaged the class of a employment without a permanent contract, constitute household, the more likely the household head the largest share of all jobs among the poor and the is unemployed or economically inactive. Only 31 vulnerable, whereas among the middle class and elite, 80 percent of household heads among the chronically poor percent of all household heads who work as employees are employed, with the remainder being economically have a permanent contract (Figure 60a). In line with the 40 Overcoming Poverty and Inequality in South Africa observed education patterns, among those who engage occupations also dominate, followed in significance by as employees, household heads of chronically poor service and sales occupations. Among the middle and elite households are most likely to be employed in elementary classes, a very high proportion of household heads are occupations. Similarly, for household heads belonging to employed in highly skilled occupations, such as managers, transient poor and vulnerable households, elementary professionals, or technicians (Figure 60b). Figure 60: South Africa’s five social classes in the labor market, 2008–2014/15 a. Economic activity of the household head b. Occupation of the household head (employees) Source: Authors’ calculations using NIDS waves 1–4 pooled sample (post-stratified weights). Notes: Figures represent employment status and occupational category limited to heads of households. Female-headed households, large families, children, food poverty line also declined about the same amount as and people in rural areas are especially vulnerable to measured by either the lower bound poverty line or upper being in poverty for a long time. Larger households face bound poverty line. In absolute terms, around 2.3 million a higher risk of experiencing a poverty spell which tend South Africans escaped poverty at the LBPL and 1.2 million to be more persistent. Specifically, the chronically poor at the UBPL. Between 2011 and 2015, however, at least 2.5 live in households that, on average, have seven members, million more South Africans slipped into poverty. Poverty which is more than twice the size in comparison to those rates not only rose between 2011 and 2015, the level of who were never poor. Chronic poverty particularly affects poverty also became deeper and more unequal. children, with every second child below age 15 growing up The level of multidimensional poverty has declined in persistent poverty. since the end of apartheid, but it has stagnated in recent years. Further, the results highlight continuing gaps SUMMARY with respect to expanding access to basic services in an Consumption poverty rates declined in South Africa inclusive manner and reducing multidimensional poverty. between 2006 and 2015, but the trend has reversed in The poor tend to be affected by these gaps more than the recent years. The share of South Africans living below the rich, with access increasing with income levels. Inequality is high when it comes to access to safe water and improved An Assessment of Drivers, Constraints and Opportunities 41 sanitation. Food insecurity, stunting, and child malnutrition A higher level of education of the household head remain a challenge and some indicators have deteriorated and access to stable labor market income are key since 2012. Reducing multidimensional poverty will involve determinants for households to achieving economic leveling the playing field in the access of children to quality stability. This implies that access to quality higher education, irrespective of location, gender, or race. Paying and tertiary education, better labor market access, special attention to water, sanitation, and health care needs and improvement of both the quantity and quality of of rural areas and townships, and overcrowding in townships employment opportunities would be important to would also be important. Policy design needs to recognize spurring the growth of the middle class. The demographic that children of certain circumstances are vulnerable to characteristics of households, such as family size, structure, deprivations in multiple dimensions simultaneously. The and race play an important role in the determination of the presence of multiple deprivations points to the need for socioeconomic status of the family and its level of poverty. policy solutions. Black South Africans consistently have the highest poverty rates, but the prevalence is falling. Poverty is persistent and the economy is highly polarized. Almost half of the population is chronically Poverty has a clear spatial dimension and spatial poor at the UBPL. That is, for a relatively large share of the patterns of poverty suggest progress toward population, poverty is a permanent state. The causes of dismantling the spatial legacy of apartheid has been chronic poverty are linked to low levels of education, low slow. Rural areas remain the regions of highest poverty labor force participation, demographic factors, and low concentration. The results reveal a notable divide in poverty skills. One in four South Africans can be considered stably levels between two sets of provinces: Free State, Gauteng, middle class or elite, whereas the rest are either poor or and Western Cape versus Eastern Cape, KwaZulu-Natal, have an elevated risk of falling into poverty. At 20 percent, and Limpopo. This divide is a clear legacy of apartheid: the size of the middle class is thus considerably smaller, compared to Eastern Cape, KwaZulu-Natal, and Limpopo; and its growth has been more sluggish than suggested by the Free State, Gauteng, and Western Cape did not have other studies. The racial composition of the middle class high concentration of “homelands” during apartheid. has changed over time: however, black South Africans are Homelands were areas set aside for black South Africans still underrepresented in the middle class relative to their along ethnic lines during apartheid. Public service delivery share in the overall population. and infrastructure was poor in these areas. 42 Overcoming Poverty and Inequality in South Africa CHAPTER 3 SOUTH AFRICA IS ONE OF THE MOST UNEQUAL COUNTRIES IN THE WORLD With a consumption expenditure Gini coefficient of 0.63 in This chapter examines inequality of both outcomes 2015, South Africa is the most unequal country in the world and opportunities. It aims to identify factors relevant to and incomes are highly polarized. The country is characterized explaining each type of inequality, as well as how they by high wealth inequality and low intergenerational mobility have changed over time. The Income and Expenditure which arise from high income inequality and inequality of Surveys (IESs) from 2005/06 and 2010/11, Living Condition opportunity for children. This also helps explain the missing Surveys (LCSs) from 2008/09 and 2015/16, and the National middle and polarization in the labor market. These inequalities Income Dynamics Studies (NIDS) from 2008 to 2014/15 are appear to be passed down from generation to generation, used. Wealth and dimensions of wage inequality, as well implying little change in inequality over time and perhaps even as the level of polarization, are examined. Also analyzed is a worsening of the already bad situation. Not only does South inequality of opportunity. Africa lag its peers on level of inequality and poverty, it lags on the inclusiveness of consumption growth. Also, changes in the inequality had an adverse impact on the reduction of extreme poverty. An Assessment of Drivers, Constraints and Opportunities 43 A. CONSUMPTION INEQUALITY IS VERY HIGH country had a Gini coefficient of 0.63 in 2015, one of the AND HAS INCREASED SINCE THE END OF highest in the world and an increase since 1994.15 APARTHEID South Africa is economically highly polarized country. Figure 62 shows an international comparison of countries’ South Africa inherited very high inequality from the Duclos-Esteban-Ray Polarization Indexes16. South Africa has time of apartheid, and it increased since.14 Analysis of the highest value of the index. This level of polarization has the distribution of consumption expenditure per capita in not changed over time: the value stays at or close to 0.37 the recent Living Conditions Survey 2014/15 found that the across a 10-year span. 14 It is important to note the differences in the Gini coefficients present- ed in this report and those presented in Statistics South Africa (2017). While both estimates are based on the same data, Stats SA uses dif- ferent welfare aggregates for poverty and inequality estimates. The 15 WDI is used as the source for the countries, using the latest available per capita welfare measure used for poverty measurement includes year after 2011. all food items while for non-food items, large-sized, or “lumpy, durable 16 Duclos-Esteban-Ray index (Duclos et al 2004) measures the extent to goods” are excluded to reduce their biasing factor in the monthly es- which groups of individuals within a country feel alienated from each timates. For inequality measurement, total consumption expenditure other, yet this alienation takes place alongside a strong within-group (including components that are excluded in the welfare aggregate identity. used for poverty measurement), in per capita terms, is used. This re- port uses the same per capita welfare aggregate for both poverty and inequality measurement, and it is the one that excludes some com- ponents of consumption. This allows for comparison across countries, as most countries tend to use the same per capita welfare aggregate for poverty and inequality estimates. Figure 61: Long-term trends in inequality, comparison Figure 62: Polarization indexes across countries to other countries Source: South Africa: authors’ calculations based on the Income and Source: Authors’ calculations based on the Income and Expenditure Surveys Expenditure Surveys for 2005/06 and 2010/11 and the Living Conditions for 2005/06 and the Living Conditions Surveys for 2014/15. Surveys for 2008/09 and 2014/15 and WDI for 1996. WDI for the rest of the Note: Methodology based on Duclos et al. (2004). countries and regional estimates. 44 Overcoming Poverty and Inequality in South Africa Consumption trends indicate growth in the median median to upper-median percentiles. Between 2011 and to upper-median percentiles, decline at the top, and 2015, the ratios did not change much, indicating stagnation. stagnation at the bottom. Figure 63 shows the growth Figure 64 shows consumption shares over time, by groups incidence curve for consumption expenditures between defined in terms of their place in the overall distribution. 2006 and 2015. The trend here corroborates the evidence The 40th to 75th percentile gains the most (5 percent) while that the very poor—those in the bottom 10 percent—grew the top 10 percentiles lose the most (6 percent) between less that the rest of the population over time. Consumption 2006 and 2015. The bottom 40 percent experienced no growth between 2006 and 2011 was concentrated in change in their share of consumption. Figure 63: Growth incidence of consumption expenditures Figure 64: Consumption shares over time by percentile, 2006 to 2015 Source: Authors’ calculations based on the Income and Expenditure Surveys for 2005/06 and 2010/11 and the Living Conditions Surveys for 2014/15. Employment income accounts for a larger share of significantly as a proportion of total income. Meanwhile, the income for median and upper-median percentiles, poor increased their dependence on grants: the bottom 40 while the poor rely on grants. The increased reliance percent experienced a 4 percent rise in the proportion of on employment incomes appears to raise income shares grants and other income sources to total income. Figure for the median and upper-median percentiles, while the 66 shows the top decile had an 8 percent decline in its bottom 40 percent rely more on grants. Figure 65 shows share of total income, and the median to upper-median these trends, indicating changes between 2006 and 2015 in percentiles, particularly those between the 40th and 75th, the composition of income by deciles. Particularly for those had an increase. in the 40th to 75th percentile, work-based income increases An Assessment of Drivers, Constraints and Opportunities 45 Figure 65: Changes in income shares by source Figure 66: Income shares over time 100% 6% 5% 3% 90% 13% 22% 80% 34% 70% 60% 64% 60% 50% 94% 95% 97% 40% 87% 78% 30% 66% 20% 40% 36% 10% 0% 2005 2015 2005 2015 2005 2015 2005 2015 Bottom 40 40th to 75th 75th to 90th Top 10 Percentile Percentile Work Grants and Others Source: Authors’ calculations based on the Income and Expenditure Surveys for 2005/06 and the Living Conditions Surveys for 2014/15. B. HIGH LEVEL OF INEQUALITY OF From a cross-country perspective, the inequality of OPPORTUNITY opportunity (and its ratio to overall inequality) is the highest in South Africa. Figure 67 shows selected i. Extent of inequality of opportunity estimates of the inequality of opportunity index and its ratio to overall inequality for South Africa, upper middle- Access to quality basic services, such as education, income, and Sub-Saharan African countries.17 health care, and essential infrastructure, provides a better understanding of the nature and causes of inequality of outcomes. 17 The Ferriera-Gignoux (2011) method for estimating the inequality of opportunity uses gender, age, race, father’s education and occupa- tion, and the district council at birth as the predictor variables. 46 Overcoming Poverty and Inequality in South Africa Figure 67: Inequality of opportunity, cross-country estimates Figure 68: Decomposition of the inequality of opportunity into constituent factors Ug anda 9% 4% Sub- Saharan Africa Ghana 11% Ratio of Inequality 5% of Opportunity to Cote d'Ivoire 14% 5% O verall Inequality Guinea 13% 6% Madag ascar 21% 9% Upper Middle Income Countries Turkey 26% 9% 23% Inequality of Colombia 13% O pportunity Ecuador 26% 15% Peru 28% 16% Panama 30% 19% Brazil 32% 22% South Africa 45% 40% 0% 10% 20% 30% 40% 50% Source: Brunori et al. (2013); for South Africa, estimates from NIDS 2008–2014/15. Race, parent’s education, and father’s occupation are ii. Human Opportunity Index in South Africa major determinants for individuals’ opportunities, and The main principle of equality of opportunity for children the latter two factors affect labor market outcomes is that predetermined circumstances such as gender, for children. Father’s occupation plays a large role, ethnicity, place of birth, or family origins should not play highlighting the importance of intergenerational labor a role in determining the ability of a person to archive networks; meanwhile, the level of a child’s education can be socioeconomic success. This way, a child born in a poor, strongly influenced by that of its parents. Figure 68 shows rural, black family should have the same chance to get that the contribution of race is less than that of parent’s quality education and be successful as a child of a white education. The father’s occupation makes the next biggest family from Sandton in the Gauteng province, one of the contribution at 11 percent.18 richest areas in South Africa. Opportunities among children are measured in this section by the Human Opportunity 18 This analysis is also carried out at the household level, using character- istics of the household head. Households are divided into four racial Index (HOI), which is the coverage rate of a basic service categories—black, colored, Asian/Indian, and white. Parents’ educa- tion consists of mother’s and father’s highest level of education, each adjusted by how equitably the service is distributed among of which has five possible values—no schooling, primary, secondary, groups differentiated by circumstances.19 matriculation, and tertiary. Father’s occupation has 10 possible val- ues—legislators/senior officials/managers, professionals, technicians/ associate professionals, clerks, service/shop/market sales workers, skilled agricultural/fishery workers, craft/related trades workers, plant 19 This means that two societies with the same coverage rate for a ser- and machinery operators/assemblers, military, and elementary occu- vice can have different HOIs if access to that service in one society is pations. determined to a greater extent by gender, race, family background, or other personal circumstance beyond their control and considered by society to be an unjust source of exclusion. An Assessment of Drivers, Constraints and Opportunities 47 A range of indicators capturing access to education, and the D-index measuring the level of inequality for health, and basic services are analyzed to show how each service or good are presented in Figure 69. Some equitable and extensive access to services are in South opportunities, such as school attendance by children Africa. The indicators included in the analysis are school under the age of 16, school instructors, adequate teachers, attendance (ages 6–11, 12–15, and 18–25), quality of access to telecommunications, and access to electricity education (ability to finish tertiary grade, starting primary are nearly universal, with an HOI above 90 percent. school on time, adequate infrastructure at school, adequate Intermediary HOI between 60 and 90 is associated with teachers), access to health insurance, access to services the quality of education such as starting primary school (severe overcrowding,20 access to water on site, improved on time, completing seventh grade, and improved water water and quality, improved sanitation, access to electricity, and sanitation. However, well below universal (an HOI waste removal service, access to telecommunications, and of 60 percent or below) are access to health insurance, environmental issues). environment issues, housing conditions without overcrowding, and access to tertiary education and school Opportunities among children in South Africa vary attendance among youth. The latter are distributed with widely across different types of services. HOI indexes high inequality among children of different circumstances. 20 Severe overcrowding is defined as habitation by more than three people per room. Figure 69: Human Opportunity Index and D-index of inequality of opportunity, 2015 Source: Authors’ calculation from LCS 2014/15 and GHS survey. HOI (0–100)—higher is better. D-index (dissimilarity)—penalty for inequality of distribution—higher index is higher inequality. 48 Overcoming Poverty and Inequality in South Africa Despite rapid improvement in access to services in the school attendance, where HOI remains low due to high 2000s, progress slowed in recent years. In comparison to inequality of opportunity. middle-income countries, South Africa fares well on school Inequality of opportunity among children in South attendance, access to electricity, and telecommunications, Africa is shaped by various circumstances. Figure 71 but ranks below most comparators on the HOI for shows the contribution of various circumstances to the completing primary school on time and access to safe D-index measuring inequality of opportunity. Whether a water on site, improved sanitation, and access to tertiary child lives in a township or rural area as opposed to an urban education. Our analysis suggests general improvements area, and education of the household head contributes the in HOI over 2006–2015, but the gaps with other most to inequality of opportunity in most cases. Location countries are generally not closing. Except for electricity, is particularly important for opportunities related to telecommunications, and access to sanitation, where South infrastructure (access to electricity, telecommunications, Africa’s average annual progress has been exceptional, water); and education of the household head contributes progress on the other dimensions was less impressive. The the most to inequality in finishing primary school on time bulk of the change in most of the HOI indicators occurred and having health insurance, underscoring the role of the between 2002 and 2010, while the improvement between family’s socioeconomic background on the future of its 2011 and 2015 was positive, but less prominent. Especially children. slow progress is observed in tertiary education and youth Figure 70: Change in the HOI and decomposition of Figure 71: Contribution circumstances to D-index, changes, 2002–15 2015 Source: Authors’ calculation from LCS 2014/15 and GHS surveys. Source: Authors’ calculation from LCS 2014/15 and other surveys. Selected with high inequality (D-index greater than 10). The overall picture of inequality of opportunity is electricity, both of which more than tripled to reach more ambiguous. On the positive side, South Africa improved than 90 percent in 2015, together with a big increase in most of the HOI indexes over 2006–2015, achieving near- HOI for sanitation and school infrastructure, are improving universal access to primary education, a necessary first opportunities for children. Major challenges are the limited step for equalizing opportunities among children and and unequal access to quality education and ability to an important success for the education system to build finish primary school on time, and inequality of access to on. The rapid rise in access to telecommunications and tertiary education. Inequality is high with respect to access An Assessment of Drivers, Constraints and Opportunities 49 to safe water on site and improved sanitation and general C. WAGE INEQUALITY IS VERY HIGH AND IS lack of physical safety—all of which affect the conditions COMPOUNDED BY HEAVY POLARIZATION for children and youth to develop their human potential. BETWEEN TWO EXTREMES Early childhood education has substantial long-term impacts that affect adult earnings. Access to safe water and i. Trends and causes of wage inequality improved sanitation are particularly critical inputs for child South Africa is characterized by extreme wage health, a determinant of nutrition status. inequality. While part of the population enjoys wages roughly equivalent to those living in developed economies, the lower-end wages are comparable to those in the poorest countries (Figure 72). Figure 72: Wage inequality Source: Data for international wages come from Oostendrop (2013) and is the average wage in US dollars for the latest year data is available. Data for South Africa come from NIDS wave 4, converted into US dollars using the conversion rate of R10.76 per dollar (taken for January 3, 2014, from https:// www.bloomberg.com/quote/USDZAR:CUR). High wage inequality is compounded by heavy than a fifth of the total working population. A little over polarization between two extremes. The number of 10 percent of the working population is white, but white workers with high-end jobs is low, while a large fraction South Africans make nearly three times the average wage of the working population is employed in very low paid for black Africans, who constitute nearly three-quarters of jobs. For instance, high-skill jobs earn nearly five times the entire labor force (Figure 73 and Figure 74). the average wage for low-skill jobs yet constitute less 50 Overcoming Poverty and Inequality in South Africa Figure 73: Average wages by groups Figure 74: Group share in the sample Source: NIDS wave 4. A skills-biased labor demand trajectory in an economy 29 percentiles ranges between 3.4 percent to about 1.7 would be suggestive of a widening level of internal percent, after which the growth rate drops to an average labor market inequality. Real growth in wages has been of 0.98 percent between the 30th and 69th percentiles. For positive for all percentiles of the distribution, including the the 70th percentile and above, the average growth rate per mean. However, real wage growth rates are heterogeneous year is 3.6 percent. Thus, while real wages at the bottom by percentile. A closer look at the average annualized of the distribution are growing at an annualized rate of 2 percentage change in wages by percentile between 1994 percent per year, and high-end real wages are growing at and 2014 (Figure 75) shows that the middle of the income almost twice the rate of the bottom, workers in the middle distribution has lost the most in the post-apartheid era. of the distribution have experienced real growth rates that The average annual real wage growth rate of the bottom barely exceed 1 percent. Figure 75: Real monthly wage by percentile, average Figure 76: Real wage inequality, 1995–2014 annualized percentage change 1994–2014 Source: Post-Apartheid Labour Market Series, authors’ calculations. An Assessment of Drivers, Constraints and Opportunities 51 Policy may have a large role to play in explaining the Wage movements have reinforced a pattern of gap in the middle of the wage distribution. Pro-poor disadvantaging those in the middle of the distribution. policies such as the Basic Conditions of Employment An examination of the real earnings distribution in 1994 and Act, employment tax incentives, and various sectoral 2014 presents the change in earnings in the post-apartheid minimum wage laws may have protected the employment era. Employees in the middle of the wage distribution, and increased the wages of more vulnerable workers those typically in semi-skilled jobs, have experienced much at the bottom of the distribution. A skills-based growth lower real wage growth than workers on either side of path has in turn maintained the relatively high demand them in the wage distribution. These findings have been for skilled workers who, being in short supply, retain a reinforced variously by a sectoral pattern of growth favoring significant premium. Ultimately then, the combination of skills-intensive services, policy choices favoring low-wage policies protecting and promoting wages at the bottom workers, and technology responses by firms, which may of the distribution, lack of a semi-skilled labor-intensive have an adverse impact on the median worker. manufacturing sector, and a growth trajectory built on high demand for highly skilled workers—has eroded the D. WEALTH INEQUALITY IS VERY HIGH, EVEN earnings of workers in the middle of the distribution. HIGHER THAN INCOME INEQUALITY Other measures of inequality indicate that wage Household net wealth is an indispensable factor in inequality increased significantly between 1995 and defining the economic well-being of the population. 