SWP715 The Household Economy of Rural Botswana An African Case Dov Chermichovsky Robert E. B. Lucas Eva Mueller WORLD BANK STAFF WORKING PAPERS Number 715 044-02 0282 Feathery jaiTies KI; K 905 WORLD BANK STAFF WORKING PAPERS Number 715 The Houselhold Economy of Rural Botswana An African Case Dov Chemichovsky Robert E. B. Lucas Eva Mueller The World Bank Washington, D.C., U.S.A. Copyright (© 1985 The Intemational Bank for Reconstruction and Development/THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First printing August 1985 This is a working document published infomlally by the World Bank. To present the results of research with the least possible delay, the typescript has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for enrors. The publication is supplied at a token charge to defray part of the cost of manufacture and distribution. The World Bank does not accept responsibility for the views expressed herein, which are those of the authors and should not be attributed to the World Bank or to its affiliated organizations. The findings, interpretations, and conclusions are the results of research supported by the Bank; they do not necessarily represent official policy of the Bank. The designations employed, the presentation of material, and any maps used in this document are solelyifor the convenience of the reader and do not imply the expression of any opinion Whatsoever on the part of the World Bank or its affiliates concerning the legal status of any country, tenitory, city, area, or of its authorities, or concerning the delimitation of its boundaries, or national affiliation. The most recent World Bank publications are described in the annual spring and fall lists; the continuing research program is described in the annual Abstracts of Current Studies. The latest edition of each is available free of charge from the Publications Sales Unit, Department T, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A., or from the European Office of the Bank, 66 avenue d'Ina, 75116 Paris, France. Dov Chemichovsky, Robert E. B. Lucas, and Eva Mueller are consultants to the World Bank; they are at, respectively, Ben Gurion University (Israel), Boston Univer- sity, and the University of Michigan. Library of Congress Cataloging-in-Publication Data Chernichovsky, Dov., The household economy of rural Botswana. (World Bank staff working papers ; no. 715) Bibliography: p. 1. Rural poor--Botswana. 2. Income distribution-- Botswana. 3. Households--Botswana. I. Lucas, Robert E. B. II. Mueller, Eva, 1920- . III. Title. IV. Series. HC930.Z9P625 1985 339.2'2 85-3176 ISBN 0-8213-0493-3 ABSTRACT In an effort to understand the causes of poverty and its perpetu- ation in an African rural economy, this study analysed household data from the Rural Income DistrLbution Survey (RIDS) conducted in 1974-75 in Botswana. The Republic of Botswana is a 'Largely rural, land-locked, and semi-nomadic nation located in Southern Africa. It had an estimated popu- lation of 1 million in 1983. About 85 percent of Botswana's population in the mid seventies comprised rural households. bore than 50 percent of these households were estimated to live below the Poverty Datum Line established by the Government. This fact: underscored Botswana's unequal income distribution; the wealthiest 10 percent received 50 percent of rural incomes in the mid-1970s. The findings of the study indicate substantial under- and unemployment in the rural areas of Botswana at the time of the survey; work in agriculture yielded low returns and employment opportunities outside the rural economy were limited. Wage employmtent, largely male specific, required out-migration primarily to mines in South Africa. Animal husbandry provided the lion's share, 44 percent, of average household income. Farming, although pursued with a relatively high share of family labor, provided only an estimated 11 percent of income. The marginal returns on working time particularly in activities of women and children were relatively low. One impediment to improvement of agricultural incomes was the relative unavailability of good cleared land for farming. This situation affected adversely mostly female-headed households because they had limited access to productive assets, including cattle, and to wage employment. Inspite of the immediate pressure on household incomes and presumably on capital formation, large rural families appeared justified at the household level; children provided security to the ill and the elderly, and worked on and off the farm. This was one implication of a lack of a widespread soc:Lal infrastructure in rural areas. Low iLncomes, Limited wage employment, the scope for child labor, and population movement, all combined to make schooling a costly and unattractive exndeavor. Yet, it appeared to attract those who had a few assets and, therefore, :Lncurred relatively low opportunity costs for schooling. ThiLs turned out to be one way:, encouraged by the governments abolition of tuition fees, to equalize opportunities in rural Botswana. Improvement in the quality of life among Botswana s rural population at the time of the survey, seemed to require raising the relative returns to crop production, and generating wage employment in the rural areas. Government policies deemedi necessary to provide capital investments in rural industry, rural commerce, cooperatives, land improvements (through rural credit and insurance), and so forth. In the short run, such policies would raise the incomes and living standards of the poor by provision of employment. The long run results might encourage smaller families and better education of Botswanna's future generations. CONDENSE La presente etude a pour objeztif de cerner les causes de la pauvrete et de sa perpetuation dans un pays africain a economie rurale. A cet effet, elle analyse les donnees concernant les menages fournies par l'Enquete sur la r6partition du revenu rural, realisee au Botswana en 1974-75. Le Botswana, Etat de l'Afrique australe, est un pays enclave, dont l'econormie repose surtout sur l'agriculture et dont la population, estimee en 1983 a 1 million d'habitants, est semi-nomadique. Vers le milieu des ann6es 70, les menages ruraux representaient environ 85 % de la population et plus de 50 % d'entre eux vivaient au-dessous du seuil de pauvrete etabli par lea Gouvernement. A lui seul, ce fait mettait l'accent sur le d6sequilibre de la repartition des revenus, puisque 10 % de la population recevaient 50 2 des revenus ruraux. Le; resultats de l'etude indiquent qu'a la date de l'enquete le ch6mage et le sous-emploi s6vissaient dans les zones rurales du pays; la rentabilite des travaux agricoles etait faible et les possibilites d'em- plois en dehors de ce secteur etaient limit6es. Les emplois salaries, pour la plupart des einplois masculins, iexigeaient l'emigration, princi- palement vers les mines d'Afrique du Sud. L'elevage, selon l'etude, procurait la plus grande part du revenu moyen des menages : 44 %. L'agriculture, malgre le recours intensif a la main-d'oeuvre familiale, n'en fournissait que 11 %. La rentabilite marginale du travail, particulierement des activites des femmes et des enfants, 6tait comparativiement faible. La penurie de terres arables et d6broussaiLlees, propices a :La culture, etait l'un des obstacles a P'am6lioration du revenu agricole. La situation etait particulierement critique dans les foye:rs dont le chef de famille etait une femme, ces dernieres n'ayant qu'un acces limite aux avoirs productifs, notamment b6tail, et A l'emploi salarie. Pour le foyer rural, une famille nombreuse semblait justifi6e malgre les pressions journalieres qu'un.grand nombre d'enfants exerce sur le revenu du menage et sur la formation de capital; les enfants repr6- sentant un e61ment de securite pour les parents en cas de maladie et pendant leurs; vieux jours, et pouvant travailler aussi bien sur l'exploi- tation qu'a L'ext6rieur, en l'absence d'une infrastructure sociale. Le faible niveau des revenus, des possibilites d'emploi r6munere limit6es, la generalisation du travail des enfants et les mouvements de population faisaient de la scolarit6 une activite coateuse et peu attrayante. Toutefois, ceux qui disposaient de quelques avoirs et pour qui, par consequent, le coat d'opportunite de la scolarite etait compara- tivement faible, semblaient s'y interesser, ce qui apparut comme un moyen d'ameliorer les possibilites d'acces a l'education, d'autant que l'Etat, pour encourager cet effort, a opte pour la gratuite de l'enseignement. Il est apparu indispensable, pour am6liorer la qualit6 de la vie de la population rurale du Botswana, d'aLccroitre la rentabilite de la production agricole et de creer des emplois remuneres dans les zones rurales. A cet effet, le Gouvernement a, juge qu'il 4tait necessaire de fournir des biens d'equipement destin6s a l'industrie rurale, au commerce rural, aux cooperatives, a l'amelioration des terres (par l'entremise de credit et d'assurance agricoles), etc. A court terme, ces strategies devraient permettre d'am6liorer le revenu et le niveau de vie des populations pauvres en leur fournissant de l'emploi. A long terme, elles pourraient encourager les familles a avoir moins d'enfants et promouvoir un meilleur niveau d'education pour les futures generations. EXTRACTO Con el fin (le comprender las causas de la pobreza y su perpetuaci6n en una economia rural de Airica, en este estudio se analizaron los datos correspondientes a unidades"familiares tomados de la Encuesta sobre Distribuci6n de los Ingresos Rurales realizada en 1974-75 en Botswana. La Repuiblica de Botswana es una naci6n predominantemente rural, mediterrinea y semin6made ubicada en el sur de Africa. En 1983, su poblaci6n se estim6 en un mill6n de habitantes. A mediados del decenio de 1970, un 85% de ella estaba constituido por familias rurales, de las cuales un 502 vivia, seguin las estimacicines, en condiciones inferiores a la Linea de Referencia de Pobreza fijada por el Gobierno. Este hecho pone de relieve la desigual. distribuci6n del ingreso en el pais; al 10% mais rico le correspondia, en esa epoca, el 50% de los ingresos rurales. Los resultados del estudio indican para esa fecha un sustancial desempleo y subempleo en las zonas rurales de Botswana; el trabajo agricola tenia rendimientos bajos y las oportunidades de empleo fuera de la economia rural eran limitadas. Para obtener empleo asalariado, principalmente en el caso de los hombres, era menester emigrar, casi siempre a las minas de Sudafrica. La cria de ganado constituia la parte principal (44%) del ingreso familiar medio. Los cultivos, aunque se realizaban con una participaci6n relativamente elevada de la mano de obra familiar, proporcionaban unicamente un 11% estimado del ingreso. El rendimiento marginal del tiempo de trabajo, en especial en las labores de mujeres y ninios, era relativamente bajo. Un impedimento para la mejora de los ingresos agricolas era la falta relativa de tierras despejadas, buenas para cultivar. Esta situaci6n afectaba adversamente a la mayoria de las familias encabezadas por una mujer porque tenian acceso limitado a los activos productivos, incluido el ganado, y al empleo asalariado. A p-esar de la presi6n inmediata en el ingreso familiar, y supuestamente en la formaci6n de capital, las familias rurales grandes parecian justificadas a nivel de la unidad familiar; los hijos proporcionaban seguridad a los enfermos y los ancianos y trabajaban dentro y fuera de la explotaci6n. Era un corolario de la falta de una infraestructura social generalizada en lias zonas rurales. Los bajos ingresos, el limitado empleo asalariado, las posibilidades de trabajo infantil y el movimiento de la poblaci6n se combinaban pa;ra que la asistencia a la escuela resultara costosa y poco atractiva. No obstante, aparentemente atraia a los que poseian unos pocos activos ya que les significaba costos de oportunidad relativamente bajos. Esta result6 una forma de igualar las oportunidades en el sector rural de Botswana, alentada por la eliminaci6n de los derechos de matricula decidida por el Gobierno. La mejora de la calidad de la vida de la poblaci6n rural de Botswana en la epoca en que se realiz6 el estudio parecia requerir el aumento de los rendimientos relativos de la producci6n de cultivos y la generaci6n de empleo asalariado en las zonas rurales. En las politicas oficiales se consider6 necesario proporcionar inversiones de capital para la industria y el comercio rurales, las cooperativas, la mejora de las tierras (mediante cr6dito rural y seguros), entre otras cosas. A corto plazo, dichas politicas aumentarian los ingresos y los niveles de vida de los pobres al suministrarles empleo. A largo plazo, los resultados podrian alentar la formaci6n de familias mas pequeiias y una mejor educaci6n para las futuras generaciones. PREFACE This volume constitutes the second large scale publication based on the Botswana Rural Income Distribution Survey, commonly known by its acronym RIDS. The first volume published in 1976 by the Central Statistics Office of the Republic of Botswana, The Rural Income Distri- bution Survey in Botswana (1974/5), summarizes extensively the events surrounding the survey and leading up to the initial report. It in- cludes data on survey design, maps, sampling procedures and related information, as well as a series of analyses of income in Botswana. After the first publication, further analysis of the data seemed warranted as RIDS pro-vides a wealth of unexplored data that portray economic life in rural Botswana and in Southern Africa in general. Furthermore, many facets of economic life in Botswana covered by the survey are not scheduled to be resurveyied in the near future. This need was answered by a World Bank research program culminating in the present volume. Socio-economic issues such as allocation of family resources, agricultural production, wage employment, schooling and fertility behavior, are studied froin the perspective of the house- hold economy. The study is based on time use data-- the scope and detail of wh:Lch are unprecedented at least in African studies. This volume is organized into eight chapters dealing with the socio- economic issues mentioned. The first chapter is an introduction to Botswana, its economy and the economic :Lssues underlying the studies which follow. The second chapter presents the conceptual framework within which the data are examined and l:he studies are organized. Six chapters follow, deal:Lng with allocation of time, crop production, wage employment, schooling and fertility behavior, and conclusions from the study. The various chapters were written by different authors. An attempt has been made to strike a balance between a set of independent essays and an integrated volume. Furthermore, although they deal with compli- cated conceptual issues, the chapters have been written with wide reader- ship in mind. Many should share any credit related/to RIDS. Special acknowledgment must be made to Dr. Derek Hudson, whose dedication and desire for per- fection led to one of the most successful household surveys. He was institutionally and technically supported by Mr. P. Nteta, the former Botswana Government Statistician, and by Mr. C. Norrlof and numerous interviewers in the field. The support for RIDS-related research on behalf of the Government of Botswana was extended throughout the research program by the current Government Statistician, Mr. F. Modise. Many other people have made their contributions to the project through technical and moral support, as well as through their research. We would like to acknowledge the support of the World Bank staff, T. King, R. Gulhati, M. C. Thalwitz, D. N. JordaLn, C. B. Boucher, as well as a group of researchers who used RIDS dataL and enhanced our studies-- most notably H. E. Dahl, C. Colclough, P. Fallon, C. Molomo, C. Smith (Allison), and R. Szal. We are especially grateful to Ms. F. Cartford for organizing and processing the data, and to Dr. P. WieEssner for her most enlightening comments throughout the research program. SUMMARY The studies in this volume use the data of Botswana Rural Income Distribution Survey (RIDS) to portray the micro-economic aspects of house- hold behavior of the traditional semi-nomadic people of rural Botswana. The prime objective of these studies has been to identify and quantify the obstacles to imaprovement in standards of 'Living and income distribution in southern Africa. In par-ticular we have looked at factors which limit the growth of household incomes, employment and productivity. Further, we have looked at constraints on the improvements of standards of living over time by reduction in family size and better education for children. By the means of these studies we hope to convey some understanding of circum- stances which perpetuate rural poverty or limit sustained rural development in southern Africa. The Republic of Botswana, situated in southern Africa, is a landlocked nation of 580,000 square kilometers with a population of about one million in 1984. Because of its relatively