2014. The wage Gini coefficient rose from 0.58 to 0.69 The importance of household wealth analysis for policy between 1995 and 2014. At the same time, the Palma ratio followed the publication of Thomas Piketty’s Capital in the (the share of the top 10 percent of earners’ wages to the 21st Century (2014). Traditionally, poverty research in South share of the bottom 40 percent) has almost doubled, from Africa focused nearly exclusively on income poverty. Such 5.11 to 10.13.21 Decomposing the Gini coefficients by sector research found that income poverty rates are generally shows the extent to which larger scale wage inequality is high. Increasingly, the focus in South African poverty driven by the interaction between intra-sectoral skills studies is shifting to exploring relationships between mismatches and sector of occupation. While real wage households’ wealth and poverty. The data required to inequality has increased in every sector since 1995, the compile distributional balance sheets were derived from size of the increase differs between sectors based on skills five nationally representative household financial wellness levels. surveys conducted by the Bureau of Market Research at the Lower skilled labor absorption influences the University of South Africa during the period 2011–2015. distribution of wage inequality. The finance and South Africa is one of the most unequal countries in community, social, and personal services22 sectors, whose terms of net wealth distribution (Figure 77 and Figure shares of skilled labor were the highest in 2015, exhibit the 78). The share of household wealth held by the top 10 largest growth (60 percent and 73 percent, respectively) percent in the distribution was 71 percent, while the in their sectoral Gini coefficients between 1995 and 2014. bottom 60 percent held 7 percent of the net wealth. Similar Conversely, the retail and wholesale trade sector, which statistics for OECD countries suggest that, on average, boosted the highest growth in unskilled labor between the top 10 percent of the wealthiest households own 50 1995 and 2014, exhibits one of the lower growth rates in percent of total wealth, while the bottom 60 percent own the wage Gini coefficient of 39 percent. only 13 percent. 21 The Gini coefficient and Palma ratio measures vary widely between periods. This could be attributable to the quality of the data collected since earnings surveys have usually low representation of higher in- come earners. 22 CSP services includes government services. 52 Overcoming Poverty and Inequality in South Africa Figure 77: Households wealth inequality, Gini coefficients Figure 78: The share of household wealth held by the across countries percentiles in the distribution Source: Authors’ calculations for South Africa, OECD Source: Authors’ calculations based on UNISA survey data. Wealth inequality is much larger than income and mortgage liabilities. Richer households have, on inequality. The bottom 50 percent of households account average, nearly 10 times more wealth than poor households for only 8 percent of incomes, 5 percent of asset values, and (Figure 79). For the poor, the financial assets represent 4 percent of net wealth. Conversely, the top 10 percent of 36 percent of total assets, while among the rich, financial households account for 55 percent of household incomes, assets represent 75 percent. Similarly, poor households about 69 percent of total household asset values, and 71 have a very small share of mortgage in total labilities (about percent of household net wealth. Clearly, wealth is much 7 percent), while for the rich this share is close to 58 percent. more unequal than income. Ownership of financial assets features prominently among the factors that influence wealth inequality. Richer households are almost 10 times wealthier than poor households and have much more financial assets Figure 79: Composition of wealth by income group Figure 80: Correlates of households’ income and wealth, coefficients from regression analysis Source: Authors’ calculations for South Africa, OECD Source: Authors’ calculations. Selected coefficients are from regression analysis. Income and log wealth dependent variables. Base age 15–20, blacks, less than primary education, grants as main source of income. An Assessment of Drivers, Constraints and Opportunities 53 Human capital (education attainment) is strongly E. LOW INTERGENERATIONAL MOBILITY IS correlated with higher wealth as well as higher incomes AN OBSTACLE TO INEQUALITY REDUCTION and earnings. The elasticities for the income and net worth regressions are presented in Figure 80. Tertiary education New data suggest low levels of intergenerational has the highest elasticity in net wealth determination—on mobility, which also relates to high income inequality. average 220 percent compared to less than primary school. The new estimate of intergenerational elasticity is 0.634 – The impact of education on net wealth is even stronger suggesting relatively low level of intergenerational mobility. than the impact on income. The second strongest correlate This is generally close to earlier estimates by Piraino with net wealth is race: white South Africans have much (2015) of 0.67 (Box 7).23 In Figure 81 (panel a) estimates higher elasticity than black South Africans and, as in the of intergenerational elasticities are plotted against Gini case of education, the impact is stronger on net worth. coefficients for 23 countries including South Africa. Given Other factors such as age, employment income, income the estimated error, 95 percent confidence intervals bound from investments, and being male contribute to income the intergenerational elasticity between 0.73 and 0.53, and wealth generation. suggesting South Africa has intergenerational mobility comparable to Brazil, Chile, China, and Peru. 23 If the data are restricted to the first three waves of NIDS, the elasticity estimate is 0.68, very similar to Piraino (2015). Box 7: Intergenerational mobility in South Africa Intergenerational mobility refers to the link between life outcomes for a given generation versus those of the preceding generation. A mobile society would be one in which this link is very weak or non-existent. Life outcomes is a very general concept and can refer to incomes, educational achievement, or occupation status, among other factors. Economic mobility varies across countries. Family structure, education, labor markets, and public policies all interact to affect the relationship between child and parental outcomes (Corak 2013). In addition, segregation either along racial or income dimensions, can affect mobility. Many of these factors were first identified by Becker and Tomes (1979, 1986). As discussed in the previous chapters, inequality is stubbornly high in South Africa and has risen more than two decades after the end of apartheid. Why this remains the case is an enduring puzzle. International evidence suggests an inverse relationship between inequality and mobility, a relationship nicknamed the “Great Gatsby Curve” (Krueger 2012). Given its level of inequality, this relationship suggests South Africa would have low mobility. Intergenerational mobility in South Africa is indeed low, with a high intergenerational elasticity, and shows persistence at the top of the distribution. Piraino (2015) estimates South Africa to have an intergenerational elasticity of 0.67 and suggests the existence of a racial component in mobility.24 Low intergenerational mobility paints a rather pessimistic scenario as it suggests current levels of inequality are likely to persist in the future. This section presents new evidence on intergenerational mobility and explores the relationship between inequality and mobility based on the new wave of the NIDS data. It also identifies explicit characteristics of intergenerational mobility and analyzes the possible causes of upward mobility. 24 24 The analysis here focuses on intergenerational income mobility. Other studies analyze intergenerational mobility using different dimensions. Magruder (2012) finds a strong intergenerational link in labor market networks between fathers and sons, which may reduce mobility if networks play a major role in job allocation. Educational mobility appears to be improving but occupational mobility is stagnant (Girdwood and Leibbrandt 2009, Nimubona and Vencatachellumn 2007). 54 Overcoming Poverty and Inequality in South Africa Intergenerational mobility and inequality are and Gini coefficients within South Africa. Estimated at the negatively correlated. Figure 81 (panel b) shows the province level, an inverse relationship between inequality relationship between intergenerational elasticity estimates and mobility is revealed. Figure 81: The relationship between intergenerational mobility and inequality Panel a: Cross-country data Panel b: South African provincial data 1.2 Eastern Cape 1 Intergenerational Elasticity KwaZulu- Western Natal 0.8 Cape Gauteng Mpumala 0.6 nga Northern Cape 0.4 Limpopo North 0.2 West Free State 0 0.6 0.65 0.7 0.75 0.8 Gini Coefficient, Household Income Source: Data for earnings elasticities taken from Corak (2016); data for the Gini coefficient taken from the WDI. NIDS survey weights were used to construct the intergenerational elasticity estimate for South Africa. Note: Higher intergenerational mobility coefficient suggests lower mobility, thus lower coefficients are preferred. At least a third of all sons born to very poor fathers— those in the first quintile—will occupy the top 40 percent of their income distribution. Sons of rich fathers—those in the fifth quintile—have a 43 percent chance of also being in the top quintile of their income distribution. Table 6 shows the frequency of transferring income quintiles across a generation. Both single and multiple imputation25 methods give similar results. The probabilities here are constrained to add up to 100 by father quintiles. 25 Full panel data on father and son incomes are lacking so the anal- ysis uses a two sample, two stage instrumental variables procedure (Bjorklund and Jantti 1997), where father incomes are first regressed on a set of characteristics using historical data. The estimated coeffi- cients are then used to predict father incomes for sons captured in the NIDS data. This is the single imputation procedure. In the multiple imputation procedure, coefficients on father characteristics as well as its variance-covariance matrix are estimated. Using these predicted means and variances, multiple imputation then draws multiple simu- lated means and variances under a specific distributional assumption. These simulations are then averaged out to provide a final estimate of the father’s income. An Assessment of Drivers, Constraints and Opportunities 55 Table 6: Frequencies of transition across income quintiles (multiple imputation estimates)   Son Quintile Father Quintile 1 2 3 4 5 1 22.77 19.43 21.22 20.38 16.19 2 15.98 16.61 21.21 22.52 23.67 3 13.19 15.44 20.35 22.25 28.77 4 10.86 14.67 16.79 22.70 34.97 5 9.88 12.42 14.12 20.88 42.71 Source: Authors’ calculations based on NIDS data. Notes: all rows add up to 100. Sons of poor fathers are more mobile than sons of at the 10th percentile.26 For comparison, from Bratsberg et rich fathers: elasticities at the 50th and 90th percentile al. (2007) similar elasticities are reported evaluated for three of father’s income are more than twice that at the 10th other countries. At the 10th percentile, South Africa has percentile. Figure 82 shows intergenerational elasticities higher mobility than the United States or United Kingdom, evaluated at the 10th, 50th, and 90th percentile of father’s but at higher percentiles mobility falls. income. For South Africa, the elasticity rises as father income percentiles increase; the estimates at the 50th and 26 The relationship between the incomes of sons and fathers is estimat- ed using a polynomial specification. The order of the polynomial is de- 90th percentile are statistically different from the estimates cided based on overall fit of the regression. Using root mean squared error, the Akaike Information Criterion or an F-test of model fit sug- gests a third-order polynomial fits best. Bratsberg et al. (2007) uses root mean squared error as the decision criterion. In addition, the plot of the incomes of sons versus fathers does not support moving up a higher order. Figure 82: Intergenerational elasticities at various percentiles of father’s income Source: Authors’ analysis. Notes: This figure shows the elasticity between father’s and son’s income at different points of the distribution of father’s income. Data for Denmark, United States, and United Kingdom are taken from Bratsberg et al. (2007); estimates here are taken from regressions that include father’s age and age squared as additional explanatory variables. For South Africa, both son’s and father’s income are age-adjusted, so father’s age is not included as a control when evaluating the elasticities; further, the elasticities are calculated at each decile. 56 Overcoming Poverty and Inequality in South Africa F. SOUTH AFRICA LAGS ITS PEERS ON (2003). In this case the GICs plot the average growth rate INCLUSIVENESS OF CONSUMPTION of real consumption between 2006 and 2015. This enables GROWTH an assessment of the role of growth and redistribution in bringing about changes in poverty in South Africa between i. Incidence of growth 2006 and 2015 as well as between any two periods under analysis. Specifically, the use of GICs sheds light on whether This section examines how consumption expenditure of the expenditure of the poor may increase more or less different groups changed between 2006 and 2015. That than that of the country overall when national income or is, it describes the distributional effects of consumption expenditure increases. This is important given the prevailing growth from 2006 to 2015. This is done using Growth high inequality in South Africa. Incidence Curves (GICs) as proposed by Ravallion and Chen Figure 83: Growth incidence curves, national Source: Authors’ calculations based on the IESs for 2005/06 and 2010/11 and the Living Conditions Surveys for 2008/09 and 2014/15. At the national level, growth in consumption zero, that is, the consumption expenditure levels of the poor expenditure between 2006 and 2015 was pro-poor have increased in absolute terms. A “relative” perspective to in absolute terms, but deteriorated in relative terms. pro-poor growth says growth is relatively pro-poor if the All segments of the population along the consumption change in the expenditure levels of the poor is larger than expenditure spectrum experienced growth in consumption the change in the expenditure levels of the non-poor. between 2006 and 2015 (Figure 83). Pro-poor growth can The shape of GIC curves changes sharply between any be considered “absolute” if the change in consumption two periods under analysis. Essentially, the trends were expenditure levels of the poor over a given period is non- reversed from one period to the next. The GIC for 2006 to An Assessment of Drivers, Constraints and Opportunities 57 2009 indicates negative growth in consumption for the The pattern of distribution of consumption expenditure poorest 2 to 25 percent of the population as well as the growth varies geographically. Considering the urban- top 5 percent. This could be reflecting the negative impacts rural delineation, between 2006 and 2015 those in the of the 2008/09 financial crisis, which led the country into middle of the consumption expenditure distribution in a recession, with the economy shrinking by 2.9 percent in urban areas benefited more, in relative terms, from growth 2009. The financial crisis likely affected the richest segment and redistribution of consumption compared to the poor of the population the most given their integration into the and those at the upper end of the expenditure distribution financial sector. The poorest of the population are likely to (Figure 84). The bottom 15 percent and the top 10 percent have been affected by the 2007/08 global food prices. The of the population registered negative growth between absolute pro-poor pattern shown in 2009–2011 is consistent 2006 and 2009. Weak economic growth prospects between with the recovery in GDP following the recession in 2009. 2011 and 2015 are shown to have negatively affected the A growth in per capita GDP of 1.8 percent was recorded rich more than they affected the poor in urban areas. between 2009 and 2011. Economic growth prospects have been weaker since then, and this is reflected by negative consumption growth across-the-board between 2011 and 2015. Figure 84: Growth incidence curves 2006–2015, urban and rural Source: Authors’ calculations based on the IESs for 2005/06 and 2010/11 and the Living Conditions Surveys for 2008/09 and 2014/15. In rural areas, on the other, hand, the rich benefited Not only does South Africa lag its peers on international more from consumption expenditure growth between poverty rates, the country is a highly unequal and 2006 and 2015 than the poor and those in the middle lags its peers on the inclusiveness of consumption (Figure 84). The overall picture in rural areas is that growth. Inclusiveness of growth in this case is examined expenditure rose slower for those in the lower part of the by considering the rate at which the consumption of the expenditure distribution than for those who were better bottom 40 percent of the population grows compared to off. The relatively positive slope of the growth incidence the growth in the consumption of the total population. curve in rural areas shows that, as a percentage of their Focusing on the bottom 40 percent is consistent with the initial consumption level, the rural rich have seen a higher shared prosperity goal of the World Bank Group. Shared percentage increase in their consumption between 2006 prosperity is an indicator used to measure and track the and 2015. income or consumption growth among the bottom 40 percent in a country. It is an indicator of economic 58 Overcoming Poverty and Inequality in South Africa growth with equity and inclusion. Growth is said to lack 2006 and 2011 (3.5 percent), the period between 2011 and inclusiveness if the income or consumption expenditure 2015 was marked by deceleration of consumption for this growth of the bottom 40 percent is consistently lower than group. The consumption of the bottom 40 percent shrank the average income or consumption expenditure growth of by 1.4 percent. This does not compare well with the median the total population. Figure 85 shows that while the bottom for the world (3.9 percent). 40 percent registered growth in consumption between Figure 85: Shared prosperity indicator in selected countries (2007–2014) Source: Authors’ calculations based on WDI http://www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity G. INEQUALITY SLOWS DOWN POVERTY poverty. Figure 86 decomposes poverty reduction into REDUCTION two components: a change in the average per capita consumption expenditure and a change in distribution The broad links between economic growth and poverty of consumption expenditure around the average (the changes can be attributed to changes in the growth (or redistribution component). At the LBPL, growth contributed lack thereof ) in consumption and changes in inequality 10.4 percentage points compared to 1.8 percentage point of consumption. The method used here to decompose contribution from the redistributive component. In urban changes in poverty into growth and redistribution and rural areas, growth reduced poverty by 7.6 and 10.6 components was developed by Datt and Ravallion (1992). percentage points. Redistribution reduced poverty by 2.2 Decomposition of changes in incidence of poverty percentage points in urban areas but increased it by a slight between 2006 and 2015 suggests growth in 0.3 percentage points in rural areas. In contrast, the finding consumption contributed more to overall poverty at the FPL shows that growth drove poverty reduction while reduction while changes in inequality (redistribution inequality slowed the process. The slowdown in poverty component) had a negative impact on the extreme reduction due to redistribution was more pronounced An Assessment of Drivers, Constraints and Opportunities 59 in rural (9.7 percentage points) than urban areas (4.3 reduction, specifically easing the increase in poverty while percentage points). The 2011–2015 period was the only the growth component contributed more and positively to period in which redistribution contributed to poverty poverty rates. Figure 86: Decomposing changes in the poverty headcount ratio into growth and redistribution Panel a Panel b Source: Authors’ calculations based on the IESs for 2005/06 and 2010/11 and the Living Conditions Surveys for 2008/09 and 2014/15. Use of the poverty gap and the squared poverty gap considering the severity (squared poverty gap) of poverty: further confirms the slowing effect of inequality on inequality dampened poverty reduction both in urban the welfare of the poor. The poverty gap, measured at and rural areas as well as at national level. Overall, these the LBPL, indicates a slowdown of reduction of depth of measures suggest inequality had a much stronger negative poverty due to redistribution in rural areas as well as at impact on poverty reduction in rural areas. the national level (Figure 87, panel a). This also holds when 60 Overcoming Poverty and Inequality in South Africa Figure 87: Decomposing changes in poverty into growth and redistribution, 2006–2015, poverty gap and squared poverty gap Panel a Panel b Source: Authors’ calculations based on the IESs for 2005/06 and 2010/11 and the Living Conditions Surveys for 2008/09 and 2014/15. SUMMARY implying little change in inequality over time and perhaps South Africa is the most unequal country in the world even a worsening of the situation. An empirical assessment by any measure. With a consumption expenditure Gini of key trends and movements in wage levels and wage coefficient of 0.63 in 2015, South Africa is the most unequal inequality in the labor market suggests further polarization country in the world and incomes are highly polarized. of wages. Wealth inequality is even larger than consumption Not only does South Africa lag its peers on level expenditure inequality and the country is also the most of inequality and poverty, it lags peers on the unequal based on wealth distribution. A significant inclusiveness of consumption growth. The expenditure determinant of this inequality is inequality of opportunity. growth of the bottom 40 percent is consistently lower than Analysis in this chapter confirms earlier estimates by Piraino the average income or consumption expenditure growth (2015) that very low intergenerational mobility paints a of the total population and below growth in other middle- rather pessimistic scenario, suggesting that current levels of income countries. Also, changes in inequality had an inequality are likely to persist in the future. These inequalities adverse impact on the extreme poverty reduction. appear to be passed down from generation to generation, An Assessment of Drivers, Constraints and Opportunities 61 CHAPTER 4 DRIVERS OF POVERTY AND INEQUALITY IN SOUTH AFRICA Consumption-based poverty declined considerably between inequality. South Africa has made progress toward creating 2006 and 2015. The largest explanatory factor in that decline an efficient social protection system, but further expansion is was increased labor income. Government social grants and likely to be fiscally unsustainable under the current low growth pensions were also important and contributed 24 percent and financially constrained scenario. The challenges of high to poverty reduction and 44 percent to reducing the poverty inequality should be solved by the creation of productive jobs gap. Improvement in education endowments, urbanization, and further improvement in the efficiency of services delivery. demographic changes, and expansion in the provision of services, also contributed to improved household welfare. A. WHAT DRIVES CHANGES IN POVERTY IN However, returns to education now are lower than they have SOUTH AFRICA? been in the past. While race continues to determine poverty and The findings presented in this chapter are based on three inequality, it has been declining in importance and the skills decomposition methods, as described in Box 8. and labor market is an increasingly important determinant of 62 Overcoming Poverty and Inequality in South Africa Box 8: Three methods for decomposing changes in poverty The non-linear Oaxaca-Blinder (1973) quantifies how much poverty reduction can be accounted for in changes in the characteristics of households (“endowments”) compared to the changing relationships between poverty and households’ endowments (“returns to endowments”). The second method uses Recentered Influence Functions (RIF, Firpo et al. 2009) in which the traditional Oaxaca-Blinder used in the first method is applied to different quintiles of the consumption distribution. Finally, the microsimulation approach proposed by Azevedo, Inchauste, and Sanfelice (2013) is used to understand the role of different sources of income in changes in the welfare of households. All decomposition methods rely on defining a counterfactual scenario and estimating what would have happened to poverty had the counterfactual scenario occurred. The Oaxaca-Blinder and the RIF analyses focus on a counterfactual scenario of a constant relationship between endowments and poverty in South Africa between 2006and 2015. This counterfactual scenario is used to determine the changes in endowments that have been important to poverty reduction and the amount of poverty reduction that could have been different due to the changes in returns to endowments. In these two decomposition methods, an interaction effect also exists and can be interpreted as a measure of the correlation between changes in endowments and returns to endowments. This interaction term is relatively small in the analyses. The third decomposition method introduced by Azevedo, Inchauste, and Sanfelice (2013a) focuses on four factors that could have a potential impact on changes in poverty: (i) a decline in household size could lead to higher levels of consumption per capita; (ii) growth in labor income could imply higher consumption; (iii) growth in non-labor income could also lead to higher consumption; and (iv) changes in the ratio of consumption to income. This approach constructs each counterfactual scenario by keeping one factor constant. Thus, poverty measures for each counterfactual distribution can be interpreted as the poverty that would have been realized in the absence of a change in that factor. Labor income remained the most important source in to an additional of 8 percent. Analysis of the lower bound of reducing the level and depth of poverty over 2006– poverty reduction suggests generally similar pattern with 2015, while income from social grants was the second. even stronger impact of the grants (Figure 88c and d). Labor income contributed 60.2 percent to the decline in In rural areas, income from grants was by far the poverty headcount and poverty gap (Figure 88a and b), largest contributor to reducing the poverty gap. Sixty- using the UBPL. The impact of transfers was also significant. nine percent of the decline in rural poverty gap can be Incomes from grants and pensions together contributed explained by income from grants alone. This observation to 24.0 percent of the upper bound poverty reduction. The may highlight the success of social assistance programs impact of grants was even more pronounced on the upper in targeting poor residents in rural areas, and the impacts bound poverty gap. Expansion of grants contributed 36.2 of such programs on reducing rural poverty have been percent of poverty gap reduction and pensions contributed encouraging. An Assessment of Drivers, Constraints and Opportunities 63 Figure 88: Contribution to poverty reduction by income sources over 2006–2015 Upper bound poverty line a. Reduction in poverty headcount b. Reduction in poverty gap Lower bound poverty line c. Reduction in poverty headcount d. Reduction in poverty gap Source: Author’s calculations based on the Income and Expenditure Surveys for 2005/06 and the Living Conditions Surveys for 2014/15. Note: Uses the methodology developed by Azevedo, Inchauste, and Sanfelice (2013). The methodology quantifies the contribution of the share of working-aged population and sources of income to poverty reduction. The ratio between consumption and income is an important component to link a change in poverty status with change in household income. In most African countries, poverty estimates are based on consumption. Meanwhile, changes in household consumption do not always align with changes in income. 64 Overcoming Poverty and Inequality in South Africa To further explore poverty reduction, the changes in 2015, but the results varied substantially by level of consumption have been decomposed using a range income. Changes in endowments represent changes of factors associated with the demographics, location, in composition of the population by factors such as education, work, and access to services for households. education, location, demographics, and labor. Changes in Distinction is made between the impact of endowments endowments of the poorest households, defined as those (composition of the respective characteristics) and returns in the bottom quartile of the consumption distribution, to endowments (changes in remunerations). The results of explain 77 percent of growth in consumption. However, the the decomposition are presented in Figure 89. contribution of endowments was much lower for the richer quintiles—47 percent and 52 percent for a household in Improvement in endowments accounted for about the top two quintiles, respectively. half of the average consumption growth over 2006– Figure 89: Endowments and Returns. The contribution of demographics, location of residence, education, access to services and labor to consumption growth, in %, LCS 2004/05–2014/15 Source: Authors’ calculation. Note: Interaction is the change in the consumption that could not be attributed to either endowments or returns. Changes in returns to endowments contributed to migration, accounted for a majority of the welfare 37 percent of the average consumption growth but improvements between 2006 and 2015. Figure 90 (panel mattered more for the rich. Returns to endowments a) presents the decomposition results for the contribution represent remuneration on the endowments. For example, of the changes in endowments of location, education, returns to education is a remuneration from getting higher demographics, labor, and access to services on households’ education. The role of the returns differed significantly consumption growth (as a share of total endowment across quartiles of the consumption distribution. Changes effect). Seventy percent of total endowment effects came in returns had almost no impact on the change in welfare from improvements in education. Improved access to of the bottom quartiles, while it constituted 61 percent of services and reallocation of population from rural to urban the growth of the consumption for the top quintile. areas each explained 30 percent of the total endowments. Improvements in employment had a small contribution to Improvements in education, followed by the total endowment. improvements in access to services and internal An Assessment of Drivers, Constraints and Opportunities 65 Decomposition analysis of returns suggests a negative on demographics have improved significantly, especially impact for returns to education but a positive effect for among poorer households. Most of the increase in returns demographic returns. Decomposition of the returns on on demographics was driven by changes in the returns to endowments is shown in Figure 90 (panel b). Changes in household size. In other words, larger households became returns to education endowments contributed negatively better off in terms of consumption in the later period, which to welfare growth, particularly for wealthier households. is generally associated with the increase in the non-labor Acquiring some secondary education no longer obtained sources of income due to the expansion of the social and the same increase in consumption in 2015 as it did in children grants that benefit larger households. Returns to 2006. While demographic endowments made relatively access to services have a smaller but positive welfare effect little contribution to welfare improvements, the returns for the poorest. Also, returns to location were positive. Figure 90: Causes of welfare changes, 2006–2015, in percent a. Endowments b. Returns c. Total impact Source: Authors’ calculations based on the Income and Expenditure Surveys for 2005/06 and the Living Conditions Surveys for 2014/15. Note: The numbers do not add up to 100 percent because of negative demographic effect (-20 percent). Combining both the endowment effects and the demographics each contributed 20 percent to welfare returns to endowments effects, education was the improvements for the total population. However, for the most prominent contributor to improvement of welfare poorest, the impact of demographic factors was negative. among the poorest, followed by location and access to While the foregoing analysis focused on the drivers of services. The combined impact of the endowments and poverty between 2006 and 2015, which allows for a long- returns to endowments of all the analyzed components term perspective on the drivers of poverty in South Africa, is shown in Figure 90c. The overall impact of education of interest could be to understand what factors explain the on the welfare improvement was significant: 50 percent increase in poverty between 2011 and 2015 as discussed of consumption growth of the poorest quartile was in Chapter two and presented in Statistics South Africa associated with education. Education was also the main (2017). According to the Stats SA’s 2017 poverty trends factor explaining improvements in welfare of the total report, the increase in the poverty levels between 2011 population, accounting for 30 percent of the overall and 2015 is associated with “a combination of international welfare improvements. Access to services, location, and and domestic factors such as low and anemic economic 66 Overcoming Poverty and Inequality in South Africa growth, continuing high unemployment levels, lower constant. Figure 91 shows that the contribution of race commodity prices, higher consumer prices (especially and education both fall by sizable amounts while the other for energy and food), lower investment levels, greater factors stay relatively constant. household dependency on credit, and policy uncertainty.” Race has become the main factor determining (Statistics South Africa 2017, pp 16). Rather than focus on inequality of opportunity.27 A Fields (2003) decomposition the most recent trends, this study takes a longer-term suggests race, education, and the labor market outcomes perspective with the aim of understanding the causes and are dominant factors explaining overall inequality. The consequences of polices and sources of poverty reduction. influence of race fell over time, while that of education This requires a longer-term perspective and makes it possible to better capture and explore factors and polices rose. Incorporating more detailed labor market information affecting inclusive growth and poverty in South Africa. from NIDS raises the contribution of the labor market (from 6 to 19 percent using NIDS data combined across all waves) B. WHAT DRIVES CHANGES IN INEQUALITY and lowers that of education (from 42 to 33 percent). AND INTERGENERATIONAL MOBILITY IN Within education, the categories that contribute most to SOUTH AFRICA? inequality are at the higher end: finishing high school and getting a college degree. i. Drivers of inequality of consumption Race, household size, education, and location are the 27 The Theil-L measure of inequality is used to investigate the possible factors behind inequality. The advantage of this measure is that it biggest contributors to inequality. Of these factors, can be broken into between-group and within-group contributions. the influence of race and education appears to have Grouping observations by various factors, permits assessment of which factors appear to contribute the most to inequality by dividing declined over time while the others stay relatively the between-group contribution by total inequality. Figure 91: Factor wise contribution to inequality (Theil-L Measure) Source: NIDS 2008–2015, authors’ calculations. Race and gender are becoming less important factors contributes to inequality in all opportunities but does not by themselves in determining the extent of inequality rank among the top two contributors for any indicator. Race (Figure 92). The gender of a child contributes appreciably and gender correlate to other factors, such as education to inequality only in finishing primary school (seventh and socioeconomic characteristics and have impact grade) on time and in youth school attendance. Race through these circumstances. An Assessment of Drivers, Constraints and Opportunities 67 Figure 92: Decomposition of inequality by contributing factors Figure 93: Inequality by income sources Source: IES/LCS 2005–2015, authors’ calculations. A detailed breakdown of the various factors affecting contribution of employment income is steady and at a very inequality suggests that education and labor market high level.28 The contribution of employment incomes to affiliation are primarily responsible for overall inequality is much greater than its share in total income. inequality. Within race, white contributes 38 percent. ii. What drives intergenerational mobility? Within education, tertiary education contributes 40 percent. The bulk of inequality from education comes from The existing literature uses six factors as correlates those completing a college degree, which indicates that of mobility: education, labor markets, race, family education provides a path to a high-paying job. Indeed, structure, migration, and location. The analysis here from the NIDS labor market data, after race and education, uses these factors and controls for the poverty status of high-skilled jobs contribute the most to inequality (17 fathers,29 the level of inequality of the fathers’ incomes, and percent), not job formality or sector. absence of the father. Inequality is measured by the Gini Employment income contributes almost entirely to coefficient on fathers’ incomes calculated separately for income inequality. Decomposing income inequality by each province. Recall information on fathers’ education, sources provides further evidence of the role of the labor market in driving inequality. Figure 93 shows that the 28 This analysis is carried out at the household level. Employment in- come includes wage and business income in 2006, wage income in 2011, and wage and household business income (farm and non- farm) in 2015. Grants include disability, worker’s compensation, and other grants (in 2006); disability, child support, dependency, foster care, grant-in-aid, grants for veterans, and other grants (in 2011); disability, child support, dependency, foster care, grant-in-aid, social relief grants, grants for veterans, and other grants (in 2015). Other in- cludes alimony, pensions, and annuities (in 2006); other income and pensions (in 2011); other income, financial income, and pensions (in 2015). If total income was zero, these observations were deleted in calculating both the source-wise contribution to inequality and the shares. 29 Poverty status of the father is the same as that for the household, but is used as predicted income of the fathers. 68 Overcoming Poverty and Inequality in South Africa occupation, and province lived in 1994 is used to predict fathers who were in the bottom 40 percent. Strong factors father absenteeism.30 In identifying causes of upward are the focus of the analysis, as this is a true representation of mobility, the factors can be weak or strong. Weak factors upward mobility. Education, labor, demographics, location, are those that correlate with the son being in the top 60 and neighborhood factors associated with upward and percent, but this takes place irrespective of the father’s downward mobility are systematically examined. The background. Strong factors both correlate with sons being analysis results are summarized in Table 7. in the top 60 percent, and they operate only over sons of 30 The PSLSD records information on whether fathers are present or ab- sent from the household. A probit model estimates coefficients on father education, occupation, and province, which are then used to predict father absenteeism for sons from the NIDS data. Table 7: Summary of regression results—upward mobility Upward mobility Downward mobility Education +, s -, w Labor market (occupation skill level) +, s -, s Race +, s -, s Location (urbanization, province) 0 0 Family structure 0 0 Migration +, s 0 Neighborhood variables +, s -, s Source. Authors’ analysis. Note: “s” denotes a strong factor; “w” denotes a weak factor. Education has a positive effect on upward mobility. also lowers the likelihood of downward mobility but for Education is measured by the highest level of education sons of all fathers. the son achieves. This can have one of five values—no Higher-skilled occupations are more likely to result in education, primary education (up to grade 6), secondary upward mobility. Compared to the lowest-skilled jobs, education (grade 10 or equivalent), matriculate (grade 12 having a high-skilled job raises the probability of being in or equivalent), and tertiary (a college degree or equivalent). the top 60 percent by 13 percent. For semi-skilled jobs, this Higher levels of education are associated with higher effect does not operate for sons of fathers who were in the probability that the son will be in the top 60 percent. These top 60 percent. Access to formal jobs raises the probability effects, however, operate even for sons whose fathers were of the son belonging to the top 60 percent by 22 percent. in the top 60 percent. Completing secondary education is However, the hypothesis that the effect is null for sons associated with a 17 percent increase, matriculation with of fathers who belong to the top 60 percent cannot be a 34 percent increase, and tertiary education with a 40 rejected. Access to a formal job results in a 19 percent lower percent increase in the likelihood of the son being in the chance of being in the bottom 40 percent. Again, the latter top 60 percent. effect operates over sons of fathers who are in the top 60 Increasing educational attainment implies a lower percent, while it is not possible to reject a null effect for the probability of moving downward. Completing secondary former. education lowers the possibility of being in the bottom 40 White South Africans and South Africans of Indian/ percent by 25 percent, matriculating lowers the probability Asian descent are more likely to rise upward than black by 30 percent, and completing tertiary education lowers the South Africans. Being white increases the probability probability by 34 percent. Completing tertiary education An Assessment of Drivers, Constraints and Opportunities 69 of being in the top 60 percent by 69 percent relative to province of residence differs from that reported under being black. Being of Indian/Asian descent increases the the recall question.31 “Moving” is defined as having lived in probability of being in the top 60 percent by 38 percent. another suburb, town, or village. Those living in provinces These effects are concentrated among those with a poor different from the one they were in 1994 are 13 percent father. Colored South Africans are 13 percent less likely to more likely to move upward. Similarly, sons who report show downward mobility; Indian and white South Africans having lived in another suburb, town, or village are 13 show a similar result, but the effects are not concentrated percent more likely to be in the top 60 percent. This latter on only sons of rich fathers. effect is estimated with the inclusion of province indicators. Both cross- and within-province movements appear to be Geographical location—being in an urban or rural relevant in explaining upward mobility. Downward mobility area, or in any province—is not strongly associated is unaffected by migration. The length of stay at the current with upward mobility. This is a particularly striking result, residence is correlated with neither upward nor downward but ought to be cautiously interpreted for two reasons. mobility. First, the location information refers to present-day status. People may have chosen to move in the past. Second, C. ACHIEVING A MORE EQUITABLE SOCIETY this information is at a relatively high level of aggregation. THROUGH EFFICIENT SOCIAL PROTECTION If segregation operates at a more disaggregated level, this will not be captured. Areas with higher teenage South Africa has a long history of designing and unemployment rates tend to be less likely to have sons implementing social protection programs. The first grants moving upward. That higher teenage unemployment were implemented during the early 1900s, although they rates, and not adult unemployment rates, negatively affect were, like other aspects of South African life, characterized upward mobility indicates that differences in mobility by race-based differences in eligibility or value (Van der arise much before sons formally join the labor force. At the Berg, Siebrits, and Lekezwa 2010). Such differences have mean, one standard deviation increase in cluster teenage now been abolished and the social protection system is unemployment rate reduces the probability of upward an important means of addressing poverty and cushioning mobility by 6.3 percent. Downward mobility is lower for vulnerable households from economic shocks. areas with a higher proportion of black South Africans, with The social protection system is relatively extensive, a single standard deviation increase corresponding to an given the level of development in South Africa (Box 8.2 percent decline in the probability of moving downward. 9). This is a result of the system having initially developed This effect operates solely on those with fathers in the top during the twentieth century for the benefit of the white 60 percent. population, and gradually expanding to cover other Sons who move and sons who live in a province groups (Van der Berg 1997). Expenditure on public social different from their parents are likely to move upward. protection, excluding health care, was estimated at almost Changing provinces is defined using recall information on 5.1 percent of GDP in 2010, sixth highest in Africa, and the province lived in in 1994: for this analysis an indicator comparable to spending in Republic of Korea (5.1 percent), variable was defined that equals one if the present-day Thailand (5.0 percent), and Mexico (5.0 percent) (ILO 2014). 31 This is the answer to question B12 “In which province were you living in 1994?” 70 Overcoming Poverty and Inequality in South Africa Box 9: Elements of the South African social security framework South Africa is characterized by well-designed, means-tested social assistance covering children, working age people, and the elderly. The system is a life-course social security framework, typically associated with European social protection systems, that provides different types of assistance at different stages of an individual’s life. The key elements of the system are shown in Table 8. Table 8: Elements of the South African social security framework Childhood Working age Old age Means-tested child support grants Work-related injury compensation Means-tested social pensions Means-tested care dependency grants Means-tested disability grants Means-tested grant for war veterans Foster care grants Temporary unemployment benefits Occupational pensions Source: Van der Berg, Siebrits, and Lekezwa 2010. The Unemployment Insurance Fund (UIF), the Compensation Fund, and the Road Accident Fund (RAF) are the three key social insurance programs. The UIF is the largest of the three, typically receiving between 700,000 and 800,000 claims annually. The Expanded Public Works Programme (EPWP) aims to provide those of working age with income, work experience, and training for the unemployed. Work opportunities are provided in four sectors: infrastructure, non-state, environment, and culture and social. In 2015/16, the EPWP provided 742,000 work opportunities or 285,000 full-time equivalent jobs. These programs are dwarfed by social grants. With almost 17 million recipients in 2015/16, social grants are the largest intervention in the social security system. Social assistance grants are funded from general tax revenue and are non- contributory. Pension and provident funds and medical schemes are voluntary insurance schemes regulated by the state. Data on membership in pension and provident funds is limited, but it is estimated that in 2011 there were close to 10 million active members. Approximately 8.9 million individuals are covered by South Africa’s various medical schemes in 2016. This number consists of 3.9 million members and their 4.9 million dependents. The population covered by medical schemes grew by 2.2 percent annually between 2007 and 2016, but between 2010 and 2014 it grew slightly slower at 1.5 percent annually. South Africa devotes substantial resources to the social is 0.8 percent of GDP and among upper middle-income assistance system (Figure 94). In 2015, spending on social countries the proportion is 1.4 percent. Thus, relative to GDP, assistance in South Africa was equivalent to 3.0 percent South Africa spends almost four times the Sub-Saharan of GDP. This figure places the country within the top 15 Africa median and 2.2 times that of upper middle-income percent of countries for which there is data in the World countries. Relative to other Sub-Saharan African countries, Bank’s Atlas of Social Protection Indicators of Resilience South Africa is an outlier in terms of its spending on social and Equity (ASPIRE), ahead of countries such as Brazil (1.3 pensions, which is almost 30 times higher as a proportion percent), the Russian Federation (1.9 percent), Colombia of GDP, and on cash transfers (6.8 times the Sub-Saharan (2.4 percent), and Kenya (2.5 percent). Africa median). In comparison with other upper middle- income countries, South Africa devotes a relatively large Compared with other African countries, South Africa amount of resources to public works programs (9.7 times allocates more to social assistance as a proportion the upper middle-income country median), cash transfers of GDP than any other country for which there is (3.4 times), and social pensions (3.1 times). data. Within Sub-Saharan Africa, the median proportion An Assessment of Drivers, Constraints and Opportunities 71 Figure 94: Spending on social assistance as percent of GDP Source: World Bank (2017a). Notes: Most recent estimates, 2010–2015. Spending on social grants has grown over the past contrast, though, real spending on the disability grant fell decade due to an increase in coverage (Figure 96). by an average 3.2 percent per year, while total spending on Between 2005/06 and 2015/16, total spending on social all other grants grew by 4.2 percent per year in real terms. grants grew from R99.4 billion to R134.3 billion (2016 Thus, the composition of spending on grants changed, prices). This is equivalent to a real growth rate of 3.1 as the old age and child support grants grew within total percent per year, with higher growth during the first half spending at the expense of the disability grant. By 2015/16, of the period (3.8 percent). Thus, real spending on the Child the old age grant accounted for 41.4 percent of spending Support Grant (CSG) rose from R27.7 billion to R49.5 billion on social grants, followed by the CSG at 36.9 percent (2016 prices) over the period. Growth in spending on the and the disability grant at 14.9 percent. All other grants old age grant was less rapid, although it still averaged accounted for just 6.8 percent of total spending. 3.8 percent per year in real terms over the full period. In 72 Overcoming Poverty and Inequality in South Africa Figure 95: Real expenditure on social grants, 2005/06– Figure 96: Social assistance coverage rates across 2015/16 quintiles Source: Authors’ calculations, SASSA (2016) and S Statistics South Africa (2017). Source: Authors’ calculations and World Bank (2017a). Notes: Figures in are expressed in 2016 prices (CPI 2016=100). Full details can be Notes: Includes direct and indirect beneficiaries of social assistance found in the background note. programs. An individual is covered if they reside in a household in which any member receives social protection transfers. Official SASSA data indicate that the system of social therefore, extremely reliant on social assistance transfers, assistance expanded its number of beneficiaries with wage income playing a very small role in enabling the even more rapidly. Between 2005/06 and 2015/16 the poorest households to support themselves. For quintile 2, number of grant beneficiaries increased by 4.7 percent grant income is more than one-third of total income and per year from just under 11.0 million to just under 17.0 is only slightly less important within total income than million. Given the mid-year population estimate for 2015 wages (37 percent of total income). At the upper end of of just under 55 million (Statistics South Africa 2015), this the distribution, grant income represents just 0.5 percent of implies a coverage rate of just under 31 percent. In terms total income, compared with 67 percent for wage income.32 of the number of beneficiaries, the CSG is the largest, with In 2015, social assistance transfers are estimated to 11.97 million beneficiaries in 2015/16, 70.3 percent of have reduced the poverty headcount rate in South the total. This is followed by the old age grant, with 3.19 Africa by 8 percent and the poverty gap by about million beneficiaries (18.5 percent of the total), and the 30 percent (Figure 97). These reductions are similar in disability grant with 1.09 million beneficiaries (6.7 percent). magnitude to those in 2010/11. In an international context, Together, these three grants account for 95.6 percent of though, South Africa does not perform particularly well in all beneficiaries. The overall expansion in the number of terms of the ability of the social assistance system to reduce beneficiaries was driven by the CSG, which accounted for the poverty rate. South Africa’s reduction is slightly above eight out of ten new beneficiaries over the decade. the global and Sub-Saharan African average and is similar For the poorest pre-transfer quintiles, grant income to that of Latin America and the Caribbean (8 percent), but represents a substantial boost to total household it is significantly lower than the reduction observed among resources: grant income accounts for 71 percent of upper middle-income countries (14 percent). However, if total income in quintile 1, compared with just 9 percent 32 This pattern—of the poorest households being extremely dependent for wage income. Thus, total grant income is more than on social grants and relatively isolated from wage earners—has been seven times the size of total wage income for the poorest previously documented in South Africa (Klasen and Woolard 2009 and Leibbrandt et al. 2010a) and highlights the critical role of social pro- 20 percent of the population. Quintile 1 households are, tection. An Assessment of Drivers, Constraints and Opportunities 73 weight is attached to individuals further below the poverty rate in South Africa and in the average country in Latin line, South Africa’s performance is better. In reducing the America and the Caribbean are similar, the impact on the poverty gap by 32 percent, the poverty-reducing impact poverty gap is almost twice as strong in the former than of South Africa’s social assistance system ranks ahead of the in the latter (32 percent compared with 17 percent). This average upper middle-income country (27 percent) and suggests that where the South African system is particularly far ahead of the average Sub-Saharan African country (15 successful is in reaching the poorest individuals. percent). Interestingly, while the impacts on the poverty Figure 97: Simulated poverty reduction associated with social assistance programs Source: Authors’ calculations and World Bank (2017a). Notes: Poverty reductions are simulated assuming the absence of social assistance programs and are expressed as a proportion of the pre-transfer poverty measure. The inequality-reducing impact of social assistance income countries, the Gini coefficient is reduced by 1.7 is significant (Figure 98). In 2014/15, social assistance percent by social assistance transfers, while the reduction transfers reduced the Gini coefficient in South Africa by an is 0.7 percent in Sub-Saharan Africa and 1.6 percent in Latin estimated 10.5 percent, a slightly stronger impact than in America and Caribbean countries. This is clearly an area 2010/11. No other regional or income grouping average where the South African social assistance system is very effect comes close to this level of impact: in upper middle- effective. 