low population density, Botswana fits a common and general perception of Africa as an "empty land." As for many other Sub-Saharan countries this perception may be erroneous, however, once other productive resources are considered. Shortage of water is the most striking reality of Botswana. Rainfall is erratic in timing of onset and amount, and there is a very little surface water. Availability of water and the fertility of the soils in the different regions of the country, largely dominated by the Kalahari desert, have established the patterns of human settlement. More than eighty percent of the population is concentrated along the eastern side of the country. Most people depend on subsistence agriculture (sorghum and maize) and cattle husbandry as prime means of livelihood. More than half the population reside in villages with populations of less than five hundred. Concentration of people in large villages that range in size up to 40,000 is mostly seasonal. People divide theLr time between distant cattle posts, lands, and village dwelling. Since its independence in 1966, Botswana has demonstrated a most remarkable economic growth. Up to 1980, GDP expanded about 13 percent a year. Botswana's economic growth has been marked by a structural shift of the economy. The mining sector replaced agriculture as a major sector by the end of the seventies. Mining has made Botswana less dependent on weather conditions and disease prone livestock, and to an extent has reduced its dependency upon foreign financial assistance. Of course, it made Botswana more dependent on prices of diamonds and copper. Botswana's rapid economic growth has nonetheless created a range of interrelated socio-economic problems mostly related to the reliance of the economy on a narrow production base characterized by capital intensive cattle production and mining. These problems include unemployment in both urban and rural areas, and a deteriorating income distribution. According to RIDS data the average household in rural Botswana receives income from five to six sources, and this pattern of multiple income generating activities is common to all income groups. The largest single source of income in rural Botswana is animal husbandry, primarily cattle raising, but also small stock such as'goats, sheep, pigs, and poultry. Other significant sources of income are crop cultivation, wage employment and transfers, including remittances from mine workers in South Africa, urban residents, and relatives in other rural areas. A comparison between the median per capita income and the Poverty Datum Line (PDL) for rural areas reveals that more than 50 percent of the households of flive or more persons lived below the PDL established at the time of the suIvey. The RIDS data point to substantial underemployment of the labor force, a logical consequence of low labor productivity. Mueller presents a thorough analysis of how people allocate their time. Time spent on income earning activities appears strongly influenced by economic endowments. Households with low levels of assets and human capital produce lower economic returns on their time and thus are seen to work less. Animal husbandry and wage labor yield much more income per unit of time input t'han do other economic pursuits. Crop cultivation is a particularly unrewarding activity, probably because of Botswana's arid climate and the relatively small capital inputs into this activity. Due to economic circumstances aFnd tradition, households headed by women own less cattle to use for draught power and for accumulation of wealth. Hence these households are relatively worse than others. Labor utilization is also low in those male-headed households which are too poor to own cattle. The agricultural sector of Botswana generates insufficient employment for its labor force, although some agricultural work is pursued to the point of very low marginal productivity. There is relatively little crop work and wage employment. During most seasons of the year there is an apparent "abundance" of leisure time. In fact, most wage employment in rural Botswana is in non-agricultural pursuits. Lucas has found that the rural wage labor market appears as a dual labor market; there exists a higlher wage market which is protected by government wage-setting policies and encompasses an elite of privileged workers, and an "informal" market. The latter probably closely reflects the opportunity costs of labor, that, according to Mueller, are substantially lower than the prevailing rural wage. However, even this informal wage labor market does not offer sufficient work opportunities to the poor to break the linkage between low asset holdings and low income. As in many other instances, government minimum wage policies do not improve the situation, and may, in fact, be an obstacle to more employment in rural areas. The need for employment and income creation in rural areas of Botswana implies that productive resources other than unskilled labor, must be greatly increased. The resources in question include infrastruc- ture such as irrigation, land improvement, rural roads, electricity, fencing, agricultural extension and research, and marketing facilities. They also include investments in human capital and rural credit. Otherwise, only those who have access to mine work in South Africa are able to form some of their own capital. This conclusion is derived from Lucas's estimates concerning agricultural productivity. He estimates a fairly high average product for land and infers that land -- or more precisely good, cleared land -- is not freely available. The chief competitor for land is cattle grazing; it is also true that overgrazing is a serious ecological and economic problem. Absorption of labor in crop production of the growing group of young job seekers -- enlarged by the population growth and declining availability of jobs in South Africa -is doubtful. While the presence of more men could raise crop output, it could do so, Lucas' analysis indicates, only with greater use of other inputs, especially if no additional land were allocaterd to crop production. Given the shortage of productive assets other than cattle (which is in many ways subsidized), it is necessary to make the diversification of assets attractive to the farm population. This objective implies that people must be made acquainted with the advantages of alternative assets, that they be given access to financing, and-that these assets provide earnings, security and/or prestige. Examples of such might be farm and business equipment, land improvements, cooperative shares, better housing, or a secondary education for their children. A further problem is that the absorptive capacity of the agricultural sector, is limited by the shortage of trained personnel particularly wLth vocational skill. The shortage of educated people, especially those educated beyond primary school, severely limits the amount of experimentation with development programs that is feasible, and the amount of flexibility permitted in administering programs. Nonetheless, inspite of the implied high social benefits from education, the private cost-benefit equation associated with schooling, may have quite a different implication than what the social benefits would imply. Chernichovsky shows that children in rural Botswana play an important role in the household economy. Hence, in spite of their low marginal productivity, children's contribution to work (especially animal herding) and housework, raises the cost of education. The limited oppor- tunities in the wage labor market for primary school leavers make for low returns to education. Wage employment in South African mines do not require schooling. High drop-out rates suggest that parents and children are not fully persuaded of the benefits of education. The data suggest that controlling for other factors, children are most likely to attend school when their families can financially afford it, when their parents are educated, and when demands on children's time for economic work and household chores are not too high. It appears that the eudcational system in Botswana will have to become more employment and vocationally oriented, so that parents anticipate a return to education other than an uncertain chance of some kind of urban employment. Educational planning will have to get around the obstacles to school attendance that result from demands on children's time on the farm and in the household. Schools will have to exploit periods of slack demand for children's work and tailor facilities more closely to the geographic location of the school age population. Women play an important role in production in Botswana with crop cultivation largely a woman's activity. Moreover, about 30 percent of households in Botswana are either permanently, or for lengthy periods, headed by females. Thus the training of women and the creation of employ- ment opportunities for women,, are essentLal aspects of the upgrading of human capital in rural areas. We have argued that the major objective for the rural sector must be the achievement of a rate of growth oi productive capital which substan- tially exceeds the rate of growth of the labor force. It remains to consider the possibility of slowing down population growth, and thereby labor force growth. An effective family planning program would be valuable component of a package of policies relating to rural development. They need to be complemented., however, by policies which reduce people's perceived need for children, because children play a significant role in providing assistance of various kinds, particularly to single women and to ill or elderly parents. Chernichovsky's work suggests that the security motive is a major reason for high levels of desired fertility in rural Botswana, in spite of the economic strain, in time and resources, childcare impose. Indeed, parents who can afford relatively more children have higher levels of fertilLty. This behavior is counteracted to a degree by higher levels of motheris' education among the well-to-do. Our findings suggest that programs to enhance female education and female earning oppor- tunities (but not children's employment) could contribute to fertility reduction. Programs to provide some minimal security for the aged would serve the same purpose. Such programs of course have justification in their own right. In summary, from the household perspective, raising the relative returns to crop production and generating wage employment in rural industries, as opposed to fostering cattle rearing, appear fundamental policy strategies for improving income distribution and economic welfare in rural Botswana. In the short run, such policies would raise the level of living and the security of the poor by provision of employment. In the long run, they might encourage smaller families and better educated people among Botswana's future generations. Table of Contents Page Chapter 1: Introduction 1.1 The Setting .......................................... 1 1.2 Th,e Economy .......................................... 4 1.3 The Rural Economy and Rural Poverty ................. 6 1.4 Policy and Research Issues .......................... 12 Chapter 2: Aspects of Household Behavior in Rural Botswana Conceptual Considerations and Framework 2.1 Introduction ........................................ 17 2.2 The Economic Environment ............................ 18 2.3 The Unit of Analysis-- The Household ................ 21 2.4 Household Income and Production ..................... 24 Chapter 3: The Value and Allocation of Time in Rural Botswana by Eva Mueller 3.1 Inl:roduction ..... . ............. 27 3.2 Dat:a and Setlting . . .............. 29 3.3 Some Analyt:Lcal Problems . . . 39 3.4 Productivity of Time . . ........................... 48 3.5 Determinants of Time Use . ........................... 57 3.6 Summary and Conclusions ............................ . 69 Table of Contents, continued: Chapter 4: The Distribution and Efficiency of Crop Production in Tribal Areas of Botswana by Robert E. B. Lucas 4.1 Introduction ......................................... 77 4.2 The Distribution of Inputs ............................ 78 4.3 Productivity and Some Deterrminants of Production.....106 4.4 Profits and Profitability ............................ 121 4.5 Concluding Remarks ................................... 130 Appendix: Regressions on Land Areas ................. 134 Chapter 5: The Distribution of Wages and Employment in Rural Botswana by Robert E. B. Lucas 5.1 Introduction ........................ 139 5.2 The Setting ......................... 141 5.3 The Distribution of Employment ......... 144 5.4 Occupations ......................... 148 5.5 The Distribution of Wage Earnings . ............. 150 Chapter 6: Socio-Economic and Demographic Aspects of School Enrollment and Attendance in Rural Botswana by Dov Chernichovsky 6.1 Introduction .. .................. 163 6.2 Setting and Data .................... 164 6.3 Conceptual and Analytical Framework . ............ 169 6.4 Correlates of School EnrolLment and Attendance ... 172 6.5 Summary and Conclusions ................... 181 Table of Contents, continued: Page Chapter 7: Socio-Economic Correlates of Fertility Behavior in Rural Botswana by Dov Chernichovsky 7.1 Introduction ......................................... 185 7.2 Conceptual and Analytic Considerations ............... 186 7.3 Economic Value and Costs of Children ................. 188 7.4 Demographic and Socio-economic Characteristics of Mothers .............................................. 192 7.5 Correlates of Fertility Behavior ..................... 196 7.6 Summary and Conclusions .............................. 206 Chapter 8: Conclusions 8.1 Introduction ......................................... 209 8.2 Analytic Framework and Data .......................... 210 8.3 Findings and Implications ............................ 212 Bibliography ................................................. 223 List of Maps and Figures Page Map 1. Botswana (IBRD 11895R) ....................... xvii Figure 4.1 Lands Against Adults Present ................. 85 Figure 4.2 Lands Against Men Present for Ploughing ...... 86 Figure 4.3 Lands Against Cattle ......................... 88 Figure 5.1 Rural School-Wage Profile at Age 30 .......... 154 Figure 5.2 Rural Age-Wage Profile at Standard 4 ......... 156 List of Tables Page 1.1 Key Economic Characteristics of the Rural Economy.. 8 1.2 The Distribution of Total Rural Household Incomes.. 10 1.3 Rural Household Per Capita and Total Incomes by Household Size ..................................... 11 1.4 Basic Welfare Indicators by Income Group ........... 12 3.1 Males: Distribution of Activity Time by Age ....... 32 3.2 Females: Distribution of Activity Time by Age..... 33 3.3 Variations in Activity Time by Season for Adults ... 36 3.4 Variations in Activity Time by Season for 7-19 Year 01s .37 3.5 Males: Percentage Distribution of Activity Time by Age and Location .40 3.6 Females: Percentage Distribution of Activity Time by Age and Location .41 3.7 Number of Days Engaged in Wage Labor Per Year by Age and Sex .............. .......................... 42 3.8 Comparison of the Contribution of Adults and Children to Economic Work .................................... 43 3.9 Determinants of Household Income for Households with at Least: One Male and One Female Aged 15-64, Wages and Wage Time Excluded . .. 50 3.10 Determinants of Household Income for Households with at Least: One Male and One Female Aged 15-64, Wages and Wage Time Included .52 3.11 Determirnants of Household Income in Female Headed Hous'eholds (with no Male aged 15-64) .54 3.12 Determinants of Individual Time Use, Males aged 20-64 .59 List of Tables, continued: Page 3.13 Determinants of Individual Time Use, Females Aged 20-64 ... .. ........................ 60 3.14 Determinants of Individual Time Use, Boys aged 7-19 .. .. ........................ 61 3.15 Determinants of Individual Time Use, Girls aged 7-19 .. .......................... 62 4.1 Mean Primary Income from Crops per Household with Crops .77 4.2 Lands Area by Income Class of Household and Sex of Head ................................................ 82 4.3 Multivariate Analysis of Lands Area ................ 90 4.4 The Distribution of Expend:Ltures on Crop Husbandry. 92 4.5 The Distribution of Farm Equipment by Value ........ 97 4.6 Frequency of Possession of Equipment Types ......... 99 4.7 Percentage of Time Spent on Crops by Non-Producing Households ......................................... 101 4.8 Family Labor Inputs ................................ 103 4.9 Distribution of Output and Land Productivity ....... 107 4.10 Crop Production Functions .......................... 115 4.11 Incremental Effect of Adult Availability ........... 119 4.12 Distribution of Profit and Profitability ........... 125 4.13 Profit per Unit of Output .......................... 127 4.14 Profit as a Function of Fixed Assets ............... 