74 Overcoming Poverty and Inequality in South Africa Figure 98: Simulated inequality reduction associated with social assistance programs Source: Authors’ calculations and World Bank (2017b). Notes: Inequality reductions are simulated assuming the absence of social assistance programs and are expressed as a proportion of the pre-transfer inequality measure, in this case the Gini coefficient. SUMMARY education have decreased. In other words, the overall population has been more educated since 2006, and that South Africa experienced significant reduction in has helped reduce poverty; however, returns to education consumption-based poverty between 2006 and 2015. are lower now than they were in the past. Urbanization, A combination of demographic, location, education, and demographic changes, and expansion in the provision of employment attributes contributed to poverty reduction. services all contributed to improvement in the welfare of Decomposition of changes in the incidence of poverty for households. the period suggests growth contributed more to overall While racial lines continue to determine poverty poverty reduction compared to redistribution. Labor and inequality levels, the skills and labor market income is the largest contributor to improving the lives incomes are an increasingly important determinant of of people at a national level and in urban settings, but inequality. The role of race is falling while skills and labor less so in rural areas. Grants and pensions contributed 24 related factors are growing in explaining inequality. Like the percent to poverty reduction but 44 percent to reducing inequality of outputs, race, education, labor are the main the poverty gap. This finding confirms the targeting factors explaining inequality of opportunity. Black South effectiveness of South Africa’s social safety net programs. Africans are less likely to be upwardly mobile and more While improvement in skills and education are key elements likely to remain at the bottom. However, racial differences to significant poverty reduction, over time, returns to An Assessment of Drivers, Constraints and Opportunities 75 are not the only reason for low mobility. Education, labor of inequality, the inequality-reducing impact of social markets, spatial segregation, and migration strongly affect assistance was significant when inequality estimates were chances of upward mobility. Skill and education matter calculated without the transfers. In 2015, social assistance for intergenerational mobility. Higher-skill occupations are transfers reduced the Gini coefficient by an estimated likely to give rise to greater mobility, as does a higher level 10.5 percent, a slightly stronger impact than in 2011. of education. Similarly, neighborhood and labor effects are Introducing redistribution polices related to wealth and important in upward mobility. land management could further reduce inequality. South Africa has made progress toward creating an efficient South Africa’s social protection system is a major social protection system, but further expansion is likely intervention aimed at ameliorating poverty and to be unsustainable due to the low growth and financial helping vulnerable households deal with unforeseen constraints. The challenges of high inequality should shocks. Close to 17 million low-income individuals got be solved by the creation of productive jobs and further access to the means-tested direct transfers. In 2015, improvement in the efficiency of services delivery. social assistance transfers are estimated to have reduced the poverty headcount rate by about 8 percent and the poverty gap by 30 percent. Despite the stagnation 76 Overcoming Poverty and Inequality in South Africa CHAPTER 5 LABOUR MARKET DYNAMICS AND POVERTY Having an employed household head is not necessarily A. DYNAMICS AND CHALLENGES IN LABOR associated with a lower vulnerability to poverty—a large MARKET OUTCOMES proportion of the population consists of the working poor who earn very low wages. To unlock the full potential of labor The South African labor force is characterized by high markets to accelerate the reduction of poverty and inequality, levels of unemployment, low participation, and many jobs need to be created and wages increased at the same time. discouraged work-seekers and non-seekers. The spatial This would include reducing the current persistent high level of separation of the country and the inaccessibility of jobs to unemployment. Race still affects the ability to find a job, as well much of the working age population in rural and remote as the wages received once employed. Although more women areas has resulted in many discouraged work-seekers and now participate in the economy, female participants find it non-seekers. While employment has increased in absolute harder to find a job, and earn less than men when they do. terms since the onset of democracy, employment growth There is strong evidence of structural mismatch between labor has not matched either population growth or the rate of demand and labor supply for unskilled workers. Moreover, growth of worker supply. Consequently, employment rates despite extremely high and rising unemployment, skilled labor as a share of the population aged 15 or older fell as share can be difficult to find. Location matters for labor market of labor force participation from 2000 to 2015 (Figure 99). outcomes, with people in urban areas having better prospects of getting a job and a higher probability of getting a formal job. Location has implications for travel costs, which can be a burden for getting jobs. An Assessment of Drivers, Constraints and Opportunities 77 Figure 99: Key labor market trends 2000–2016 Figure 100: Labor force participation rates, unemployment, and dependency ratios, by country (selected years) Source: World Bank Indicators, http://data.worldbank.org/ Note: The dependency ratio is the ratio of the non-working age population to the indicator/, 2017. working age population, represented as the proportion of dependents per 100 working-age population. The unemployment rate has been high and persistent. Economic growth is too low to generate sufficient jobs. The narrow measure of the unemployment rate remained According to the World Bank’s South African Economic consistently high (21–26 percent) throughout the Update of September 2017a, since 2008, 3.5 million 2005–2015 period. The unemployment rate increased people have entered the labor force, but only 1.6 million from 22.5 percent in 2008 to 25.1 percent in 2015, and additional jobs have been created. Nearly 6.2 million to 27.7 percent in the first half of 2017. Unemployment, people are unemployed, or 9.3 million if those who have in the narrow sense,33 has therefore increased by about 5 stopped looking for work are included. Of those looking for percentage points. The labor force participation rate was employment, 3.5 million (57 percent) have not worked in virtually unchanged at the end of 2015 compared to early the past five years. This number has increased by nearly 34 2005, at just over 53 percent. The broader unemployment percent since 2008. rate, which includes those in the labor force who were South Africa has a very high unemployment rate discouraged and no longer searching for jobs, was between compared to its peer economies or those within 10 and 15 percentage points higher than the narrow rate, the region. Figure 100 puts these figures into context depending on the period considered. Accounting for non- by presenting labor market indicators for South Africa searching unemployed, the proportion of the labor force alongside international comparators. A potential reason employed dropped to 68 percent in 2015. Including those for this is South Africa’s high proportion of discouraged discouraged workers, South Africa’s unemployment rate work-seekers (non-searching unemployed). As was reached 36.6 percent in the first half of 2017. observed in the expanded unemployment rate, while the unemployment rate in comparator regions has generally decreased over time, South Africa’s unemployment rate 33 Narrow unemployment is defined as unemployed, willing to work, has increased by more than 8 percent. South Africa also and having actively searched for a job in the last four weeks. 78 Overcoming Poverty and Inequality in South Africa has a high proportion of dependents relative to those who differences in unemployment were evident in each year, can participate in the labor market, and this proportion with the unemployment rate among black South Africans decreased by almost 13 percent between 1995 and 2015. the highest at around 28 percent, and unemployment among white South Africans the lowest at 5–6 percent. Youth joblessness was extremely high throughout the Finally, although the male narrow unemployment rate period, and post-secondary education became less of a increased by about 1 percentage point over the period, the buffer against unemployment. According to the narrow female narrow unemployment rate dropped. In the final definition of unemployment, 40 percent of those between quarter of 2015, the male unemployment rates were 22.3 ages 20 and 29 were unemployed throughout the 2005– percent compared to a 26.3 percent rate for females. 2015 period. Unemployment rates were lower for the older age cohorts, generally around 22 percent for the 30–39 The employment rate, defined as the employment- cohort and 15 percent for the 40–49 cohort. The relationship to-population ratio for those aged 15 and above, between education and unemployment changed over remained around 40 percent throughout the period. the period. The unemployment rate for those with post- Some of the main changes in the composition of the labor secondary education was 7.2 percent at the end of 2005 market are highlighted in Figure 101. and rose to 11 percent by the end of 2015. Very large racial Figure 101: Trends in South African employment Source: LCS surveys, staff calculations from PALMS V3.2 data. An Assessment of Drivers, Constraints and Opportunities 79 The employment outcome is worse for females than with fewer than 50 workers), which dropped to 68 percent for males; however, the gender-employment gap has in 2010, and to 67 percent in 2015. The nature of work been closing. In 1995, females were 9 percentage points changed over the period as well, as measured by hours less likely to be employed than males, but in 2015 the worked per week. In 2005, the median and mean number figure had decreased to 5 percentage points. As expected, of hours worked per week was 45. This decreased to a labor market outcomes are also better for individuals with median of 40 and a mean of 43 in 2010, and both remained a high level of education, although the gap in employment at this level in 2015. outcomes between those with no education and those The structural transformation in the economy saw with tertiary education has decreased. significant increases in jobs in the service and finance The share of black South African workers in the labor sectors, but large drops in the number of agriculture force increased, as did the share of women. In 2005, and manufacturing jobs. Services added more jobs than about 69 percent of workers were black South Africans, any other sector. The finance sector added just under but this had increased to 73 percent by the end of 2015. 800,000 jobs and made up 14 percent of the labor force in Simultaneously, the proportion of colored South African 2015 compared to 10 percent in 2005. A huge fall occurred workers decreased in both the number and proportion to in the number of manufacturing jobs, while the number white South African workers over that period. Just under of workers in the construction sector rose. The closing gap one million more men were employed in 2015 compared between the number and share of workers in these two to 2005, while the corresponding increase for women sectors over the period is notable. The number of workers was just under 1.5 million. It resulted in an increase of the in agriculture dropped almost by half between 2005 female share of the labor force from about 42 percent to 45 and 2010. The sector recovered over the next five years percent over the period. by adding about 170,000 jobs, but overall there was a 4 percentage point drop in the share of the agricultural labor There was a shift to a more educated labor force, force. Although the number of jobs in the mining sector leading to an increasing share of high-skilled jobs was relatively stable, the share of mining in the total labor in the economy. The proportion of workers with post- force decreased as the overall number of jobs. secondary education rose by 4.5 percentage points over the period, but almost all the gains took place between Labor market productivity has increased in all sectors 2005 and 2010. There was also an increase in the share but one; financial services had lower employment of workers who completed secondary school but did not growth relative to the growth of the sector. Figure 102 go on to post-secondary education. This increase—from estimates the value-added growth between 2000 and 2016 27 to 31 percent—was spread quite evenly over the full and the corresponding change in sectoral job creation. period. In line with this increase in the supply of more Each bubble represents the relative size of employment in highly educated labor force participants, the share of those that sector in 2016. Bubbles above the 45-degree line are working in high-skilled jobs increased by 5 percentage sectors where employment growth exceeded their output points from 2005 to 2010, mainly due to a relative shift out growth. The exception to this is the financial services sector, of medium-skilled jobs. although this result is driven by the rapid expansion of the temporary employment services. The retail; utilities; and The public sector added about 700,000 jobs, and there community, social, and public (CSP) services (including was a decrease in the proportion of the labor force government services) sectors have been important in employed in small and medium enterprises (SMEs). Just increasing their ability to create employment. under 72 percent of jobs in 2005 were in SMEs (enterprises 80 Overcoming Poverty and Inequality in South Africa Figure 102: Sectoral gross value-added and employment growth, 2000–2016 Source: Labour Force Survey, South African Reserve Bank, 2017; authors’ calculations. Notes: AGR = agriculture; MAN = manufacturing; MIN = mining; WRT = wholesale and retail trade; TRS = transport; PHH = private households; UTI = utilities; CSP = community, social, personal services; FIN = financial services; CONT = construction. Mining and agriculture performed poorly over the Skills intensity increased in most sectors. Figure 103 and period. Growth averaged 0 percent for mining between Figure 104 show the proportion of skilled, semi-skilled, and 2001 and 2016, coupled with a decrease in employment unskilled labor by sector and their growth between 1995 of 2 percent on average. Agriculture grew by a diminutive and 2015. In the post-apartheid era, every major sector has 1.9 percent and faced an employment contraction of witnessed an increase in skills intensity, pointing to a labor 2 percent. While the construction sector is the fastest- demand trend that has become skills-intensive over time. growing sector in employment and GDP terms, it is one of Excluding domestic work, the highest increases in skills the smaller sectors. intensity are in the financial services, construction, and agricultural sectors. Figure 103: Growth of employment shares by sector and Figure 104: Composition of employment by sector skills level, percent share: 1995–2015 and skills level, percent share: 2015 Source: Labour Force Survey, 1995, 2015, authors’ calculations. An Assessment of Drivers, Constraints and Opportunities 81 This increase has been at the expense of semi-skilled participation outcome is important to understanding the and unskilled workers: the share of semi-skilled work South African labor market. decreased for all sectors except agriculture between Education has a strong influence on the probability 1995 and 2015. This decreasing share of semi-skilled of labor market participation. Figure 105 shows labor across all non-agricultural industries is in part a that education is strongly associated with labor force function of the growth of capital intensity, the adoption participation and the probability of participation increases of advanced technologies, and possibly an avoidance with education level. In 1995, those with post-secondary of perceived regulatory burdens. The shrinkage of semi- education were 34 percent more likely to participate in the skilled employment points clearly to the existence of a labor market than those with no education. In 2015, this “missing middle” in the labor market. That is, the rise in probability increased to 48 percent. Similarly, people with skills-intensive employment has hollowed out the middle secondary education have increased their probability to of the distribution and is a likely contributor to increased participate from 7 percent in 1995 to 23 percent in 2015. labor market inequality. Females participate less than males, but black B. EXPLAINING LABOR MARKET South Africans and married individuals have higher PARTICIPATION AND EMPLOYMENT participation rates. For an individual in a household with a higher number of children under age 7, and between Pre-labor market differences affecting the way 8 and 15, there is a negative impact on the probability of individuals choose to participate in the labor force labor market participation. The same applies for individuals are widespread. Differential provision of education, in households with at least one pensioner. Women are less training, and access to public services for different race likely to participate in the labor market compared to men, groups all affect the labor market participation outcome, but this probability decreased from 20.4 percent in 1995, which precedes the employment outcome (Knight and to 12.8 percent in 2015. Married individuals are more likely McGrath 1987, Moll 1991, Case and Deaton 1997). With the to be labor force participants, but the probability is falling widespread incidence of discouraged workers, modeling over time. Figure 105: Determinants of labor force participation outcome, marginal effects for selected years a. Effect of education b. Gender, race, and family structure Source: Post-Apartheid Labor Market Series, authors’ calculations. Notes: To understand the determinants of labor force participation, logit models were estimated with labor market participation as the dependent variable, taking on a unitary value if an individual is either employed or unemployed, and a zero value if an individual is not economically active. * p < 0.1, ** p < 0.05, *** p < 0.001. Controls include province and urban status not reported here, urban/rural status not reported in survey between 2005 and 2007. 82 Overcoming Poverty and Inequality in South Africa Like labor force participation, employment depends black South Africans facing a 17 percent lower likelihood of on human capital characteristics.34 Following existing employment than white South Africans. Similarly, colored literature,35 age, education, gender, marital status, race, and South Africans were 14 percent less likely to be employed location are all assumed to be correlated with employment than white South Africans in 2015. outcomes in South Africa. After controlling for location and South Africa’s path of structural transformation has human capital characteristics such as education and age, been unique. Unlike other Sub-Saharan African countries, race remains a significant determinant of employment the proportion of employment created by subsistence outcomes. In 1995, black South Africans were 15 percent agriculture in South Africa has always been small. For less likely to be employed than white South Africans. The example, in 2001 the agricultural sector contributed only gap between white and colored South Africans was smaller 10 percent to the total employment share. Therefore, in but still significant, at 9 percent. By 2015, the difference in estimating structural transformation models, the primary employment probability due to race had increased, with sectors (mining and agriculture) was considered as the base against which transformation would be measured, instead 34 A probit model is used to analyze the discrete states “employed” and “unemployed” among labor force participants, with “unemployed” as of just agriculture. Models estimating the probability of the reference category. The employment outcome is estimated sepa- rately in five-year intervals between 1995 and 2015. working in the non-services versus the services sectors 35 See, for example, Bhorat and Goga 2013, and Kingdon and Knight are estimated to identify what it takes for an individual to 2004. obtain a job in the fast-growing services sector (Box 10). Box 10: What does it take for an individual to obtain a job in the fast-growing services sector? Following the methodology posed by (Paci 2016) this section adopts models to focus on the determinants of economic transformation. The model exploits the heterogeneity in individual micro- and macro-level endowments to identify the drivers of structural transformation. To explore the relationship between individual and household characteristics and whether an individual is likely to be employed in the services versus the non-services sector, the following model is estimated: Pr(yi,t | Xi,t)= G(β0 + Xi,t’β ) where G is a logistic function Where yi,t = 1 if the individual is employed in the services sector, including retail and wholesale trade, transport, financial services, or the CSP sector. Similarly, yi,t = 0 if the individual is employed in any non-services sector, which includes agriculture, mining, manufacturing, electricity, and construction. The vector of explanatory variables Xi,t consists of individual and household characteristics that control for sex, age, age squared, marital status, and highest level of education attained (no schooling, primary schooling, secondary schooling, or post-secondary education). Xi,t also contains household-level controls, including the proportion of the household under age 7, between 8 and 15, between 16 and 65, and over 65, and dummies for residence in urban areas, province, and a set of interaction variables between province and geographical location. An Assessment of Drivers, Constraints and Opportunities 83 Figure 106: Probability of services sector employment, individual effects: 1994–2015 Gender Location Human capital effects Race Source: Post-Apartheid Labour Market Series, authors’ calculations. Women are more likely than men to be employed in had 12 years of education (corresponding to secondary the services sector. Figure 106 shows that the probability education), alluding to a large skills mismatch between of working in the services sector is higher for already employment and potential labor absorption. employed females relative to males, and for those who Conditioned on already having a job, black and colored reside in urban areas. The probability of finding a job in the South Africans face a lower probability of working in services sector, conditioned on already being employed, the services sector as opposed to white South Africans. has increased over time for females. This coincides with the Put differently—the services sector in South Africa is a gendered structure of the primary and secondary sectors. disproportionate employer of white workers, showing that The probability of being employed in the services the economic gains of job security and the higher pay sector is only positive for those who have post- associated with working in the services sector belongs to secondary education. The returns to primary and a population group that is still in the minority, significantly secondary education measured by the probability of perpetuating a specific pattern of inequality observed in employment in services (conditional on being employed) the labor market. is negative, and these probabilities have been decreasing Notably, however, the probability of black or colored over time. This reinforces the descriptive evidence that the workers in the services sector has increased since the services sector, which corresponds to 71 percent of the mid-1990s. The figure also shows that colored individuals, employment share of the country, is highly skills-biased. who constitute about 11 percent of the labor force, are the At the same time, in 2015 the median employed individual 84 Overcoming Poverty and Inequality in South Africa most marginalized group, facing even lower probabilities services sector has driven the demand for skilled labor. of employment in the services sector than the black African However, skilled labor makes up only a small proportion population, which accounted for over 70 percent of the of the labor force, implying that the largely unskilled and labor force in 2015. semi-skilled workers who have found themselves without work because of the contracting primary sector, have C. STRUCTURAL MISMATCH BETWEEN not been able to enter jobs offered in the services sector. LABOR DEMAND AND LABOR SUPPLY FOR This structural mismatch between labor demand and the UNSKILLED WORKERS supply of unskilled workers remains a key marker of South Africa’s skills-biased labor demand trajectory. Using four waves of the NIDS survey data, this section investigates factors associated with getting a job, labor Education is important in transition to labor force participation, and wage levels. These data span eight force participation, but less affiliated with finding years and attempt to follow the same people over time. The employment. As Table 9 shows, a