129 APPENDIX Chapter 4 Table A.1: Regressions of Lands Against Adults by Months Present During Year ...................... 135 Table A.2: Regressions of Lands Against Men Present During Ploughing . ... . ...... . 136 Table A.3: Regressions of Lands Against Cattle Owned .................,,,,,,,,.......................... 137 List of Tables, continued: Page 5.1 Rural Adult Population by Sex, Age, Education and Village Category ................................ 142 5.2 Rural Time Allocation to Wage Earning by Sex, Age, Education and Village Category .. 146 5.3 Occupation of Rural Wage Earners by Sex and Village Category ........................................... 149 5.4 Rural Age and Schooling Wage Profiles .............. 153 5.5 "Rates oE Return" to SchooLing and Experience ...... 159 6.1 School Enrollment by Sex and Age Group ............. 167 6.2 Distribution of Activities of Children Out of School (OUT) and In School (IN) bt Age and Sex . 168 6.3 Distribution of Activities of Boys and Girls Aged 9-13, by Month .. 170 6.4 Schooling Indicators by Household Demographic Characteristics . 173 6.5 Schooling Indicators by Household Socio-economic Characte*ristics .. 174 6.6 Regression Coefficients, Schooling Indicators as Dependent Variables . 176 7.1 Distribution of Activities of Boys and Girls byAge 189 7.2 Household Income and TransEer Incomes by Number of Living Children ......................... 191 7.3 Characteristics of Mothers ......................... 193 7.4 Mean Number of Children Ever Born by Maternal and Household Socio-economic Characteristics and Age Group .... .... 198 7.5 Regression Coefficients, Number of Children Ever Born as a Dependent Variable ....199 7.6 Household Gross Available Income, Ratio of Surviving Children and Ratio of Children in School by Mother's Years of Schooling .. 203 I Chapter 1: Introduction 1.1 The Setting The Republic of ]3otswana is situated in Southern Africa. A large landlocked natiorn, it is bordered by Zambia in the north, by Zimbawe in the east, by South Africa in the south, and by Namibia in the west. Botswana is a large plateau, 1000 meters above sea level which is mostly a shallow sand- filled basin with limited rainfall. Three distinct ecological zones typify Botswana's 580,000 square kilometers. The first is the catchment area of the Limpopo River in the eastern part of the country. While lacking rich soil and a reliable rainfall relative to the terrain in the other two areas, the land allows for some dry- land crops and cattle-raisiing. The second ecological zone, the Okavango Delta, is a swampland of 15,000 square kilometers in the northwestern region of the country. Here wildlife abounds, but the land is not arable on a regular basis. The third zone, covering about two thirds of the country, is the semi-arid Kgalagadi (Xalahari) desert. Extending well into South Africa and Namibia, this vast area occupies the central and southwestern areas of the country. With a sparse covering of grass, acacia bush and small trees, this area is suitable for closely controlled cattle-grazing. Hitherto, the dire shortage of water has inhibited full exploitation of its potential. Shortage of water is the most striking reality of Botswana. The long-term average rainfall is about 530 mm annually, with drought occurring in about four in ten years. Rainfall is erratic in timing of onset and amount, and there is very little surface water. - 2 - Availability of water and the fertility of the soils in the different ecological zones have dominated the patterns of human settlement. More than eighty percent of the population, estimated at 1.04 million in 1985, is concentrated in the first ecological zone, along the eastern side of the country. This region, which consists of 30,000 square kilometers and is less than one tenth of the land area, enjoys relatively favorable conditions: reasonably fertile soils, and in most years sufficient rainfall to produce good pasture for cattle and to permit arable agriculture to be pursued. In this predominantly rural nation, most people depend on subsistence agriculture (sorghum and maize) and cattle husbandry as a means of livelihood. More than half of the people reside in villages with populations of less than five hundred. Concentratio,n of people in large villages that range in size up to 40,000 is mostly seasonal. People divide their time between distant cattle posts, lands and village dwelling. They move their cattle herds in search of grazing grounds and water during part of the year; and during the rest they cultivate their lands, returning to the villages for rest and festivities. While the arid conditions and scarcity of water do not allow plant production to prosper significantly, cattle grazing is more drought resistant. Crops may fail with poor rainfall; at the same time, a herd of cattle is likely to survive the fragile desert conditions. Thus, there is a tendency to emphasize the importance of cattle and to increase herd size when possible. Size of cattle herds have consequently become a measure of wealth as well as of social prestige and political power. A system of cattle tenancy (mafisa) enables poor farmers to raise herds of wealthy owners whereby they can receive in return the milk and milk products of the herd as well as the - 3 - meat of fallen cattle. In addition, the poor farmers can keep some of the offspring born during the grazing tenure, and thus they have some potential of developing a herd of their own. Botswana's population comprises eight major tribes. The largest tribe is the Ngwats and the smallest is the Masarwa known as the Bushmen, representing the early stages of man's technological development. This population grows at an annual rate of about 3 percent. Until Independence in 1966 the popllation had been almost entirely dependent upon cattle rearing, subsistence production, and remittances from migrant laborers in South Africa. About one-third of the Botswana male population still make theiLr living by working in the mines and on the farms of South Africa. By 1978, some 15 percent of the population were living in the urban centers, and this proporti.on has been increasing rapidly at about 12 percent annually. The town of Gaborone, the largest: in Botswana with a population of 40,000, hosts government facilities and serves as an administration center for the local industry. Lobal:se, another smaller modern urban center in the southeast provides facilities for the meat industry. Francistown in the northeast has a variety of industries and ccmmercial activities. Largely as a result of mineral discoveries in 1973-74, two new towns have become important centers: Orapa, where! one of the world's largest diamond pipes has been opened, and Selbi-pikwe where major copper and nickel deposits were discovered. These deveLopments have helped generate broad improvements in the infrastructure and consequently in job opportunities for the population. - 4 - Rapid urbanization in Botswana creates as many problems as it may try to solve. As elsewhere, phenomena associated with rapid urbanization prevail: overcrowding, inadequate sanitary conditions and lack of social services. The focus of this book is, however, the rural economy where the majority of the Botswana people live, and to which the Rural Income Distribution Survey (RIDS) was confined. 1.2 The Economy Botswana was declared a British protectorate in 1861. It became an independent republic under the presidency of the late Sir Seretse Khama in 1966. Under British rule, the Botswana economy was by and large neglected.1 The British assumed that the country would eventually be incorporated into South Africa, and therefore never attempted to develop an internally prosperous society. They left in their wake an uneducated, unskilled and poor population, with a subsistence economy, largely dependent on migrant labor in the mines of South Africa. With Independence, a concentrated effort was made to develop the country's economy.2 A key element of this effort has been Botswana's desire to reduce its economic dependence on South Africa through diversification of 1 Jack Parson, "The Labor Reserve in Historical Perspective", a paper presented in 1980 at the African Studies Association in Philadelphia. 