higher level of education panel nature of the data is used to analyze what leads an is associated with a higher probability of being part of the individual to find employment over time. The results of the labor market (either employed or unemployed). However, logistic multinomial analysis are presented in Table 9. only tertiary education gives higher probability of finding employment in general. Other coefficients are insignificant A structural mismatch between labor demand and suggesting low correlation with ability to find a job.36 labor supply for unskilled workers is strongly evident in the South African economy. Sectoral growth has 36 The result is generally in line with a recent study by the International Monetary Fund that suggested “previous experience is an important primarily been serviced-based, and the growth in the determinant of job-finding rates, while education has almost no ef- fect.” The study based the conclusion on the QLFs panel data and sug- gested that the job-finding rate does not differ substantially across different education groups and race. An Assessment of Drivers, Constraints and Opportunities 85 Table 9: Determinants of labor force participation and employment transitions   (1) (2) (3) (4) (5) (6) (7) Participate Dependent variable in LF Find employment   Overall Overall Skill level 1 Skill level 2 Skill level 3 Informal Formal Race (black=base)               Colored 0.024 0.028* -0.015 0.017 0.019 0.009 0.004 Indian -0.099*** -0.046 0.0394 -0.181*** 0.022 -0.049 -0.076 White -0.047* 0.007 -0.137*** -0.077 0.015 -0.034 -0.150* Marital Status (married=base) Living with partner 0.071*** 0.029 0.0732*** -0.014 -0.010 0.028 0.018 Widow/widower 0.037 0.028 0.0163 -0.031 -0.019 -0.060 0.049 Divorced/separated 0.056 0.063 0.0653 0.034 -0.063 0.094 -0.039 Never married 0.056*** 0.020 0.0753*** 0.032 0.021 0.069*** 0.042* Education (no education=base) Primary education 0.044** -0.007 -0.0258 -0.026 -0.009 -0.051 0.011 Secondary education 0.088*** 0.007 -0.166*** 0.102*** -0.018 -0.068 0.015 Matric 0.130*** 0.025 -0.164*** 0.127*** 0.0174 -0.066 0.073*** Tertiary 0.294*** 0.189*** -0.207*** 0.137*** 0.275*** -0.093* 0.254*** Demographics location               Male 0.094*** 0.109*** 0.012 0.146*** 0.014 0.102*** 0.078*** Urban 0.026*** 0.023*** 0.005 0.006 0.021** -0.007 0.035** Household size 0.000 -0.002 -0.004* -0.007*** -0.001 -0.008*** -0.002 Household head 0.089*** 0.100*** 0.061*** 0.089*** 0.014 0.085*** 0.063*** Age 0.072*** 0.050*** 0.017*** 0.025*** 0.001 0.022*** 0.019*** Age squared -0.0009*** -0.0006*** -0.0002*** -0.0003*** 0.000 -0.0003*** -0.0002*** Transfers and other               Log (state transfer) -0.005** -0.008*** 0.001 -0.003 0.002 0.003 -0.006 Poor -0.05** -0.200*** -0.071** -0.219*** -0.037* -0.112*** -0.219*** Poor X log transfer -0.003 0.003 -0.008* 0.004 -0.003 -0.011** 0.006 Log (transport) 0.009*** 0.010*** 0.014*** 0.017*** 0.006** 0.015*** 0.017*** Constant -0.942*** -0.593*** -0.0675 -0.274** 0.000 -0.114 -0.204**                 Observations 23,763 30,945 5,595 6,123 5,040 6,266 5,704 R-squared 0.192 0.208 0.095 0.163 0.181 0.102 0.186 Source: Authors’ estimation based on NIDS panel data. Logistic multinomial. Low correlation between education and the probability probability of entry into jobs requiring low or intermediary of finding employment masks heterogeneity in the skills is higher for people with lower levels of education. role of education in finding jobs in different skills However, highly skilled jobs are associated with tertiary requirement categories. People at different levels of education. Jobs that require low and intermediate skills education compete for different types of jobs. It is not are not attractive enough for people who have invested surprising that the probability of getting low-skill jobs in education, who prefer to wait for jobs appropriate to is negatively associated with the level of education. The their training. Having secondary or matriculate education 86 Overcoming Poverty and Inequality in South Africa helps in getting low- and mid-skill jobs but not enough Once employed, education and skills result in to get highly skilled positions. Having tertiary education substantial wage increases. Wages are higher for each gives a better chance of getting highly skilled job, but the successive level of education: by wave 4, a college degree number of these positions is relatively small. In other words, results in a 148 percent increase in wages relative to secondary and matriculate education does not necessarily no education, while matriculating implies a 63 percent give a better chance of getting a high-skill job in South rise. Figure 107 (panel a) shows coefficients from Mincer Africa. Tertiary education gives higher probability of getting regressions of log wages on education levels: these mid- and high-skilled job, but the number of these positions coefficients are estimated relative to no education. A similar is low, keeping high proportion of highly educated people pattern exists with skills levels: by the fourth NIDS wave, unemployed. jobs at the highest skill level have wages that are 80 percent higher than jobs for the lowest skill level. Figure 107: Skill mismatch a. Returns on education and composition of b. Returns on skills unemployment Source: NIDS, base = no education for wage regressions. Coefficients from Source: NIDS, base = skill level 1 for wage regressions. Coefficients from Mincer regression with log wages dependent variables, education, skills, Mincer regression with log wages dependent variables, education, sectors, skills, and other repressors are included. sectors, skills, and other repressors are included. D. RACIAL AND DEMOGRAPHIC FACTORS to enter formal sector employment. Wages show a distinct DEFINING EMPLOYMENT racial divide across all job categories. Black South Africans earn much less, on average, than white South Africans, who One of the more distinct features of South Africa is its legacy earned 87 percent higher wages in wave 4. of apartheid, a system designed to exclude black South The dichotomy in finding employment can be Africans from full participation in the labor force. Even 24 explained by rising disparity within the black South years after the end of the system, race still affects the ability African group (Leibbrandt et al. 2010b, Bhorat 2004), with to find jobs, as well as the wages received once employed. some black South Africans earning substantially higher Racial differences alter the probability of finding incomes. The coefficient of variation for wages of black employment for low-skill and formal jobs. Black South South Africans has risen substantially over the four waves Africans are 16 percent more likely than white South (from 3.83 to 6.37, a 66 percent rise), while it has fallen for Africans to enter low-skilled jobs and 18 percent more likely white South Africans (from 8.36 to 6.44, a 23 percent fall). An Assessment of Drivers, Constraints and Opportunities 87 Although an increased number of women participate wages for men are 28 percent higher than for women in in South Africa’s economy, female participants have a wave 1. Men and women are on an equal footing only for harder time finding a job and earn less than men when low-skilled jobs: for these jobs, there are no statistically they do. From 1993 to 2008, the participation rate for significant differences by gender in the probability of women increased by 38 percent (Leibbrandt et al. 2010b) finding a job. Figure 108 lays out these results: panel A shows and has remained constant since then. Despite this, men the elasticities of entering employment across different job are 11 percent more likely to transition into employment.37 categories, and panel B shows estimated coefficients from When employed, women earn substantially less than men: a Mincer regression on an indicator variable for whether the 37 In the NIDS data, 54 percent of men are employed compared to 39 individual is male. percent of women. Of those women who are employed, 42 percent work low-skill jobs, while only 27 percent of employed men work low- skill jobs. Figure 108: A gender gap holds except for low-skill jobs a. Elasticity of entering employment to wage increase; b. Wages for men, relative to women, percent male relative to female (base = female) difference (base = female) 37% 34% 34% 0,11 0,12 0,10 28% 0,07 0,08 0,01 Overall Skill Skill Skill Informal Formal Level 1 Level 2 Level 3 Wave 1 Wave 2 Wave 3 Wave 4 Source: NIDS wave 4. Results from logistics multinomial regressions and Mincer regressions. Authors’ estimations. Males have a higher probability of getting skilled higher probability of getting a formal job, but there employment. Males are 11 percent more likely to get is no significant difference across provinces. Generally, employment than females. The highest probability urban areas have a 3 percent higher probability of finding for males are in informal (11 percent) and mid-skills a job, and especially formal jobs. There are no significant occupations (12 percent). Age is a proxy for experience and differences across provinces in terms of probability of older people have a higher probability of getting jobs than finding employment. young counterparts. The probability to get employment is A transition from rural to urban areas would also increasing by almost 5 percent per year. This means people accompany structural change in South Africa. It is easier with 20 years of experience have almost a 50 percent to find jobs in urban areas: the probability of finding an greater chance of getting employment than young new urban job is 3 percent higher compared to rural areas. Urban entrants to the labor market. The middle-aged have the jobs pay more, but the differential falls over time. In wave highest probability of getting jobs. 1, wages for urban jobs were 32 percent more than rural jobs. By wave 4, however, the gap between urban and rural E. GEOGRAPHICAL SEGREGATION AND ROLE wages was reduced to 17 percent (Figure 109). Consistent OF INTERNAL MIGRATION with this trend, from the Oaxaca-Blinder decomposition, People in urban areas have better job prospects and returns on jobs in urban areas also fall over time. 88 Overcoming Poverty and Inequality in South Africa Figure 109: Urban wage differentials and formal sector wages a. Urban wage differential declines (base = rural) b. Constant formal sector premium (base = informal) 32% 45% 41% 36% 33% 18% 17% 12% Wave 1 Wave 2 Wave 3 Wave 4 Wave 1 Wave 2 Wave 3 Wave 4 Source: NIDS waves 1–4. High travel cost is a burden for getting jobs. In South F. DIMINISHED ROLE OF SMALL, Africa, workers travel long distances for work and spend MEDIUM, AND MICRO ENTERPRISES IN significant time and money commuting, with a large share EMPLOYMENT GENERATION of resources spent on taxes, work uniforms and clothes, and child care. Workers, especially in the townships, Small, medium, and micro enterprises (SMMEs) have commute far to work and high travel costs are a necessary been identified as a key component to advancing burden of having a job. Such costs are burdensome for all inclusive growth and development in South Africa. forms of employment, including formal and high-skilled The NDP highlights the importance of these businesses occupations, and are negatively associated with the for job creation, innovation, and competitiveness, with probability that a person will accept a job. For the working the goal that 90 percent of new jobs will be created by poor, these costs consume a large portion of their earnings. SMMEs in South Africa by 2030. The successful entry and The unemployed, especially youth, tend to lack resources growth of an SMME may create a sustainable mechanism and mobility for a job search or the ability to relocate for through which the wages of those at the bottom of the a distant job. In some cases, underdeveloped transport distribution can be increased and the level of inequality systems, high cost of commuting, and crime makes the job reduced. Entrepreneurship has often been presented as search more difficult and raises associated expenses and an alternative for the unemployed who are unable to be reservation wages. absorbed into formal employment. This view is supported by the international literature. For example, van Praag and Poor people generally have a significantly lower Versloot (2007), in a systematic review of 56 studies, finds probability of getting a job. Controlling for other factors, that entrepreneurs are an important source of job creation being poor reduces the probability of getting a job by 20 and that there are positive, long-term spill-over effects to percent. The probability has especially low association with entrepreneurship that increase employment growth rates. formal jobs and skilled professions. Furthermore, supporting the growth of existing SMMEs Government transfers have a very small impact on could encourage innovation and employment creation in employment. The impact of the transfers on the decision these businesses. to participate in the job market was estimated by inclusion of the level of transfers to households in the logistic regression. An Assessment of Drivers, Constraints and Opportunities 89 The extent to which SMMEs,38 and entrepreneurship community and social services (23 percent), financial (14 more generally, have been harnessed to increase percent), and construction (11 percent). The breakdown of employment and reduce inequality in South Africa SMMEs to more disaggregated firm sizes reveals that more has been disappointing. In low-income countries, formal than 50 percent of own-account workers operate within the and informal SMMEs contribute more than 70 percent to wholesale and retail sector, a proportion that decreases as employment and 60 percent to GDP. In middle-income the size of the SMME increases. These SMME wholesale and countries, the SMME contribution to employment and GDP retail jobs are typically categorized as low-skill occupations, is higher, at 95 and 70 percent respectively (Ayyagari et al. such as shop salespeople, petrol attendants, street 2007). Conversely, South African SMMEs employ around vendors, and cashiers. Of the elementary workers, most 56 percent of the labor force (DTI 2008) and contribute an report functioning as farm hands and laborers, street food estimated 45 to 50 percent to GDP (DTI 2004). Forty-five vendors, and helpers and cleaners in offices. Other major percent of firms are small in South Africa—considerably job functions reported among SMME workers were shop lower than any of the regional averages. Furthermore, salespeople and petrol attendants, other protective service South Africa has a relatively large share of large firms. workers (rangers and game wardens), cooks, bricklayers and stonemasons, and motor vehicle mechanics. The SMME sector has declined over the 2005–2016 period last decade and tends to focus on low-skill The unemployed are more likely to find a job in small wholesale operations. Based on Quarterly Labour Force firm than in a large firm. Figure 110 (panel a) shows that Survey (QLFS) data, the share of employment in the SMME more than two-thirds of those working are in small firms, sector declined from 68 percent in 2005 to 62 percent in a trend that is more pronounced for new entrants. The 2016. Over 70 percent of SMME employees are functioning probability of finding employment in a small firm from in low- to medium-skill level occupations. The largest share being inactive or unemployed is more than three times that of them is in wholesale and retail (30 percent), followed by of finding employment in a large firm. Further, Figure 110 (c) suggests those of prime working age (between 24 and 38 SMMEs are defined as follows: Businesses made up of the entrepre- neur only and employing no workers, known as “Own-account”; 55 years old) newly entering the labor force are 10 percent businesses with 1–4 employees (excluding the owner) are “Micro”; businesses with 5–9 employees are “Small”; businesses with 10–49 more likely to enter small firms. employees are “Medium”; and businesses with 50 employees or more are “Large.” 90 Overcoming Poverty and Inequality in South Africa Figure 110: Employment probabilities, comparing small and large firms a. Frequency of employment b. Transition matrices 77% 68% Employed 33% 19% 31% 17% 3… Wave 3 Unemployed 47% 40% 10% 32% 1% 23% Inactive 77% 19% 4% wave 4 Status In Wave 4 Inactive Small (1- Larg e (50+ ) Small (1- Larg e (50+ ) Status In Wave 4 Un-employed 49) 49) Status In Wave 4 Employed Small Firm Status In Wave 4 Employed Large Firm Full Sample New Entrance c. Frequency of employment by age group d. Probability of finding employment, regression estimates 70% 66% 60% 56% 50% 33% 42% 40% 15% Small (1- 49) 9% 29% 8% 7% 30% Larg e (50+ ) 0% 1% 2% 20% Matric Matric Primary Primary Tertiary Tertiary Secondary Secondary 10% 4% 2% 1% 0% 0% 15- 24 24- 55 55- 64 65+ Large Small Source: Panel A: NIDS wave 4. Panel B: NIDS wave 3 and 4. Panel C: NIDS wave 4. Wages in larger firms are higher and an incremental than smaller firms. In small firms, the relative increase in increase in wages with an increase of skills or wages for a matriculate level of education from a base of education levels greater for larger than smaller firms. no education results equals 65 percent, while for a tertiary On average, wages in large firms are one and a half times level of education it equals 171 percent. Similar figures for more than that of small firms. For new entrants, this large large firms are 104 percent and 197 percent, respectively. firm premium drops a little to 1.45 times. Estimates suggest Similarly, for highly skilled jobs, the relative increase in wages in large firms are 19 percent greater than those in wages for small firms from a base of low-skilled jobs equals small firms. In addition, the incremental increase in wages 64 percent while for large firms it is 96 percent. as skills or education levels increase is greater for larger An Assessment of Drivers, Constraints and Opportunities 91 Figure 111: The large firm premiums a. Wages by firm size b. Mincer regressions showing the large firm premium 197% 171% 96% 104% 57% 53% 64% 65% 32% 36% 19% 23% 2% Matric Matric Tertiary Tertiary Skill Level 2 Skill Level 3 Skill Level 2 Skill Level 3 Primary Education Primary Education Secondary Education Secondary Education Large Firm Full Sample Small Firm Large Firm Source: Panel A: NIDS save 4. Panel B: NIDS wave 4. The analysis showed that SMMEs pay significantly less environment are often said to contribute to the high level than larger firms and their shares are falling. However, of unemployment and wage disparities. Among the factors SMMEs are very important for absorbing younger, less generally mentioned are the rigid labor market, the extent skilled, and less productive people. The general trajectory collective bargaining, the prevalence of labor brokering, to obtain a job in the formal sector goes through the initial and problems with the implementation of minimum employment in SMMEs. The NDP highlights the importance wages, which are set at a regional level. of these businesses for job creation, innovation, and Union membership is integral to the structure of the competitiveness. The unemployed are more likely to find South African economy. Unions played an important a job in small rather than large firms. At the same time, sociopolitical role in the movement toward democracy. wages in larger firms are prominently higher. In addition, For the better part of the twentieth century, black South the incremental increase in wages as skills or education African workers were disenfranchised and excluded from levels increase is greater for larger than smaller firms many jobs. Union membership, as allowed for under the This bifurcated market for SMMEs requires a nuanced Industrial Conciliation Act of 1910, was not extended to set of policy solutions for each component of the SMME black African workers until the amendment act of 1979 cohort to achieve a more inclusive and equal growth (Bhorat, Jacobs, and Yu 2013). With this history, trade unions agenda. Assistance to smaller firms, which are more likely in were inextricably political, acting as the voice of the to be in the informal sector and be survivalist, may primarily African working class in opposition to apartheid. In the late be viewed as part of a poverty reduction strategy. 1980s, African trade unions successfully managed to lobby for the creation of a national bargaining council (Godfrey, G. THE ROLE OF LABOR UNIONS IN WAGE Clark, and Theron 2005), which led the way toward more DETERMINATION centralized collective bargaining from the 1990s onward (Bhorat, Naidoo, and Yu 2014). As of 2016, there were 195 A debate in South Africa academic and policy research registered trade unions in South Africa (Department of is focused on the role of institutions on labor market Labour 2016). outcomes. Labor market institutions and a rigid regulatory 92 Overcoming Poverty and Inequality in South Africa The union density estimates for South Africa are not an council, while Bhorat, Goga and van der Westhuizen outlier when compared to other OECD countries. The (2012), estimates that the unionization premium outside average union density for OECD countries was 30 percent of bargaining councils is about 7 percent. The wage gains in 2013 while South Africa’s was 37 percent (Bhorat, Naidoo, from unionization are particularly large in the middle of and Yu 2014). Using a dataset from the 1990s, Botero et the wage distribution, and the level of the union premium al. (2004) shows that South Africa’s relative union power, is not excessive compared to other developing countries measured by a labor union power index, is much higher such as Brazil, Ghana, and Mexico. than the mean value of other countries. In turn, its protection Public sector union membership as a percentage of workers index, capturing how the country fares during of public sector workers increased between 1997 collective disputes, shows that South Africa falls below the and 2016. On the other hand, there has been a trend global average in all income classified country categories. of decreasing private sector union membership as a The authors argue that while South Africa exhibits a strong percentage of total workers in the private sector. These legal right to unionize, the levels of union power are not trends show possible segmentation between public and disproportionately high when measured by the collective private sector workers in the South African labor market. dispute index (Bhorat, Naidoo, and Yu 2014). Across the income distribution, unionized workers There is wide union coverage among employees, and earn more than non-unionized workers, with public the premiums associated with union membership are sector unionized workers earning the highest wages. substantial. There are close to 200 registered trade unions The impact of this trend of separation between the public in South Africa, covering around one-fifth of private sector and private sector union membership on the distribution workers, and two-thirds of public sector workers. Private of wages is captured in the distribution of wages by sector unionization has been trending downward since the sector and union status presented in Figure 112. This early 2000s, while public sector unionization has increased segmentation is cemented by the modes of the non- over the same period. Unions can negotiate substantial union wage distributions—these modes are significantly gains for their members within the bargaining council to the left of the modes of the unionized workers’ wage system. Bhorat et al. (2012) estimates a wage premium distributions. of 22 percent for unionized workers within a bargaining Figure 112: Trade union membership of formal sector Figure 113: Percentile distribution of log wages by union employees by public and private sector status, selected status and public/non-public sector status, 2014 years Source: Adapted from Bhorat, Naidoo, and Yu, 2014; 2016 Figures from Source: LMDS, Q4 2016, authors’ calculations. Quarterly Labour Force Survey, Q3 2016. An Assessment of Drivers, Constraints and Opportunities 93 The gap between public union wages and private non- sector unionized work and private sector non-unionized union wages is the largest toward the middle of the work is the largest, showing that, of the missing middle, distribution (Figure 113). At the bottom of the distribution, it is the private non-unionized workers who have lost out the minimum wage seems to be at work protecting the the most. At the lower percentiles, the ratio of public union earnings of workers irrespective of union status, while the wages to private non-union wages are the smallest, most skills premiums at the top of the distribution remunerate likely because the minimum wage protects the earnings of workers equally, irrespective of union status or sector. all workers at this end of the distribution. However, between the 20th and 80th percentiles of the Unions appear to restrict supply, but they offer wage distribution, a clear ranking of earnings is visible. substantially higher wages. Even controlling for other Unionized workers in the public sector earn the most, factors, wages for union jobs are 42–49 percent higher followed by unionized workers in the private sector. This than wages for non-union jobs (Figure 114). In addition, is followed by non-unionized public sector workers and the returns on union jobs, estimated by the Oaxaca-Blinder those who earn the least, private non-union workers. This is decomposition, rise over time. These results imply that further evidence of the hollowing out of the middle of the unions introduce some rigidity in the labor market. Firms distribution, suggesting that those who are not unionized may see such workers as too costly and thus job offers may and in the private sector have lost the most in the labor be restricted. For instance, the number of hours worked by market, and thus presenting a key channel through which non-union workers tend to be nearly the same as those in rising wage inequality has manifested in the domestic labor unions; yet the wages for union workers are much higher. market. In terms of the data estimates, it is at the middle of At the same time, Casale and Posel 2010 have argued that the distribution that the gap between the wage for public unions tend to provide more equitable wages. Figure 114: Union restrict supply but raise wages a. Wage elasticity of union b. Size of the union premium (base = non-union job) 0,0917 -0,0195 Union Non-Union Source: NIDS wave 4. Source: NIDS waves 1–4. Note: The coefficients are labor demand elasticities from panel regressions Note: Coefficient from Mincer regression that are premium to union suggesting percentage increase in labor supply to increase in wages. membership. Sectoral centralization of the collective bargaining the wage premiums between public sector unionized and instituted in South Africa is generally larger among private sector non-unionized workers have produced wage former employers resulting in greater incidence of gaps that are the largest in the middle of the distribution. fixed wages across sectors unrelated to the firm or Ultimately, those workers who have lost out the most on individual productivity level. Trends in unionization levels wage returns are not only in the middle of the income show the stark segmentation between public