2 For an extensive discussion of Botswana's economic development, see C. Colclough and S. McCarthy, The Political Economy of Botswana, Oxford University Press, London, 1980. - 5 - the economy with multi- and bilateral technical and financial assistance.3 The country wenit through a series of custonn agreements with South Africa within the Soutih African Customs Union (SACWO) and established its own monetary system. Since its indepesndence Botswana has demonstrated a most remarkable economic growth. Official statistics put ithe Gross Domestic Product (GDP) at current market lprices at about 660 Pula per capita in 1978/79.4 This figure represents an eiLghtfold Lncrease since 1961;. Botswana' s economic growth is marlced by a structural shift of the economy as well. The miniing sector replaced agriculture as a major sector by the end of the seventies. In 1978/9, mining accounted for about 23 percent of the GDP, as compared withl the 16 percent share of agriculture. The former sector accountecl for only 63 percent of the GPD at the time of Independence. Mining has made Botswana less dependent on weather conditions and disease prone livestock, and to an extent has reduced its dependency upon foreign financial assistance. Since the early seventies Botswana has started to accumulate foreign resierves, mostly frcai its foreign trade surplus and revenues from the South African Customs Union agreement. Although substantially relieved fron financial constraints on economic development, a 3 Colclough and McCarthy, ibid., especially Ch. 6. See also, Richard Dale, "BotswEtna and its Southern Neighbors; the Patterns of Linkage and the Opinions in Stat:ecraft", Papers in International Studies, African Series No. 6, Ohlio University Cent:er for International Studies, Athens, Ohio, 1970. 4 The currency of Botswana at that period was the South African Rand (=$1.41). We adlopt here the common approach of expressing all values in Botswana current currency, the Pula. For this and related statistics the reader is referred to the Statistical Bulleti; published quarterly by Botswana's Central Statistic Office. major impediment to Botswana's econamic development has been the severe lack of human capital or skilled labor. 1.3 The Rural Economy and Rural Poverty Botswana's rapid economic growth has created a range of interrelated socio-economic problems mostly related to the reliance of the economy on a narrow production base characterized by capital intensive cattle production and mining. These problems include unemployment in both urban and rural areas, and a deteriorating income distribution. 5 Attempts to develop modern labor-intensive crop production in rural areas to alleviate these problems have not produced any significant results as of yet. The size of the rural sector in the economy is measured by its share in the Gross Domestic Product. Official statistics originally valued the GDP (in current prices) at 205.7 million Pula for the period 1974-75. 6 Thirty seven percent of this value, or 76 million Pula, was the rural value added. Based on RIDS data the rural contribution to the GDP was adjusted to 91.9 million Pula -- 21 percent above the original value.7 Table 1.1 presents some key economic characteristics of the rural economy as calculated from RIDS data. The income data include income in-cash 5 These problems are best articulated by Michael Lipton, Employment and Labor Use in Botswana: Final Report, Republic of Botswana, Ministry of Finance and Development Planning, Vol. I, Gaborone, 1978. 6 Republic of Botswana, Central Statistics Office, Statistical Abstract 1976, Gaborone, 1977. Table 58, p. 83. Hans-Erik Dahl, Rural Production in Botswana: A National Accounts Analysis of the Income Distribution Survey, Institute of Economics, University of Bergen, 1979. -7- and kind, but exclude inputed housing benefits. The largest single source of income in rural Botswana is animal husbandry, primarily cattle raising, but also small stock such as goats, sheep, pigs, and poultry. Other significant sources of income are crop cultivation, wage employment and transfers, including remittances from mine workers, trban residents, and relatives in other rural areas. In addition, there are a number of minor sources of income such as gathering wild food and fuel, hunting, manufacturing (largely food processing and crafts), trading and vendirg, construction and services. Together, these minor sources account for less than 15 percent of rural income. The second column of Table 1.1 presents figures showing the frequency with which various sources of income were reported by households. The average household in rural Botswana receives income from five to six sources and this pattern of multiple income generating activities is similar for all income groups. About nine out of ten households reported income from crop cultivation and animal husbandry. These frequencies somewhat exceed the frequency of land and animal ownership, indicating that among the poorer sections of the population some of this income is earned from helping others. Transfer incomes, and income from food gathering are also received by almost all hous,eholds. The proportion of gross income obtained in kind from each source is shown in the third column of Table 1.1. Agricultural income from crop production and animal husbandry is received in the form of cash. In the lower income groups, lhowever, both wages and transfers are more likely to be in kind. Overall, it is clear that the largest part of rural production is for household use or is exchanged in kind. Table 1.1 Key Economic Characteristics of the Rural Economy Activity of Percentage Percentage Percentage Percentage Source of Share of of Households of Receipts Distribution of Income Income by Reporting Received in Household Work Source Income from Kind by Income Time by Activitya this Source (1) (2) (3) (4) Crops 10.8 89.0 76.8 31.7 Animal Husbandry 44.4 90.2 66.0 41.4 Wage Employment 19.2 49.7 5.1 11.0 Manufacturing 3.5 67.2 2.0 -- Trading & Vending 1.2 11.5 1.7 5.3 Services & Construction 2.4 65.8 35.4 -- Hunting & Fishing 1.3 14.3 67.9 10.5 Gathering 5.2 95.8 93.6 -- Trans fers 11.9 92.9 27.8 -- Propertyb 0.0 0.1 0.0 -- All 100.0 -- -- 100.0 a This is a distribution of the sum of hours spent on income earning activities by all household members over 6 years of age. b Property income covers only income from financial assets, which is uncommon in rural areas. Income from cattle is included in cattle income; income from business in trading or manufacturing, etc. The distribution of income earning time among economic activities is shown in column 4. A comparison of this column with column 1 is instruc- tive. It discloses substantial disparities between time inputs into various activities and income earned from these activities. Animal husbandry and wage labor yield much more income per unit of time input than do other economic pursuits. Crop cultivation is a particularly unrewarding activity, probably because of Botswana' s arid climate and the small capital inputs into this activity. - 9 - Several points emerge from these data. First, most rural households piece together a living by engaging in a variety of income earning activities rather than specializing in one or two. Second, almost all households must be fairly self-sufficient. They raise at least part of their own food or obtain it by helping other households with agricultural activities. They also obtain some of their food and fuel by hunting and gathering. Third, receipts of gifts and other transfer income are very widespread, suggesting widespread reciprocity rather than a unidirectional pattern of transfers exclusively from better off to poorer households. The ruiral income distribution is presented in Table 1.20 8 On the basis of this dLstribution, the Gini coefficient, indicating the inequality of income distribution, is 0.60, According to the data, the wealthiest 10 percent of the households received above 50 percent of total rural incomes in 1974/75. 9 8 Numerous studies deal with various aspects of income distribution in Botswana based on RIDS in addition to the analysis in the original report, Barbara Watanabe and Eva Mueller, "A Poverty Profile for Rural Botswana"., DEDPH Discussion Paper Series, No. 81-83, World Bank, 1L981; Colelough and McCarthy, op. cit., Chapter 7; Peter Fallon,, "Incomesi, Poverty and Income Distribution in Rural Botswana: A Synthesis of Evidence Based on the Rural Income Distribution Survey", University of Sussex, 1980; Chris Molomo, "Rural Income Distribution Survey: Further Analysis", mimeo, presented at the Rural Income Distribution Seminar, Gaborone, July 1979; Christopher Colelough and Peter Fallen, "Rural Poverty in BotswanLa - Dimensions, Causes and Constraints", mimeo, presented at the RIDS Seminar, Gaborone, July 1.979. The original Gini coefficient reported in RIDS was 0.52. This figure was revised on the basis of a consensus that the upper income groups were seriousEly under-represented in the initial estimates. The new Gini coefficient was estimated by applying higher weights, as suggested by the CSO and Banb of Botswana officials, to the upper income groups (stratum 23). Actual estimates were not undertaken because the relevant data hcve not become public. The Gini of .60 is based on an extrapolation, expand- ing the Lorentz curve at the end of the - 10 - Table :L.2 The Distribution of Total Rural Household Incomes Percent Total Mean Percentile Income Annual Income Incone Group Received (Pula) Bottom 10% 1.3 176 10-20% 2.1 285 20-30% 2.8 380 30-40% 3.8 516 40-50% 4.4 597 50-60% 5.5 746 60-70% 6.8 923 70-80% 8.8 1,194 80-90% 13.3 1,805 Top 10%a 51.2 5,699 Top 5% 36.1 9,798 Top 1% 17.5 23,748 Top 0.1% 4.9 66,493 a Sampling weights of upper income households, shown in the survey as strata 12, were increased from 2.5 to 15 and freehold farm weights strata, from 2.6 to 3.0 per discussions and suggestions of Botswana's Central Statistics Office and Mr. D. Hudson of the Bank of Botswana. Source: RIDS, p. 84. This distribution has some clear welfare implications. The Poverty Datum Line (PDL) - the minimum required level of household income for basic needs, given household size and composition - was set at an income of $76 per capita.10 Table 1.3 presents figure showing that although total household income increased with household size, income per capita decreases for larger households. A comparison between the median per capita income and the PDL for 10 These adjustments were made by P. Fallon, "Incomes, Poverty ..." op. cit. Republic of Botswana, The Rural Incone Distribution Survey of Botswana 1974/5, Gaborone, 1976, Appendix 15, pp. 211-223. - 11 - Table 1. 3 Rural Household Per Capita and Total Incomes by Household Size Median per Total Income of Mean per Household Size Capita Income(Pula)a Median Household(Pula) Capita PDL 1 465 465 160 2 210 420 140 3-4 139 455 120 5-7 110 660 110 8-10 95 955 110 11+ 90 - 105 a Original RIDS Sample: Incttmes Inflated by 1.06 plus 20., See RIDS, op. cit., pp. 75. Source: Molomo, op. cit. rural areas reveals that more than 50 percent of the households of five or more persons in rural areas lived below the PDL defined minimum standards. Overall, it was estimated that about 50 percent of rural households lived below this minimnum in 1974/75.11 The relationship between income and some basic welfare indicators is illustrated in Table 1.4. The figures show that the ratio of surviving 11 Initial RIDS estimates put the figures at 45 percent, CSO, RIDS, op. cit., p. 223. Subsequent estimates by Muel-ler and Watanabe, op. cit., put the figure at 55 percent, after modifying the PDL given more information on household size and composition. - 12 - Table 1.4 Basic Welfare Indicators by Income Group Gross Ratio of Ratio of Number Total Value Available Surviving Children in of of Huts Incomea To Ever Born School to Huts (P) Children Children 6-14 Up to 192 0.72 0.50 1.93 117.4 193-1086 0.79 0.40 2.82 178.2 1087-4344 0.82 0.50 4.46 232.4 4345 or more b 0.93 0.62 4.55 580.3 a Breakdown of groups, according to the breakdown originally used in RIDS. b Does not include upper income groups. children of total everborn rises with incame, as does the ratio of children aged 6-14 in school, of total number in the group as well as the number and value of huts owned by the household. The data illustrate the implications of poverty, and point out the process of potential perpetuation of poverty across generations through poor health, housing, and low levels of education. 1.4 Policy and Research Issues The Goverment of Botswana has sought to improve rural inctmes and to alleviate poverty through diversification of the economy and creation of - 13 - employment.12 Policy in this direction has become increasingly important in view of the excess supply of Botswana males wishing to go to the mines in South Africa in search of work. Development programs were geared at creation and expansion of labor intensive activities. The Arable Lands Development Programme launched in 1975 (ALDEP) is the best example of a policy program in this direction. The purpose of this and similar programs is twofold: a) the creation of work opportunities and an increase of rural incomes, and b) self- sufficiency for Botswana in food production, particularly grain. At the same time, the Government of Botswana has developed a network of social welfare activities; providing food, particularly in drought striken areas, raising the peoples' level of education, and bettering the population's health. In view of the very general characterization of Botswana's social and economic policy, this research program strived for a better understanding of how Botswana's rural economy operates, and what people do to earn their livelihood. Particular attention was given to an examination of the impediments to achievement of higher and more equally distributed incomes, to universal schooling, and to achievement of smaller and presumably better-off families. From a-more specific policy perspactive, a few major issues could possibly be addressed with the data: 12 Republic of Botswana, Ministry of Finance and Development Planning, National Development Plan 1976-19831, Gaborone, 1977. - 14 - (A) The degree to which labor is underutilized, and the extent to which labor utilization depends on other inputs as well as on the individual's and household's demographic and socio-economic characteristics; (B) The nature of income generation in rural Botswana, with particular attention paid to crop production and wage employment; (C) The extent to which school enrollment and attendance depends on household characteristics and interferes with economic activities; (D) The dependence of fertility behavior and thus family size on the mothers' and the households' characteristics. These issues are examined in the following six chapters. In the next chapter, we present the conceptual framework on the basis of which the various studies are organized. It deals with the basic functional and structural relationships being explored by the study, and serves as the basis for the hypotheses tested in the chapters that follow. Chapter 3 by Eva Mueller deals with the household's allocation of its prime asset, time. This issue is crucial to the study of patterns of employment and the possibility of their change. For the first time in the African context an attempt is made to establish, on a large scale, the correlates of time use by household members in various households of different socio-economic strata. We learn about the distribution of time to income earning activities, including wage employment, to household chores, including raising children, to schooling, and leisure. Mueller's essay is followed by two chapters by R. E. Lucas that focus on two major income generating activities: crop production and wage employment. The first, Chapter 4, deals with the determinants of crop - 15 - production and thereby deEines the parameters which are crucial to its spread throughout Botswana. The chapter on wage employment, Chapter 5, deals with the rural labor market an, the consequences of schooling for usage labor. It delineates, on the one hand, the tenuous nature of the rural labor market, and on the other, tha importance and relevance of skill in this market, and how education contrilbutes to gainful employment. The sixith chapter builds on the previous chapter through discussing the market benefits of schooling. In this chapter, Dov Chernichovsky deals with the determinants of school enrollment and attendance. Its main focus is identification of the factors inhibiting sclhooling which is the major means of social mobility in rural Botswana. Correlates of fertility behavior are examined in the last chapter, Chapter 7. This behavior is of course a major determinant of family size and as outlined in earlier sections, is an important aspect of poverty and family welfare. Chernichovsky examines the motives underlying particular levels of fertility. - 17 - Chapter 2: Aspects of Household Behavior in Rural Botswana: Conceptual Considerations and Framework 2.1 Introduction The data from the Rural Income Distribution Survey (RIDS) underscored many of the prior notions concerning economic life in rural Botswana. It is the typical traditional rural settirLg of a semi-nomadic population. People earn most of their income from self-*employment. A variety of economic activities in