and private distribution, but generally work in the private sector, and sector union membership. This has had the strongest are non-unionized. impact on individuals in the middle of the distribution, as 94 Overcoming Poverty and Inequality in South Africa H. HIGH RESERVATION WAGES AND VERY types, giving it a dual structure. A small number of people HIGH WAGE DISPARITIES can access highly paid jobs while the majority works at less well-paying jobs. The highly paid jobs also are highly sticky: The wages of the unskilled and informally employed once found, people are unlikely to give them up. The less are extremely low. Figure 115 shows that the wages well-paying jobs are fluid by contrast, being more likely of the poor, those with low skills, and those employed in to employ new entrants into the labor market and more agriculture are very low compared to the average grants likely to witness exits from employment. As noted earlier, some households are receiving. This level of wage is race may affect the ability to find jobs, as well as the wages unattractively low. The data clearly indicate that wages for received once employed. The employment outcome is workers with scarce skills are too low compared to wages worse for females than for males, however the gender- for workers with a more abundant skill set. The reservation employment gap has been closing over time. wages are too high for many people to enter the current Figure 116 shows the estimated returns to various factors labor market. based on a standard log-linear wage equation. These South Africa has a highly unequal distribution of factors may be affecting wage levels and indirectly, wage wages. The labor market is polarized into two extreme job inequality, in the South African labor market. Figure 115: Average wages and transfers Figure 116: Returns from Mincer regression Source: NIDS wave 4. Source: Authors’ calculations from QLFS data 1995, 2000, and 2013. Selected variables presented. Dependent variable log of monthly wage. Independent variables include demographic, location, sector, and education variables. Wage gaps in race and gender are still prominent but percent in 1995. More recent estimates, with the data falling. The results indicate, in the first instance, that all caveat noted above, have seen this gender penalty decline else constant, older workers are likely to earn more than to about 20 percent. The mean racial wage gap has younger workers with approximately 6 percent increase declined from 65 percent for African workers to about 40 per year. Non-linearities in this age-earnings relationship percent in 2013. The results indicate that living in an urban are observed. Race and gender effects continue to predict area continues to afford wage premiums ranging from 16 earnings in the South African labor market. Hence, the to 20 percent over the 1995–2013 period. conditional mean gender wage gap stood at about 29 An Assessment of Drivers, Constraints and Opportunities 95 The results reinforce the pattern of skills-biased labor most cases this difference is decreasing. This supports the demand in the South African economy. Together, the econometric evidence showing a rise in mean farmworker education and occupation coefficients suggest that labor wages arising out of the minimum wage in the sector demand is, and has increasingly become, skills-intensive. (Bhorat, Kanbur, and Stanwix 2014). The results for 2010 Individuals with secondary education earn significantly suggest that the mining industry, followed by the public more than those with no schooling or only primary sector and transport sector—continue to offer the highest schooling, while those with post-secondary education earn sectoral mean wage premiums. a greater premium than those employed with some form of I. LABOR FACTORS AFFECTING TRANSI- secondary schooling. There is clearly a monotonic return to TIONS INTO AND OUT OF POVERTY— human capital across the entire 1995–2013 period. In 2013, RESULT OF PANEL ANALYSIS39 for example, a post-secondary educated worker earned on average about 116 percent more than an individual This section examines the factors that correlate with with no or only primary schooling, rising from 89 percent risks to poverty during 2008/9–2014/15. NIDS panel data in 1995. This return on tertiary education is consistently are used to analyze factors contributing to the transition growing. However, returns on semi-skilled or matriculation of households into and out of poverty. The results of the education is falling over time. Similar patterns are observed probability of falling into poverty are illustrated in Figure in terms of skills variables—returns to high-skills professions 117. are increasing. Demographic factors matter for the risk of falling into Returns to formality and unions are growing. As poverty. Female-headed households, black South Africans, expected, formality yields a higher average return, as does and youth have a higher risk of falling into poverty. For those possession of a formal written contract. The union wage living in initially non-poor households, the risk decreases premium stands at about 32 percent in 2013, above 1995, with the age of the household head. Members of female- although more detailed analytical work, with more careful headed households are up to 10 percent more likely to slip controls around bargaining council membership and trade into poverty and 2 percent less likely to escape poverty than union representation provides a union wage premium of members of households with male heads. Race remains a about 7 percent (Bhorat, Goga, and van der Westhuizen strong predictor of poverty, with black South Africans at 2012). Being unionized remains a key predictor for higher the highest risk of being poor. In comparison, white South conditional earnings across the entire distribution, relative Africans are about 25 percent less likely to fall into poverty to non-unionized workers in the private sector. and more than 50 percent less likely to remain poor, even The sectoral wage premium results confirm that all after controlling for other characteristics. industries pay significantly higher wages than the agricultural sector (the base category), although in 39 The analysis is based on the upper bound poverty line. 96 Overcoming Poverty and Inequality in South Africa Figure 117: Marginal effects for transitioning into poverty Source: Authors’ calculations. Compilation of the results from panel regressions. Note: The figure reports the average marginal effects of a probit regression with the individual poverty status at time t as the dependent variable. That is, the dependent variable is one if an individual is classified as poor at time t and zero otherwise. The explanatory variables include characteristics of the household that the individual lived in at time t-1. All explanatory variables were measured with a time lag (that is, prior to a potential poverty transition) and, in line with most of the poverty modeling literature, are assumed to be predetermined. The impact of having a working head on risk to falling sector is associated with a 2 percent higher chance of into poverty vulnerability depends on the type of exit out of poverty, those living in households where employment that the head engages in, especially the head runs a formal sector business (registered for regarding its stability and duration. income tax and/or VAT) face an 11 percent higher chance of making it out of poverty. Similarly, among • Persons living in a household where the head is the non-poor, self-employment of the household unemployed face a similar risk of poverty as those head in the informal or the formal sector is respectively with an economically inactive head or a head who associated with a 6 or 12 percent lower risk of poverty engages in subsistence farming. entry. • Those living in households where the head is casually • Persons living in a household where the head works as employed or helps others with a business are 3.8 an employee face a 3 percent lower risk of remaining percent more likely to remain poor than those with in poverty and 4 percent lower risk of transitioning inactive heads. More substantial is the difference into poverty. The effect is mainly driven by those who among the presently non-poor, where such an have a permanent work contract, which is associated unstable job position of the household head is with an about 5 percent lower vulnerability to associated with an 18 percent higher risk of falling poverty. Among the non-poor, the strongest effect is into poverty, making this an important vulnerability estimated for those where the head is a member of a factor. trade union, related to about an 8 percent lower risk • Self-employment of the household head can provide of slipping into poverty. This effect is likely explained an avenue out of poverty. However, while self- by higher wages and higher job security associated employment of the household head in the informal with union membership. An Assessment of Drivers, Constraints and Opportunities 97 • Persons living in a household with the head smaller for the initially poor than the non-poor. In addition employed in the services and especially in higher- to the explanations suggested earlier, it can be argued that skilled occupations, such as professionals, technicians, being poor can bring difficulties in finding good quality or clerical support workers, are considerably less jobs—through social network effects for example— vulnerable to poverty. This applies to jobs in electricity, reducing the probability of exiting poverty. gas, and water supply, as well as community, social, and Data suggest that urban and initially richer provinces personal services, where public sector employment had lower vulnerability to poverty. The risk of falling into tends to be an important contributor. In fact, there poverty is about 7 percent lower in urban than in traditional is a strong and significant relationship between the areas, whereas the chances to escape poverty are not average share of employment in the public sector significantly different between regions. Everything else and reduced poverty risks.40 In addition, mining being equal, mobility out of poverty and especially mobility sector jobs are associated with a 16 percent lower into poverty is highest in the Western Cape (although the chance of remaining poor and an 11 percent lower difference in not statistically significant for all provinces). risk of falling into poverty. By contrast, households Poverty persistence is highest in KwaZulu-Natal, followed with the head working in agriculture generally face a by the Eastern Cape. Here, both movements into and out of higher vulnerability to poverty. poverty are comparatively infrequent, which may indicate Higher levels of education of the household head are lower volatility, but may also be due to a more rigid social strong predictors for lower vulnerability to poverty. structure. Living in a household whose head has attained some tertiary education reduces the average risk to poverty by SUMMARY about 30 percent compared to those living in households with a head who has no schooling. The effect of primary The South African labor force is characterized by high and secondary schooling, by contrast, differs considerably levels of unemployment, low participation, and many between initially poor versus non-poor households. unemployed and discouraged work-seekers or non- Specifically, those living in households where the head seekers. The two decades following the end of apartheid has attained at least some secondary education are, on have yielded a growth path characterized by a rapid relative average, 4 percent less likely to remain poor, whereas the expansion in the services (or tertiary) sector. A simultaneous risk of falling into poverty is reduced by 17 percent. For shift to a more educated labor force led to an increasing those where the head has completed secondary schooling, share of semi-skilled and high-skilled jobs. Labor market the average poverty risk is reduced by 10 percent if initially productivity increased in sectors other than the financial poor and 26 percent if initially non-poor. Primary schooling services sector, which had growth in employment that was of the household head is associated with a 7 percent lower lower than growth in the sector. Skills intensity increased in average risk of falling into poverty compared to those in most sectors. with no schooling, whereas there is hardly any statistically Having an employed household head is not necessarily significant difference with respect to the likelihood to associated with a lower vulnerability to poverty. A large remain in poverty. proportion of the population consists of working poor who Presence of economically dependent household earn very low wages. The effect seems to depend on the members causes an elevated vulnerability to poverty. type of employment that the head engages in, especially The number of employed household members has an regarding its stability and duration. To unlock the full important effect on reducing vulnerability although it is potential of labor markets in accelerating the reduction of poverty and inequality, it is important to create jobs and 40 Public sector employment is not reported in NIDS. The sector level shares have been calculated from the Quarterly Labor Force Surveys increase wages at the same time. (QLFS) by sub-period (2008, 2010/11, 2012, 2014/15) and imputed to NIDS data. 98 Overcoming Poverty and Inequality in South Africa South Africa has a highly unequal distribution of categories. Once employed, education and skills result in wages and relatively high reservation wages. The labor substantial wage increases. Racial differences alter the market is polarized into two extreme job types, giving it a probability of finding employment for low-skilled and dual structure. A small number of people can access highly formal jobs. The dichotomy in finding employment can paid jobs while the majority work at less well-paying jobs. be explained by rising disparity within the black South The high-skill jobs are very sticky: once found, people are African group. Although an increased number of women unlikely to give them up. The less well-paying jobs are fluid participate in the economy, female participants have a and more likely to employ new entrants into the labor harder time finding a job and earn less than men when market and more likely to witness exits from employment. they do. People in urban areas have better job prospects One of the more distinct features of South Africa is its and higher probability of getting a formal job, but there apartheid legacy. Race may still affect the ability to find are no significant differences across provinces. High travel jobs, as well as the wages received once employed. The costs are a burden for getting jobs. employment outcome is worse for females than for males, There is stark segmentation between public and private though the gender-employment gap has been closing sector union membership in South Africa. This has had over time. the strongest impact on individuals in the middle of the A structural mismatch between labor demand distribution, as the difference in wage premiums between and labor supply for unskilled workers is strongly public sector unionized and private sector non-unionized evident. Education is important in transition to labor force workers have produced wage gaps that are largest in the participation, but less affiliated with finding employment. middle of the distribution. Ultimately, those workers who Low correlation between education and the probability have lost out the most in terms of wage returns are not of finding employment masks heterogeneity in the role only in the middle of the income distribution, but generally of education in finding jobs in different skills requirement work in the private sector and are non-unionized. An Assessment of Drivers, Constraints and Opportunities 99 CHAPTER 6 GOOD JOBS ARE THE KEY TO FUTURE REDUCTIONS IN POVERTY AND INEQUALITY Absent new policy interventions, the prospects for reduced poverty trends until 2030 under different scenarios. poverty and especially for reduced inequalities are very limited These projections were done using the dynamic World but would benefit from progress in access to education. Poverty Bank Computable General Equilibrium (CGE) model for rates (at the lower bound national poverty line) are projected South Africa, which includes a microsimulations module to decrease from 40 percent of the population in 2016 to 33 to measure the poverty impact of demographic variables percent in 2030, and inequality would fall, with a Gini coefficient (composition of the population by age and education), dropping from 62.8 in 2017 to 59.5 in 2030. Interventions that labor market variables (employment by sector, wages, simultaneously stimulate growth and reduce inequalities are and firm profits), and exogenous income variables (public likely to have much more impact than interventions that only transfers and taxes, private transfers).41 stimulate growth or reduce inequalities. Analysis of current Long-term policy impacts are measured by comparing policy interventions, such as the employment tax incentive and a baseline scenario to alternative policy scenarios. The the national minimum wage, suggests that their impact on baseline scenario is developed to project the economy inequality, and thus on poverty, is very modest. Creating good until 2030 in the absence of any major shock or radical jobs for the poor will have a much larger impact on inequality shift from the current policy stance. This scenario should and poverty. 41 The CGE model is described in detail in forthcoming SCD analysis. The model aims to provide a consistent framework to explore medi- A. PROJECTING POVERTY REDUCTION um-term developments, based on the main structural features of the THROUGH 2030 economy. The model is calibrated for the year 2012, based on a Social Accounting Matrix (SAM) built by Chitiga-Mabugu (2016). The SAM and the model cover 55 sectors of activity (and corresponding prod- The complex nexus between growth and inequality in ucts), 10 household types (corresponding to the 10 income deciles), 12 trading partners, and 7 factors of production: informal labor, formal South Africa is illustrated through the projections of unskilled, formal semi-skilled, formal skilled, capital, mineral resources, and water resources. 100 Overcoming Poverty and Inequality in South Africa not be considered a projection, but a possible future, from (0.3 percent and 0.8 percent, respectively), real GDP would which the impact of alternative policy stances can be grow at the annual average rate of 1.4 percent between evaluated. It does not either prejudge the political feasibility 2018 and 2030 (slightly above population growth, 1.1 of such a future, which can be considered uncertain given percent), generating 215,000 new jobs per year, two-thirds the persistent high level of inequalities, combined with of them among skilled and highly skilled workers. From 27.3 people’s high access to political and judicial instruments percent in 2017, the unemployment rate would go down to to redress them. The baseline scenario is influenced by 26.7 percent in 2030. The improvement in the employment several exogenous drivers, including world prices (slowly rate, that is, the proportion of the working age population rebounding mining prices), water scarcity and need to that is employed, would be more pronounced, from 42 contain carbon emissions (through taxation of carbon percent in 2017 to 43 percent in 2030, reflecting education content), and the changing composition of the labor progress discussed below. force (in terms of skills) with past and ongoing education Absent new policy interventions, the prospects for reduced efforts.42 Rebounding from the low levels of 2016 and 2017 poverty and inequality are very limited, but would benefit 42 The baseline scenario includes several assumptions. Population is from progress in access to education. set to grow at the annual average of 1.1 percent from 2018 to 2030 (from 57 to 65 million based on UN population projections). Keep- In the event, poverty (lower bound) would decline ing constant pass rates (matriculation and tertiary education) at their from 40 to 33 percent of the population despite low per 2016 levels, the supply of skilled and highly skilled labor is projected to grow faster (1.6 and 2.0 percent annually, respectively) than that capita income growth, as inequalities would narrow. of formal and informal unskilled labor (0.7 percent) between 2018 and 2030 (a total labor supply growth of 1.3 percent). Water supply The Gini coefficient would drop from 0.63 in 2016 to 59.5 is assumed to stay constant at its current level until 2030, as all pos- in 2030, and the share of real disposable income accruing sible water reserves are already being exploited. In contrast, mineral reserves (coal, gold, other mining) are considered infinite, and their to the poorest four deciles—the “bottom 40”—would depletion rate is being driven by world prices (using World Bank pro- increase from 8.6 to 10.3 percent (Table 10). jections, foreseeing a modest rebound in prices) versus production costs. Technological progress is (optimistically, given recent trends, and after accounting for the projected change in the skills mix, and factor reallocation), set to stagnate over the period 2018–30. Net for- eign financial flows are set to grow at 2 percent annually. But the pro- gressive introduction of a carbon tax, all direct and indirect tax rates (including import tariffs) are assumed to stay unchanged from 2017 in the baseline scenario. Public consumption and public transfers (social assistance) to households are assumed to stay constant in real per capita terms over the period 2018–30. Table 10: Projected poverty and inequality rates - baseline scenario Food Lower bound Upper bound $1.9 a day Gini coefficient 2017 24.7 39.8 55.5 18.6 62.8 2030 18.8 32.7 51.3 12.7 59.5 Change 2030–2017 -5.9 -7.1 -4.2 -5.9 -3.3 Source: World Bank staff calculations. As public transfers to the poor are assumed to to a redistribution of skills (and related labor incomes) remain constant in per capita terms, most of the across deciles. At current pass rates, and accounting for projected reduction in inequalities can be attributed the slow renewal of generations and the long time it takes to a reduction in inequalities of education. Analysis of for youth to enter labor markets, the proportion of semi- enrollment and attainment across deciles suggests that skilled labor (matriculation level) incomes accruing to the progress in education among the poorest deciles could bottom 40 percent would rise from 4.5 percent in 2012 be faster than among richer ones, contributing over time to 11.2 percent in 2030 (while 23 percent of the students An Assessment of Drivers, Constraints and Opportunities 101 eventually matriculating currently originates from the B. POLICY INTERVENTIONS TO GAIN bottom 40 percent); likewise, the proportion of highly FURTHER POVERTY AND INEQUALITY skilled labor (university degree level) incomes accruing REDUCTION to the bottom 40 percent would rise from 0.5 percent in 2012 to 3.6 percent in 2030 (while 11 percent of the cohort Employment and labor earning is a strong avenue out eventually getting a degree currently originates from the of poverty. The importance of the labor market in lifting bottom 40 percent). Such progress is consistent with the a household out of poverty can be seen when examining observation of a reduction in inequalities of opportunity in the drivers of escaping poverty (Figure 118). Movement the last decade, which is eventually affecting labor markets out of poverty is more likely to take place if the share of with a delay. employment income in total income increases; finding a job has nearly as strong an effect. A change in job skill levels also increases the chance of movement by a relatively smaller amount, while an increase in the share of children in a household lowers the probability of escaping poverty. Figure 118: Moving out of poverty: contributing factors 21% 19% 8% -11% Increase Work Find Employment Change Job Skill Change Share of Income Share Level Children Source: Authors’ calculations based on NIDS 2014/15 data. Labor market participation is important to reduce the impact of hiring people out of unemployment on poverty, but as the labor chapter of this report the total economy and for various economic sectors. recognized, the lack in aggregate demand is On average, moving 10 people from unemployment to complimented by the supply side deficiency. South employment reduces poverty for 7 people, but the effect Africa has a skills mismatch and a structural unemployment varies by sector. Thus, adding 10 workers in mining and problem; many workers do not possess the skills employers agriculture will reduce poverty for 13 people (the effect is demand. Demand is low for low- and semi-skilled greater than 10 as wages affect not just workers, but also workers, while high demand for high-skilled workers led their households). Increasing employment in construction to tremendous wage polarization and the emergence and manufacturing sectors also significantly affects of a missing middle. This is associated with the very low poverty, though in these sectors the exchange is almost earnings for less skilled informal workers making scarcely 1 to 1. Getting people into trade, financial services, and available low-skill jobs unattractive. A large proportion of community services has a smaller impact. Employment the population consists of the working poor who earn very in financial intermediation is geared toward the relatively low wages. Improvement of the lives of the poor could better-off educated population, so the impact on poverty be achieved through creating jobs and providing better is smaller than for other sectors. In some sectors, such as earning opportunities through developing skills and raising employees of private households, the impact on poverty is labor productivity. small because of low wages paid in these sectors and the Generating employment will reduce poverty. Figure 119 impact of the loss in transfers on poverty is significant. shows the results of a microsimulation exercise assessing 102 Overcoming Poverty and Inequality in South Africa Figure 119: Change in poverty due to employment Figure 120: Change in the Gini coefficient due to generation employment generation 0,4 0,2 Percentage change in poverty 0,0 -0,2 -0,4 -0,6 -0,8 -0,7 -0,7 -0,6 -1,0 -0,7 -0,7 -1,2 -0,9 -0,8 -1,4 -1,1 -1,0 -1,6 -1,3 -1,3 le so ring tri str ure . ed y r ic g W ty p fact r e an To nic s m M a as w n n l in Ec on an riv ret serv u ate rm om d t c hou rad Ag ryin y g tio tio m hol cia tal ati El Co ult u cit uc ia te on t l ar na om se sp ate ail qu u n un n ho er d e an l sa g ec in i or in P M m Fin Co Tr employment effect grants/UIF effect Source: Estimations based on NIDS 2014/15 data. Low bound poverty line Source: Authors’ calculation based on NIDS 2014/15 data. adjusted for inflation is used. Note: The figure shows two effects on poverty that work in opposite directions: potential reduction in poverty due to the increase in employed income (above the zero line), and potential increase in poverty due to the loss of unemployment insurance or means-tested child support grants (below the zero line). The analysis was done first for the whole economy, disregarding sectoral affiliation, and second by targeting employment to each sector. The impact of job creation on inequality depends An increase in wages for the working poor has positive, on the magnitude of the increase in employment but relatively small impact on extreme poverty as and sector affiliation of the employment growth. As employment income is not the main source of income presented in Figure 120, an increase in employment by for the poor. Figure 122 shows the poverty reduction 500 thousand individuals will reduce the Gini index by associated with a 10 percent increase in sectoral wages. 0.6 percent. The range of the Gini index reduction is 0.4 to A 10 percent increase in wages will, on average, decrease 0.7 percent, depending on the sector where employment poverty by 3.7 percent. The impact is stronger for the wage is growing. The increase in employment in the sectors beneficiaries, where a direct impact of 10 percent increase with higher wages has more pronounced results on the in wages results in 7.3 percent poverty reduction. The reduction of income inequality. The highest impact on the impact of wages varies by sector. The strongest impact on Gini index is due to the increase in employment in financial poverty is observed due the increase in manufacturing and intermediation, mining, transport, and electricity sectors. A trade sectors. smaller impact on inequality is associated with an increase in employment in community services, agriculture, and private households. An Assessment of Drivers, Constraints and Opportunities 103 Figure 121: Changes in simulated poverty rates due to Figure 122: Percent reduction in poverty rates increase in total wages, all economy and beneficiaries following 10 percent wages growth Simulated wag es g rowth in percent 0 Private households 0.4 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 -5 Ag riculture 0.4 Percentag e chang e in the poverty rate -10 Mining and quarrying 0.3 -15 Manufacturing 1.1 -20 Electricity g as water 0.0 -25 Construction 0.2 -30 Wholesale retail trade 0.7 Transport communication 0.2 -35 Financial intermediation 0.1 -40 Community personal serv. 0.5 -45 -50 0.0 2.0 4.0 Percentag e chang e in the poverty rate Beneficiaries Total Economy Source: Simulations are based on NIDS 2014/15 data. LBPL adjusted for inflation is used for the analysis. In the above simulations, employment elasticity is not assumed, thus the increase in wages is not associated with decrease of employment. Figure 122 suggests a poverty reduction of 1.27 percentage points, which is 3.8 percent of the lower bound poverty rate. Both employment generation and wages for the poor quality of jobs needed to reduce poverty and inequality. are important for reducing poverty and inequality. Thus, the access of the poor to skilled jobs needs to be Raising labor demand will ultimately be the driver of rapid accelerated to improve inequality, in raising labor demand reductions in poverty and inequality. As in the baseline through structural reforms, and in preparing the labor force scenario, the ability of poor South Africans to eventually to meet the new needs of the South African economy, as get skilled jobs is the most promising avenue to reduce its comparative advantages evolve over time. The following poverty and inequality. However, long-term economic sections in this chapter explore a few policy options in this growth prospects are grim and projected labor demand regard. is unlikely to be high enough to create the quantity and 104 Overcoming Poverty and Inequality in South Africa C. DISTRIBUTIONAL IMPACT OF LABOR This section analyzes the current set of labor policies and MARKET POLICIES AND LEGAL their projected impact on wages and poverty and then INSTITUTIONAL CHANGES IN RECENT focuses on interventions that would potentially have YEARS stronger impact on poverty. Box 11 summarizes the aims of the various amendments to the Labour Relations Act (LRA). Box 11: Policy, legal, and institutional changes The purpose of the Labour Relations Act of 1995 was to advance economic development, social justice, labor peace, and the democratization of the workplace by complying with labor standards set by the ILO (Oosthuizen et al. 2016). The act provides a framework within which employees and their employers can bargain collectively on wages and terms of employment and that supports the formulation of industrial policy. Over time, amendments have been passed to allow for the organizational rights of trade unions, the provision of pension and medical scheme coverage of employees, and the power of bargaining councils to provide industrial support. Table 11: Amendments to the Labour Relations Act 1996 1998 2000 2002 2014 To facilitate and Provisions for Specified the laws To enhance the To provide greater protection regulate the pension and around bargaining enforcement of for workers placed by organizational medical schemes. council registration, collective bargaining temporary employment rights of trade extension agreements. services by: To adjust the unions. agreements, and requirements Extended services Regulating the employment council agents. To promote and for extending and functions of of fixed-term contracts facilitate collective any collective Gave bargaining bargaining councils and earnings of part-time bargaining. agreements councils the power to the informal sector. employees below the concluded in a to provide industrial earnings threshold; bargaining council support services Specifying the liability for to non-parties. to participating employer obligations; parties. Limiting temporary employment to work that does not exceed six months. Source: Oosthuizen et al. 2017. i. The Labour Relations Amendment Act of 2014, distinguishing factor of TES arrangements is that the labor brokering, temporary employment services firm that receives the service does not directly hire the individual providing that service. The services provided The LRA Amendment Act 6 of 2014 provided greater by TES employees range in skill level, but as noted, TES protection for workers in temporary employment employees are usually more vulnerable, consisting of either services (TES). This amendment was introduced due to youth, or individuals from households close to the national growth in the number of TES workers being employed, poverty line (Bhorat, Cassim, and Yu 2014). as well as the prevailing view that working conditions for these workers were worse than for permanently employed The LRA Amendment Act specified that workers who individuals. Under TES employment third-party companies earned less than an annual threshold were deemed provide workers to fill various jobs in formal sector firms. permanent employees after three continuous months In South Africa, these are called labor brokering services. of employment. As a result, the amendment made it The occupations they fill include cleaning, accounting, illegal to employ temporary staff for a duration of longer secretarial services, security services, and others. The than three months. The amendment also states that all An Assessment of Drivers, Constraints and Opportunities 105 temporarily employed persons must receive the same in 2013 and was to last two years, from January 2014 to wage and non-wage benefits as permanently employed December 2016. The rationale of the policy was to offset persons. the costs of hiring young, typically inexperienced workers in a country where education is not always a reliable The employment growth in TES has exceeded the national employment growth rate of most sectors, indicator of job readiness (National Treasury 2016). The including the financial sector. TES employment, as a policy consists of a tax incentive to firms to stimulate youth proportion of financial employment, increased from 27 employment.43 percent in 1996 to 47 percent in 2014 (Bhorat et al. 2015), The ETI is currently the only demand-side incentive the and as a proportion of total employment it went from 2.2 government employs to absorb excess labor supply. percent to 6.44 percent in the same period. The TES sector Between the introduction of the incentive to the end of has been instrumental in maintaining, and arguably raising, 2015, over R2.26 billion in tax incentives were claimed by employment levels. In its attempt to protect vulnerable firms, supporting a total of 686,402 jobs, which equates workers, the conditions presented by the LRA Amendment to 5 percent of total jobs on the labor market. In general, Act of 2014 may thus have had adverse effects on the workers supported by ETI were not highly experienced pattern of employment levels in the TES sector. The extent and 57 percent of them were not registered for tax before to which firms are compliant with basic employment acquiring their job at the ETI-claiming firm. On the caution condition legislation, such as paying unemployment side, the natural job turnover rate for youth in the South insurance, is an important determinant of the way TES African labor market is high. The ETI does not require a new workers are treated (Bhorat, Cassim, and Magadla 2015). job to be created, it only requires a position to be filled by a Nonetheless, the LRA Amendment Act, which is an attempt young person. The natural turnover rate of jobs is sufficient at creating permanent employment, is targeted at all firms to generate enough positions (without creating additional irrespective of compliance with legislation. jobs) to exhaust the budget of the ETI. The unintended consequences of this amendment may Bhorat and Thornton (2016) show that the ETI had be an increase in labor shedding as firms try to shirk the differing impacts across sectors. Figure 123 shows the responsibility of having to permanently employ more eligible and supported jobs by sector, with the highest workers. The impact of this amendment was evaluated by numbers of potential or eligible workers in the financial and Bhorat, Magadla, and Steenkamp (2015) using data from business services sectors, wholesale and retail trade, and a survey conducted by the confederation of associations in the private employment sector. Using data from the manufacturing. Actual uptake of the incentive was highest post-legislation period, the authors show that the LRA in the sectors with high eligibility—first financial and Amendment Act had the effect of reducing jobs across the business services, followed by retail and wholesale trade, TES industry, notwithstanding the effects of external shocks then agriculture and manufacturing. The highest number to each of the industries (Bhorat, Magadla, and Steenkamp of claiming firms came from manufacturing, followed by 2015). The authors show that the dominant firm response to financial and business services. The uptake rate was highest the LRA Amendment was to terminate employment, with in tourism, with a rate of 26 percent of firms. a very small proportion of total jobs ending in permanent employment. The negative effects were largest in the metal 43 Firms are meant to pay less income tax per eligible employee be- and engineering, public, manufacturing, healthcare, white tween ages 18 and 29 who was hired after October 1, 2013, and earns less than R6,500 per month. Firms have 24 months (or until December collar, and education industries. 31, 2016) to claim a rebate for these workers, by which time, the work- ers are expected to have accrued enough experience to either keep ii. The Employment Tax Incentive their current job or qualify for a new one. The incentive is structured so that for the first year the full tax rebate is due to the employer, and in the second year of employment the rebate halves. The incentive The Employment Tax Incentive (ETI) is a demand-side is designed to discourage a “race to the bottom” whereby employers policy intended to counter the structurally high youth stand to benefit by paying lower wages to prospective candidates. To this end the size of the incentive is designed to rise then fall as unemployment rate. The policy was signed into law monthly wages increase. 106 Overcoming Poverty and Inequality in South Africa Figure 123: ETI eligible and supported jobs by sector Source: Bhorat and Thornton 2016. Displacement of older workers and wage depression to agreement, workers will receive a minimum of R20 per are the main concerns for the efficiency of the hour, which translates into a monthly wage of about R3,500 program. Econometric evidence by MacLeod and Rankin for a 40-hour week, and about R3,900 for those who work (2016) found a drop in the growth of full-time equivalent 45 hours a week. This section analyzes short- and long-term jobs for workers age 30–35 for the firms that claimed implications of this agreement. ETI, but the absolute number of this drop was small. Minimum wages in South Africa are covered by the LRA This kind of displacement is an adverse effect of the tax and the Basic Conditions of Employment Act (BCEA). The incentive, as employers substitute younger subsidized LRA guarantees the right to collective bargaining and is labor for older workers. Aside from displacement, another negotiated between unions and employers. The Minister of concern regarding the ETI would be that wages would be Labor can extend wage agreements to cover all employers depressed, or destructive churn would be created around and workers in a sector, regardless of whether those workers firms shuffling employees to maximize benefits obtained are part of the relevant bargaining council (see Box 12). from the incentive. There are currently 47 bargaining councils, of which 38 are iii. Expected poverty impact of national minimum private, 6 are public, and 3 are statutory. The BCEA outlines wage legislation the work conditions for all employees in the country, as well as the process for the sectoral determination (SD) of wages. In February 2017, representatives of government, business, The SD mechanism is aimed at vulnerable workers, and at the community sector and two of the three labor federations sectors that are not represented by workers’ organizations. signed the national minimum wage agreement. According South Africa has 11 SDs, with over 120 different wage rates. An Assessment of Drivers, Constraints and Opportunities 107 Box 12: Application national minimum wage A national minimum wage (NMW) will be applied to all sectors of the economy from May 2018. The value of the NMW has been set at R3,500 per month, or R20 per hour (equivalent to R2,976 in 2015 rand). Exceptions have been made for various sectors, with the agriculture and domestic service minimum set at 90 percent and 75 percent of the NMW, respectively. The NMW uses the definition of economic vulnerability set out in the BCEA to determine the initial subsample of workers to whom the NMW could apply. That is, the BCEA sets an income threshold below which workers are considered economically vulnerable, in the sense that their bargaining power is compromised. Figure 124 presents the typology of workers below this threshold in 2014. Of the economically vulnerable in the labor force, more than half are covered by a sectoral determination (SD), 10 percent belong to a private trade union, 8 percent to a private bargaining council, and 14 percent to the public sector, and 22 percent are uncovered. Approximately 40 percent of full-time workers (at least 35 hours per week) would be covered by a NMW of R2,976 in 2015 rand, but the coverage varies significantly by sector. The two sectors in which more than half of workers earn less than the proposed NMW are domestic services and agriculture, where 87 percent and 82 percent of workers earn less than R2,976. Wages in the construction and trade sectors are very similar, with just over 40 percent of workers earning below R3,000 a month. The percentages shown here do not indicate the extent of the distance that workers are below each line. For example, while the percentages affected at various levels in agriculture and domestic services are similar, the extent to which they impact will vary, as 50 percent of full-time workers in agriculture earn below R2,253 per month, compared to the 50 percent who earn below R1,577 per month in domestic services. Figure 124: Earning bands by sector (2015 rand) Figure 125: Ratio of NMW to lowest and highest SD wages 1,4 1,2 1,0 Ratio 0,8 0,6 0,4 0,2 0,0 il Co Hos axi re W y rs ct ality an ity rs ta tr W ivat rke tu ne T r es Re sa ecu ul t ea or o pi d ric Cl F S Ag tic e le es ra nt m Pr le Do ho NMW to lowest SD wage NMW to highest SD wage Source: Finn 2015. Source: Adapted from Bhorat et al. 2016. The relationship between the NMW and the current SD minimums differ widely by sector. The figure above presents the ratio of the NMW to the lowest and highest legislated minimum wages for eight sectors. The complexity of the current SD regime means that there are some large within-sector differences in minimum wages. For example, the lowest minimum wage in the taxi industry is R2,113 per month, while the highest legislated minimum is R3,021 per month. In contrast, the agriculture and forestry sectors have no within-sector minimum wage variation. The private security sector shows the 108 Overcoming Poverty and Inequality in South Africa biggest differences between existing minimums and the NMW. The NMW is over 40 percent higher than the current lowest minimum wage in private security but is less than half of the highest minimum in that sector. The overall impact will, of course, depend on the within-sector distribution of wages. The ratios for the lowest and highest SDs for contract cleaners are much closer, with the NMW being just higher than the lowest SD minimum, and just lower than the highest SD minimum in that sector. Lack of compliance by employers drives a wedge between wages that are legislated and wages that workers receive. Given the complexity of the existing SD wage setting mechanism, it is possible that the simplicity of an across-the- board NMW will have positive implications for compliance. A microsimulation was used to assess the first order a higher level of employment adjustment. In addition effect of the distributional impact of the proposed to adjustment of total employment, the adjustment in minimum wage on poverty and inequality. A modified individual hours worked was also simulated. The results version of the Bhorat et al. (2016)44 methodology was obtained from adjustment in employment and changing followed by evaluating the impact of the minimum wage work hours were not significantly different. Counterfactual on sectoral wages. Sector-level increases in wages are wage, employment, and total household income were assumed as a difference between current and proposed estimated based on the proposed methodology using legislation. Three wage elasticities were used to generate a NIDS 2014/15 data. To understand the extent to which the set of employment effects for these NMW scenarios: 0.1— minimum wage has the potential to affect the distribution low level elasticity suggesting maximum impact on incomes of wage inequality in South Africa, the income Gini and minimal impact of employment, -0.3—moderate level coefficients, poverty rates, and growth in incomes were of elasticity, and -0.5—a relatively high elasticity suggesting calculated. 44 Job losses of those who were employed at the time the survey was conducted were derived using a probability distribution of those most likely to lose their jobs. The probability distribution was estimat- ed using a two-step Heckman model of employment equation, con- sidering sample selection bias of those who will keep their jobs, based on five characteristics: race, gender, education, location, and age. This probability was then appended to the “wage gap”—the “distance” be- tween an employee’s current wage and the new legislated wage—as a weight, and thus jointly determined a ranking or queue of those individuals most likely to lose their jobs following a minimum wage introduction. This was then used to estimate the impact on house- hold inequality for households with at least one wage earner, then for all households including those with no wage earner. An Assessment of Drivers, Constraints and Opportunities 109 Figure 126: First order effect: impact of projected Figure 127: First order effect: impact of projected minimum wage legislation on poverty and inequality minimum wage legislation on income, by decile Source: Authors’ calculations based on NIDS 2014/15 data. Implementation of the NMW would have uncertain, This stickiness of the Gini coefficient points to a larger and at best relatively marginal impact on poverty problem with addressing the extent of inequality. While and inequality. As expected, the lower the impact of the the NMW has the potential to positively affect many low- minimum wage on job shedding, the higher its impact on wage earners and employed households, the impact that poverty and inequality reduction: a small labor demand the NMW has on the broader inequality of the population to wage elasticity (-0.1) would generate a 1.2 percentage becomes negligible. Tackling inequalities calls for solutions point decline in the Gini coefficient, and a 3.5 percentage that would increase the participation of the poor in a more point decline in the lower bound poverty rate. At the rapidly growing economy—that is, promoting inclusive other extreme, a large labor demand to wage elasticity growth in a meaningful way. (-0.5) would generate a 0.7 percentage point decline in the Gini coefficient, and a 2.0 percentage point decline in D. FUTURE POLICY MEASURES THAT COULD the lower bound poverty rate. These mechanical estimates HELP REDUCE POVERTY AND INEQUALITY remain subject to a number of uncertainties, as many Authorities acknowledge the need to accelerate other second order effects could come to play, including growth to address poverty and inequality. Recognizing imperfect enforcement of the minimum wage and greater the need to accelerate GDP growth from a low potential, resort to informality, impact on workers’ additional level authorities underlined in the Budget Review 2018 put of effort with higher wages, impact on the price of goods before the Parliament in February the need to undertake disproportionally consumed by the poor, agricultural structural reforms to forge a new compact between the goods notably, impact on the wage of unskilled labor social partners and provide investors with the certainty whose remuneration is already above the minimum wage, required that would encourage increased investment. possible shift in labor demand toward skilled labor, and Raising the level of investment and improving the ease of deepened capital intensity at the expense of unskilled doing business in the country will support job creation. labor. The government aims to finalize many outstanding policy 110 Overcoming Poverty and Inequality in South Africa and administrative reforms in sectors with high growth (Figure 130). The World Bank CGE model reflects this growth potential. The government envisions mining sector acceleration though higher productivity, domestic savings, policies that support investment and transformation, and investment, and measures its impact on jobs, poverty, telecommunication reforms, lowering barriers of entry and inequality by 2030, in comparison to the baseline and anticompetitive practices, supporting agriculture scenario discussed earlier. The results of this simulation and tourism sectors, and increasing skill levels across (TFP1) are presented in the column 3 of Figure 131. Such the country. The National Treasury estimated that, if the higher growth results having significant impact on poverty, international environment remains supportive, effective but not on inequality. As reflected in the table, the low implementation of the reforms could boost economic bound poverty rate would be 23 percent in 2030 (column growth in the coming decades (Figure 130) 3, TFP-2030) in comparison to 33 percent in the baseline scenario (column 2, BS1-2030). Inequality, however, will The World Bank CGE model was used to assess the remain at the same level. The scenario also suggests slight effect of higher growth on poverty and inequality in improvement in labor indicators—a fall of unemployment the medium- to long-term. As suggested in the Budget to about 24 percent in comparison to the 27 percent in the Review, improvement in confidence, telecommunications baseline and some improvement in the employment rates. reforms, the reduction of barriers to entry, transport This is because of the currently weak labor supply response reforms, and support to tourism and agriculture would to new economic opportunities discussed in previous encourage investment and raise productivity to eventually chapters, including the skills mismatch. raise GDP growth potential by about 2 percentage points An Assessment of Drivers, Constraints and Opportunities 111 Box 13: Growth to poverty elasticity in South Africa The poverty reduction response to growth differs substantially across countries. The percent of poverty reduction due to average growth is measured by the growth to poverty elasticity. Growth to poverty elasticity is the percentage reduction in poverty rates associated with a percentage change in mean per capita consumption. Generally, increases in per capita consumption decrease the poverty rate, hence the elasticity is negative. Growth to poverty elasticity ranges from -1 to -6 in developing countries, with a median estimate of around -3. Thus, on average in developing countries, a 1 percent increase in per capita consumption is associated with a 3-percentage point decrease in the poverty rate. Several factors affect the growth to poverty elasticity, the most important of which are the initial income distribution and the poverty line. Generally, countries with a more equal distribution of income have a higher elasticity and thus greater reduction in the poverty rate for a given increase in per capita consumption. South Africa has very low growth to poverty elasticities due to the extremely high levels of income inequality. The country’s growth to poverty elasticity in 2014/15 was -1.22 for FPL, -0.58 for the LBPL, and -0.97 for upper bound poverty (Figure 128). The growth to poverty elasticity in rural areas is the lowest, ranging from -0.33 to -0.71, depending on the choice of the poverty line. A relatively high proportion of the population lives far below the poverty line, and economic growth leads to relatively slow poverty reduction. Figure 128: Elasticity of poverty to consumption growth, Figure 129: Elasticity of poverty to consumption 2014/15 growth, 2005–15 Source: Authors’ calculations based on IES and LCS survey 2005–2015. South Africa’s low growth to poverty elasticities underline the critical importance of reducing inequality by developing social and economic policies that foster pro-poor growth. South Africa’s growth to poverty elasticities are lower than in most of the middle-income countries worldwide, but comparable to that of other highly unequal African countries. For instance, the elasticity in the relatively equal Mauritius is -3.2, while elasticity in Botswana is -1 and in Namibia -2. Over 2005–2015 growth led to a reduction in poverty (Figure 129), but it remained insufficient to make a significant dent in poverty, given the high inequality levels. Thus, future interventions that stimulate growth and reduce inequalities are likely to be much more effective than interventions that only stimulate growth or reduce inequalities. 112 Overcoming Poverty and Inequality in South Africa Coupling growth-acceleration reforms with efforts Figure 130: Potential impact of selected NDP reforms to narrow the skills gap would generate synergetic on GDP growth effects, and help South Africa attain the goals articulated in the NDP. Higher growth should provide the fiscal space to generate more job opportunities for the poor through education and provide a dignified life to those unable to reap growth opportunities through more generous social assistance. TFP2 simulation adds to TFP1 improved basic education and financial support to access university for the bottom 40 percent,45 and increased social assistance. Accelerated efforts to improve the quality of basic education and access to tertiary education would be rewarded by a significant reduction of inequalities by 2030—with a Gini down to 58, significantly amplifying the poverty-reducing impact of accelerated growth. As with strengthening the social compact through reduced inequalities, combined efforts would also likely improve the confidence of investors. Hence, costs to narrow the skills gap (of about 1 percent of GDP by 2030, comparing TFP1 Source: National Treasury, February 2018. and TFP2) could be partially offset by higher growth. 45 Two broad policy sets can be envisaged to improve the skills of youth from poor backgrounds: improving teachers’ capacity and accounta- bility to raise primary and secondary school achievements among the poorest deciles; and facilitating access to university for poor eligible students through financial support. Figure 131: Projected impact of the policies on poverty and social indicators 160.0 2017 BS1-2030 140.0 120.0 100.0 TFP1-2030 TFP2-2030 80.0 60.0 40.0 20.0 0.0 Bottom 40 share of real Real GDP index Unemployment Employment Food Lower Bound Upper Bound US$1.9 day Gini Coefficient disposable (2016= 100) rate rate income 2017 24.7 39.8 55.5 18.6 62.8 8.6 100.0 27.3 42.1 BS1-2030 18.8 32.7 51.3 12.7 59.5 10.3 119.5 26.7 43.1 TFP1-2030 12.9 22.6 40.8 7.4 59.4 10.2 149.8 23.5 45.0 TFP2-2030 8.9 19.9 38.7 5.7 58.0 11.1 149.0 23.8 44.9 Source: Authors’ calculations. An Assessment of Drivers, Constraints and Opportunities 113 SUMMARY Future interventions that simultaneously stimulate growth and reduce inequalities are likely to be much Poverty reduction prospects by 2030 will depend on more effective than interventions that only stimulate GDP growth and the reduction of income inequalities, growth or reduce inequalities. Analysis of current policy the former being affected by access of the poorest interventions, such as the ETI and the NMW, suggests that groups to economic opportunities, and fiscal their impact on inequality, and thus on poverty, is very redistribution. South Africa has slow growth to poverty modest. Creating good jobs for the poor will have a much elasticities due to the extremely high level of income larger effect on inequality and poverty. The social impact inequality. Projected sluggish growth, coupled with of reforms currently envisaged by authorities to boost recorded improvements in access of the poor to education growth would be significantly amplified with reforms to (and eventually, skilled jobs) is likely to somewhat reduce equip the poor to reap growth opportunities through the inequality and poverty in the coming years (baseline acquisition of skills. In doing so, the social compact would scenario). Poverty rates (at the lower bound national be further strengthened, with a likely positive impact on poverty line) are projected to decrease from 40 percent investment. Nonetheless, recognizing the time needed to of the population in 2016 to 33 percent in 2030 despite increase the economic participation of the poor over future slow growth, as inequality would fall, with a Gini coefficient generations, such a package of reforms would still need to dropping from 62.8 in 2017 to 59.5 in 2030. pay attention to maintain social assistance to the poor and vulnerable. Higher fiscal revenue from accelerated growth would provide the fiscal space to do so. 114 Overcoming Poverty and Inequality in South Africa REFERENCES Suggestions.’’ Development Policy Research Unit Working Paper 09/139. DPRU, University of Cape Town. AfDB (African Development Bank) (2011). The middle of Bhorat, H. and Goga, S. (2013). ‘’The Gender Wage Gap in the pyramid: Dynamics of the middle class in Africa. AfDB Post-Apartheid South Africa: A Re-examination.’’ Journal of Market Brief, 20 April 2011. Tunis: African Development African Economies 22(5): 827-848. Bank. Bhorat, H., and Mayet, N. (2012). ‘’Employment outcomes Alderman, H., Babita, M., Demombynes, G., Makhatha, N., and returns to earnings in post-apartheid South Africa.’’ and Ozler, B. (2002). ‘’How Small Can You Go? Combining Development Policy Research Unit Working Paper 12/152. Census and Survey Data for Mapping Poverty in South DPRU, University of Cape Town. Africa.’’ Journal of African Economies 11: 169–200. Bhorat, H., S. Goga, and van der Westhuizen C. (2012). Alkire, S., and Foster, J. (2011). ‘’Counting and ‘’Institutional Wage Effects: Revisiting Union and Bargaining multidimensional poverty measurement.’’ Journal of Public Council Wage Premia in South Africa.’’ South African Journal Economics 95(7): 476-487. of Economics 80(3): 400–414. Ayyagari, M., Beck, T., and Demirguc-Kunt, A. (2007). ‘’Small Bhorat, H., Kanbur, R., and Mayet, N. (2013). ‘’The impact and medium enterprises across the globe.’’ Small Business of sectoral minimum wage laws on employment, wages, Economics 29(4): 415–434. and hours of work in South Africa.’’ IZA Journal of Labor & Development 2(1): 1-27. Azevedo, J.P., Inchauste, G., and Sanfelice, V. (2013a). ‘’Decomposing the Recent Inequality Decline in Latin Bhorat, H., Kanbur, R., and Stanwix, B. (2014). ‘’Estimating America.’’ World Bank Policy Research Working Paper No. the Impact of Minimum Wages on Employment, Wages 6715. Washington: World Bank. and Non-Wage Benefits: The Case of Agriculture in South Africa.’’ American Journal of Agricultural Economics 96(5): Azevedo, J.P., Inchauste, G., Olivieri, S., Saavedra, J., and 1402–1419. Winkler, H. (2013b). ‘’Is Labor Income Responsible for Poverty Reduction? A Decomposition Approach.’’ World Bhorat, H., Cassim, A., and Hirsch, A. (2014). ‘’Policy co- Bank Policy Research Working Paper No. 6414. Washington: ordination and growth traps in a middle-income country World Bank. setting: The case of South Africa.’’ UNU-WIDER Working Paper No. 2014/155. Bane, M., and Ellwood, D. (1986). ‘’Slipping into and out of poverty: the dynamics of spells.’’ Journal of Human Bhorat, H., Naidoo, K., and Yu, D. (2014). ‘’Trade Unions Resources 21(1): 1–23. in an Emerging Economy: The case of South Africa.’’ Development Policy Research Unit Working Paper 201402. Becker, G.S., and Tomes, N. (1986). ‘’Human Capital and the DPRU, University of Cape Town. Rise and Fall of Families.’’ Journal of Labor Economics 4(3, Part 2): S1–S39. Bhorat, H., Naidoo, K., Oosthuizen, M., and Pillay, K. (2015). ‘’Demographic, employment, and wage trends in South Becker, G.S., and Tomes, N. (1979). ‘’An Equilibrium Theory of Africa.’’ UNU-WIDER Working Paper No. 2015/141. the Distribution of Income and Intergenerational Mobility.’’ Journal of Political Economy 87(6), 1153–89. Bhorat, H., Magadla, S. and Steenkamp, F. (2015). Employment Effects in the Temporary Employment Bhorat, H. (2004). ‘’Labour market challenges in the post Services (TES) Sector: Post Labour Relation Amendment apartheid South Africa.’’ South African Journal of Economics Act s198. Published mimeograph, Development Policy 72(5), 940-977. Research Unit: University of Cape Town. Bhorat, H., and Cheadle, H. (2009). ‘’Labour Reform in South Bhorat, H., Cassim, A., and Magadla, S. (2015). Relative Wage Africa: Measuring Regulation and a Synthesis of Policy Levels amidst Legislative Compliance: Remuneration in the An Assessment of Drivers, Constraints and Opportunities 115 Temporary Employment Services Sector in South Africa. Calvo, C., and Dercon, S. (2009). “Chronic Poverty and Unpublished mimeo, Development Policy Research Unit: All That: The Measurement of Poverty over Time”, In T. Cape Town. Addison, D. Hulme, and R. Kanbur (eds.), Poverty Dynamics: Interdisciplinary Perspectives, pp. 29–58. Oxford University Bhorat, H., Magadla, S. and Steenkamp, F. (2015). Press. Employment Effects in the Temporary Employment Services (TES) Sector: post Labour Relation Amendment Cappellari, L., and Jenkins, S. (2002). “Who Stays Poor? Who Act s198. Published mimeograph, Development Policy Becomes Poor? Evidence from the British Household Panel Research Unit: University of Cape Town. Survey.” The Economic Journal 112(478): C60–C67. Bhorat, H., and Thorton, A. (2016). The Employment Tax Cappellari, L., and Jenkins, S. (2004). “Modelling Low Income Incentive in South Africa: A Brief Assessment of the Initial Transitions.” Journal of Applied Econometrics 19(5): 593–610. Impact. Unpublished mimeograph, Development Policy Cappellari, L., and Jenkins, S. (2008). “Estimating low Research Unit: University of Cape Town. pay transition probabilities accounting for endogenous Bhorat, H., T. Caetano, B. Jourdan, R. Kanbur, C. Rooney, selection mechanisms.” Journal of the Royal Statistical B. Stanwix, and I. Woolard (2016). ‘’Investigating the Society: Series C (Applied Statistics) 57(2): 165–186. Feasibility of a National Minimum Wage for South Africa.’’ Case, A., and Deaton, A. (1997). School quality and Development Policy Research Unit Working Paper 201601. educational outcomes in South Africa. Mimeo, Research DPRU, University of Cape Town. Program in Development Studies, Princeton University. Bhorat, H., Jacobs, T. and Yu, D. (2013). An Overview of Strike Chitiga-Mabugu, M. (2016). South Africa Social Accounting Activities in South Africa, 1999–2011. Published Report, Matrix. University of Pretoria. Development Policy Research Unit Factsheet, University of Cape Town. Cichello, P.L., Almeleh, C., Mncube, L., and Oosthuizen, M. (2011). “Perceived Barriers to Entry into Self-Employment in Bjorklund, A., and Markus, J. (1997) ‘’Intergenerational Khayelitsha, South Africa: Crime, Risk and Start-up Capital Income Mobility in Sweden Compared to the United States.’’ Dominate Profit Concerns.” CSSR working Paper No. 300. American Economic Review 87(5): 1009–1018. Corak, M. (2013). “Income Inequality, Equality of Opportunity Botero, J., Djankov, S., La Porta, R., Lopez-de-Silanes, F., and and Intergenerational Mobility.” The Journal of Economic Shleifer, A. (2004). ‘’The Regulation of Labor.’’ The Quarterly Perspectives 27(3): 79–102. Journal of Economics 119(4): 1339–1382. Retrieved from: http://www.jstor.org.ezproxy.uct.ac.za/stable/ 25098720 Corak, M. (2016). “Inequality from Generation to Generation: The United States in Comparison.” IZA Discussion Paper No. Bratsberg, B., Roed, K., Raaum, O., Naylor, R., Jantti, M., 9929. Eriksson, T., and Osterbacka, E. (2007). ‘’Nonlinearities in intergenerational earnings mobility: Consequences Datt, G., and Martin, M. (1992). “Growth and Redistribution for Cross-Country Comparisons.’’ The Economic Journal Components of Changes in Poverty Measures: A 117(519): C72-C92. Decomposition with Applications to Brazil and India in the 1980s.” Journal of Development Economics 38(2): 275–95. Brunori, P., Ferreira, F., and Peragine V. (2013). “Inequality of Opportunity, Income Inequality and Economic Mobility: Deaton, A., and Zaidi, S. (2002). “Guidelines for Constructing Some International Comparisons” Policy Research Working Consumption Aggregates for Welfare Analysis.” LSMS Paper 6304. World Bank. Working Paper; No. 135. World Bank. Casale, D., and Posel, D. (2011). “Unions and the Gender Department of Labour (Republic of South Africa). (2016). Wage Gap in South Africa.” Journal of African Economies Annual Industrial Action Report. Pretoria: Department of 20(1): 27–59. Labour, South Africa. 116 Overcoming Poverty and Inequality in South Africa Department of Labour (Republic of South Africa). (2016). List distribution of earnings in the United States.” Research in of registered trade unions in South Africa on 1 December Labor Economics 22: 1–38. 2016. Pretoria: DoL. Available: http://www.labour.gov.za/ Finn, A. (2015). “A national minimum wage in the context DOL/downloads/documents/useful-documents/labour- of the South African labour market.” Southern Africa Labour relations/regtradeunions_dec2016.pdf and Development Research Unit (SALDRU) Working Paper Department of Trade and Industry (Republic of South 153. SALDRU, University of Cape Town. Africa). (2008). An Annual Review of Small Business in South Firpo, S., Fortin, N.M., and T. Lemieux (2009). “Unconditional Africa, (2005-2007). Department of Trade and Industry, Quantile Regressions.” Econometrica 77(3): 953-973. South Africa. FinScope South Africa (2011). Small Business Survey 2010. Du Plessis, P.G. (1978). “Concentration of economic power FinMark Trust. [Online]. Available: http://www.finmark.org. in the South African manufacturing industry.” South African za/wp-content/uploads/2016/01/FS-Small-Business_- Journal of Economics 46(3): 172–182. reportFNL2.pdf [Accessed 17 March 2017]. Duclos, J-Y., Esteban, J., and Ray, D. (2004). “Polarization: Fortin, N., Lemieux, T., and S. Firpo (2011): “Decomposition Concepts, Measurement, Estimation.” Econometrica 72(6): Methods in Economics,” in: O. Ashenfelter and D. Card (ed.), 1737 – 1772. Handbook of Labor Economics, volume 4A, chapter 1, Elbers, C., Lanjouw, J.O., and Lanjouw, P. (2002). “Micro-Level Elsevier. Estimation of Welfare.” World Bank Policy Research Working Foster, J. (2009). “A Class of Chronic Poverty Measures,” in T. Paper no. 2911. World Bank. Addison, D. Hulme, and R. Kanbur (eds.), Poverty Dynamics: Faber, M., Wenhold, F.A.M., and Laurie S.M. (2015). “Dietary Interdisciplinary Perspectives, pp. 59–76. Oxford, Oxford diversity and vegetable and fruit consumption of University Press. households in a resource-poor peri-urban South African Giesbert, L., and Schotte, S. (2016). Africa’s new middle class: community differ by food security status.” Ecology of Food Fact and fiction of its transformative power. GIGA Focus and Nutrition 56(1): 62–82. Africa, 1. Fedderke, J., and Naumann, D. (2011). “An analysis of industry Godfrey, S., Clarke, M. and Theron, J. with Greenburg, J. concentration in South African manufacturing, 1972–2001.” (2005). On the Outskirts but still in Fashion. Homeworking Applied Economics 43(22): 2919–2939. in the South African Clothing Industry: The Challenge to Fedderke, J., and Szalontai, G. (2009). “Industry concentration Organisation and Regulation. Monograph 2, Labour and in South African manufacturing industry: Trends and Enterprise Project, University of Cape Town. Restructuring consequences, 1972–96.” Economic Modelling 26(1): 241– in the clothing industry. 250. Girdwood, S., and Leibbrandt, M. (2009). Intergenerational Fedderke, J., Obikili, N., and Viegi, N. (2016). “Markups and Mobility: Analysis of the NIDS Wave 1 Dataset. NIDS concentration in South African manufacturing sectors: an Discussion Paper No. 15, 1-26. analysis with administrative data.” WIDER Working Paper International Labour Organisation (ILO) (2014). World Social 2016/40. Protection Report 2014/15: Building economic recovery, Ferreira, F., and Gignoux, J. (2011). “The Measurement of inclusive development and social justice. International Inequality of Opportunity: Theory and an Application to Labour Organisation, Geneva. Latin America.” The Review of Income and Wealth 57(4): Jalan, J., and Ravallion, M. (1998). “Transient poverty in post- 622-57. reform rural China.” Journal of Comparative Economics Fields, G. S. (2003). “Accounting for income inequality 26(2): 338–357. and its change: a new method, with application to the An Assessment of Drivers, Constraints and Opportunities 117 Kingdon, G.G. and Knight, J., (2004). “Race and the incidence Magruder, J.R. (2010). “Intergenerational Networks, of unemployment in South Africa.” Review of Development Unemployment and Persistent Inequality in South Africa.” Economics 8(2): 198-222. American Economic Journal: Applied Economics 2(1): 62– 85. Klasen, S., and Woolard, I. (2009). “Surviving Unemployment Without State Support: Unemployment and Household May, J. Govender, J., Budlender, D., Mokate, R., Rogerson, Formation in South Africa.” Journal of African Economies C., and Stavrou, A. (1998). Poverty and Inequality in South 18(1): 1-51. Africa: Report prepared for the Office of the Executive Deputy President and the inter-ministerial committee for Knight, J.B., and McGrath, M.D. (1987). The erosion of poverty and inequality. apartheid in the South African labour market: measures and mechanisms. University of Oxford Institute of Economics Mazumder, B. (2012). Is Intergenerational Economic Mobility and Statistics. Lower Now than in the Past? The Federal Reserve Bank of Chicago 297, April. Krueger, A. (2012). The Rise and Consequence of Inequality in the United States. Speech at the Center for American Moll, P.G. (1991). “The decline of discrimination against Progress. Washington DC. colored people in South Africa, 1970 to 1980.” Journal of Development Economics 37(1): 289–307. Labadarios D., Steyn, N.P., and Nel, J. (2011). “How diverse is the diet of adult South Africans?” Nutrition Journal 10(33). Nimubona, A.-D., and Vencatachellum, D. (2007). “Intergenerational education mobility of black and white Leach, D.F. (1992). “Absolute vs. Relative Concentration in South Africans.” Journal of Population Economics 20: 149- Manufacturing Industry, 1972–1985.” South African Journal 182. of Economics 60(4): 229–238. National Treasury (Republic of South Africa). (2018). Budget Leibbrandt, M., Woolard, I., Finn, A., and Argent, J. (2010a). Review, (February 2018). South Africa “Trends in South African Income Distribution and Poverty since the Fall of Apartheid.” OECD Working Papers, No. 101. National Treasury (Republic of South Africa). (2016). Employment tax incentive descriptive report. South Africa Leibbrandt, M., Woolard, I., McEwen, H., and Koep, C. Pretoria: National Treasury. (2010b). Employment and Inequality Outcomes in South Africa. University of Cape Town: Southern Africa Labour and Oaxaca, R. (1973). “Male-Female wage differentials in urban Development Research Unit. labor markets.” International Economic Review 14(3): 693– 709. Leventhal, T., and Newman, S. (2010). “Housing and child development.” Children and Youth Services Review 32: Oosthuizen, M., Mudiriza, G., Khan, S., Rooney, C. and 1165–1174. Thornton, A. (2016). Analysis on the Impact and Effectiveness of the BCEA Earnings Threshold. Unpublished mimeograph, López-Calva, L.F., and Ortiz-Juarez, E. (2014). “A vulnerability Development Policy Research Unit, University of Cape approach to the definition of the middle class.” The Journal Town. of Economic Inequality 12(1): 23–47. Oostendorp, R.H. (2013). Occupational Wages around Lund, C., Breen, A., Flisher, A.J., Kakuma, R., Corrigall, J., Joska, the World (OWW) Database; Update for 1983–2008. J.A., Swartz, L., and Patel, V. (2010). “Poverty and common Amsterdam. mental disorders in low- and middle-income countries: a systematic review.” Social Science and Medicine 71(3): 517– Paci, P. (2016). Beyond a Single African Story: A Multifaceted 528. Tale of Structural Transformation. Unpublished mimeo, World Bank Group, Poverty Global Practice. MacLeod, C. & Rankin, N. (2016). Econometric Analysis of the Employment Tax Incentive. Presentation to National Piketty, T. (2014). About Capital in the 21st Century (PDF). Treasury. August 2016. Thomas Piketty PSE Website. 118 Overcoming Poverty and Inequality in South Africa Piraino, P. (2015). Intergenerational Earnings Mobility and Statistics South Africa (2017). Estimates of National Equality of Opportunity in South Africa. World Development, Expenditure 2017. National Treasury: Pretoria. Republic of 67, 396–405. South Africa Plagerson, S., and Ulriksen, M.S. (2015). Cash transfer Tarozzi A., and Deaton, A. (2009). “Using census and survey programmes, poverty reduction and empowerment of data to estimate poverty and inequality for small areas.” The women in South Africa. GED Working Paper No. 4/2015. Review of Economics and Statistics 91: 773–792. International Labour Organisation, Gender, Equality Van der Berg, S. (1997). “South African social security under and Diversity Branch, Conditions of Work and Equality apartheid and beyond.” Development Southern Africa 14(4): Department. Geneva: ILO. 481–503. Ravallion, M., and Chen, S. (2003). Measuring pro-poor Van der Berg, S., Siebrits, K., and Lekezwa, B. (2010). growth. Economics Letters 78(1): 93–99. “Efficiency and equity effects of social grants in South Africa.” Senkal, A. (2017). Documenting the Stylized Facts on Stellenbosch Economic Working Papers 15/10. Department Structural Transformation in Southern Africa, forthcoming of Economics and Bureau for Economic Research, University as a background note for Systematic Country Diagnosis of Stellenbosch. (SCD), The World Bank. Van der Berg, S. (2014). “Inequality, poverty and prospects South Africa Social Security Agency (SASSA). (2016). Annual for redistribution.” Development Southern Africa 31(2): 197– Report. South African Social Security Agency (SASSA): 218. Pretoria. Van Praag, C.M., and Versloot, P.H. (2007). “What is the value Schotte S, Zizzamia, R., and Leibbrandt, M. (2017). Assessing of entrepreneurship? A review of recent research.” Small the Extent and Nature of Chronic Poverty, Poverty Dynamics, Business Economics 29(4): 351–382. and Vulnerability to Poverty in South Africa, forthcoming as World Bank. (2014). South Africa Economic Update: a background note for this report. Fiscal Policy and Redistribution in an Unequal Society. Statistics South Africa (2008). Measuring Poverty in South Washington, DC: World Bank. Africa: Methodological report on the Development of the World Bank. (2016). South Africa Economic Update #8: Poverty Lines for Statistical Reporting. Pretoria: Statistics Promoting faster growth and poverty alleviation through South Africa. competition. Washington, DC: World Bank. Statistics South Africa (2015). Methodological report on World Bank. (2017a). South Africa Economic Update #9: rebasing of national poverty lines and development of Private Investment for Jobs. Washington, DC: World Bank. pilot provincial poverty lines. Report no. 03-10-11. Statistics South Africa. World Bank. (2017b). Does Fiscal Policy Benefit the Poor and Reduce Inequality in Namibia? Washington, DC: World Statistics South Africa (2016). Quarterly Labour Force Bank. Survey, Quarter 3: 2016. Available: http://www.statssa. gov.za/publications/P0211/P02113rdQuarter2016.pdf [31 World Bank. (2017c). South Africa Economic Update #10: January 2017]. Statistics South Africa. Innovation for Productivity and Inclusiveness. Washington, DC: World Bank. Statistics South Africa (2017). Poverty Trends in South Africa: An examination of absolute poverty between 2006 & 2015. World Bank. 2018. Republic of South Africa Systematic Statistics South Africa. Country Diagnostic. An Incomplete Transition: Overcoming the Legacy of Exclusion in South Africa. Forthcoming, Statistics South Africa (2016b). Quest for nodal development: Washington, DC: World Bank. Evidence from Census 2001 and Census 2011. South Africa http://www.statssa.gov.za/publications/Report-03-10-14/ Report-03-10-142011.pdf An Assessment of Drivers, Constraints and Opportunities 119 ANNEX Technical background papers produced for this report Non-Income Dimension of Poverty, The nexus between food security and nutrition, and income poverty in South Africa, by Julian May and Coretta Jonah. Understanding changes in welfare: The case of South Africa 2005 – 2014, by Nga Thi Viet Nguyen and Victor Sulla. Structural Change and Patterns of Inequality in the South African Labour Market, by Haroon Bhorat, Safia Khan. SMMEs in South Africa. Understanding the Constraints on Growth and Performance, by Haroon Bhorat, Zaakhir Asmal, Kezia Lilenstein, Kirsten van der Zee. Social Assistance in South Africa, by Morné Oosthuizen. The Structure and Evolution of Inequality in South Africa from 2005 to 2015, by Kanishka Kacker. Something in the way they move? Patterns of labor mobility and earnings mobility in South Africa, by Kanishka Kacker. The Structure and Evolution in South African Inequality 2005 – 2015, by Kanishka Kacker. The Dual Nature of the South African Labor Market, by Kanishka Kacker. Assessing the Extent and Nature of Chronic Poverty, Poverty Dynamics, and Vulnerability to Poverty in South Africa, by Simone Schotte, Rocco Zizzamia, and Murray Leibbrandt. SOCIAL STRATIFICATION, LIFE CHANCES AND VULNERABILITY TO POVERTY IN SOUTH AFRICA, by Simone Schotte, Rocco Zizzamiab and Murray Leibbrandt. How useful is the South African National Income Dynamics Survey for dynamic social welfare analysis? by Kanishka Kacker. How does the Son Rise in the Rainbow Nation? Intergenerational Mobility in South Africa, by Kanishka Kacker. Background Note on Household Capability and the Distribution of Households Wealth with a Specific Focus on Wealth Inequality, by Carel van Aardt, Bernadene de Clercq, Johann van Tonder. 120 Overcoming Poverty and Inequality in South Africa NOTES: This report documents the progress South Africa has made in reducing poverty and inequality since the end of apartheid in 1994, with a focus on the period between 2006 and 2015. The main conclusions are as follows: First, by any measure, South Africa is one of the most unequal countries in the world. Inequality is high, persistent, and has increased since 1994. Second, although South Africa has made progress in reducing poverty since 1994, the trajectory of poverty reduction was reversed between 2011 and 2015, threatening to erode some of the gains made since 1994. High levels of inequality and low intergenerational mobility act as a brake on poverty reduction and as a result poverty is high for an upper middle-income country. Poverty is consistently highest among black South Africans, the less educated, the unemployed, female-headed households, large families, and children. Further, poverty has a strong spatial dimension in South Africa, a demonstration of the enduring legacy of apartheid. Poverty remains concentrated in previously disadvantaged areas, such as the former homelands – areas that were set aside for black South Africans along ethnic lines during apartheid. Third, high levels of income polarization are manifested in very high levels of chronic poverty, a few high-income earners and a relatively small middle class. Fourth, the role of skills and labor market factors have grown in importance in explaining poverty and inequality while the role of gender and race, though still important, has declined, presenting an opportunity for policy to influence poverty and inequality outcomes. Social protection remains important in reducing extreme poverty, but the fiscal space for further expansion is limited. Low growth perspectives in the coming years suggest poor prospects of eliminating poverty by 2030 as envisaged in the National Development Plan. Looking ahead, accelerating poverty and inequality reduction will require a combination of policies that seek to unlock the full potential of labor markets and promote inclusive growth through skilled job creation.