different locations comprise the.household enter- prises in which people engage to generate the major part of their means of subsistence. In this way, households also spread the risk associated with particular sources of income. The division. of labor by age and sex is fairly delineated, assuring that work is shared by all able household members. Wage employment is limited, the prime source being the mines of South Africa where many of the young and able males spend a substantial part of the year.1 The objective of this chapter is to outline the considerations and framework structuring the various essays comprising this volume. First, we broadly delineate the unit of analysis - the household - and its economic environment. Then, we attempt to identify, from an economic perspective, factors which stimulate particular household behavior and thereby determine the levels of particular "outcomes" which they produce. The main behavior 1 On the ways of life in rural Botswana, see Isaac Schaper, Married Life in an African Tribe, London: Faber and Faber, 1940. - 18 -- examined is the allocation of time to activities which are assumed to enhance household welfare. The outcomes considered are different aspects of household income, schooling, and fertility.2 This chapter serves a few interrelated purposes. First, it states the simplifying assumptions underlying our approach. Secondly, it is used to identify the relationships which are believed to be relevant to the household economy and which can be studied with the data. Thirdly, although not formally derived here, it provides the general framework within which these basic hypotheses are tested. Fourthly, this framework serves as an organizational tool according to which the various topics are addressed. The following sections of this chapter are concerned with the economic environment pertinent to our subject, the unit of analysis, and the processes of household production. 2.2 The Economic Environment The labor market where individuals buy and sell labor for wages, and the capital market where households or individuals buy and sell capital services, are crucial aspects in the study of economic behavior. As shown by the data presented in the previous chapter, opportunities to earn wages in rural Botswana are dismal. Our data indicate that there is almost no labor market for young people under 18-20 years of age, and there is persistent evidence that the country is stricken by under employment and 2 The notion of "behavior" and "behavioral outcomes" are rather arbitrary. As discussed in a section which follows, lines of causality are complex and, as a consequence, cause and effect or "behavior" and ''outcome" are often a matter of assumption. - 19 - unemployment.3 One segment of the labor market is the mines of South Africa. However, as pointed out already, data concerning the decision to become a migrant: worker, as well as data oni remittances, is scanty and could not be considered useful for our analysis. Under these circumstances, an assumption which is common in economic studies, that of "free" movement into and out of the wage labor force at all times, has been unrealistic here. This means that for some minimum "reservation wage" or institutional wage, some individuals cannot find employment :Ln the wage labor market. A few other matt:ers which complicate the employment issue must be considered. Sea.sonality is a crucial factor in agricultural production in Botswana. Many may be busy during the peak season; may even leave other jobs in favor of their own agricultural activities during this time; and yet may sit "idle" during other periods. Hence, any notions concerning "labor surplus" must take this factor into account. Traditional and institutional constraints are significant in Botswana. There is a fairly delineated division of labor between. males and females. This fact limits the reliability of a strictly economic perspective on allocation of time to production of "work", including wage employment. The RIDS data did not yield. information concerning capital transactions and capital formation that could give us a picture of the capital market. Yet some prevailing knowledge and notations concerning transaction and acquisition of assets, are pertinent to our discussion. Land in Botswana is tribal and has been traditionally allocated on the basis of "need" and 3 M. Lipton, Enployment and Labor Use in Botswana: Final Report, Vol. I, Republic of Botswana, Ministry of Finance, Gaborone, 1979. - 20 - usability. This reality, which is discussed in more detail in Chapter 4, has led us to assume that while land in Botswana is by-and-large free, usable land may be rather scarce. Drought-power is a critical element in agricultural production in Botswana. There appears to be a limited market for the renting of farm implements and cattle. However, as agricultural activities in Botswana depend ,critically on timing (often to the day) because of rains, drought-power can generally be rented by households who do not own their own equipment and cattle only "off" optimal time. We have no information on the scope of borrowing and lending in rural Botswana. Institutional borrowing in the large villages has been growing with the introduction of banks. However, it has been known that the people have always preferred investing in cattle rather than in financial assets, and banks' lending has been limited primariLy to housing and other assets which can serve as collateral. Since cattle serve as poor collateral, and apparently yield a higher private rate of return than financial assets, there is little economic scope for development of a formal capital market in this traditional setting. Hence, households in rural Botswana operate in an economic environment with little scope for dependence upon a market for buying and selling services. At the same time, people seize economic opportunities when they appear, demonstrating decidedly entrepreneurial behavior. In spite of the distances separating households and settlements, and of poor transportation, information seems to travel rapidly. A glass of home-brewed beer usually has a uniform price across the country. Probably because long distance and extended visiting is common, people appear well-informed on many - 21 - matters, including wages and job opportunit:ies. Such phenomena are believed to explain some consistent and systematic elements of economic behavior which we will attempt to formulate here. 2.3 The Unit of Analysis - The Household The household is the basic behavioral unit we attempt to study here. Yet, it could not be precisely identified and defined for analytical purposes. As in many other developing areas, in Botswana members of the same family who share the same familiar and economic endowments, tend often to be in different locations at any given time. This has complicated not only the count of the size of the household, but particularly the documentation of the flow of remittances and transfers, most significant in Botswana, from and into a specific dwellLng which was observed. Within dwellings we could identify about twenty "types" of families, in terms of relationships existing under the same roof. Variations have resulted largely from customary support sysl:ems for the aged as well as for the relatively young. In the case of the latter, for instance, it is common in Botswana to find children of virtually all ages being cared for by relatives. In recent years schooling has become a major reason for children to live with relatives who reside near schools. By and large the discussion here is confined to those family members who were present in the sampled dwelling during at least one of the interviews. While other family members are registered, information concerning their economic links with the household is extremely tenuous. For example, we - 22 - had to disregard an explicit account of the demographic and economic roles of mine workers.4 For analytical purposes, the household is considered to be an entity that, for its well-being, wishes to enhance the following welfare attributes: i) level of consumption, ii) the amount of leisure time its members enjoy, and iii) the number of children it has, as well as, iv) the educational levels of these children. In a situation where people work for their living and can work as much as they desire, people allocate part of their time to work in order to earn an income. Following analytical frameworks by Lancaster and Becker, the household is assumed to "produce" these attributes with inputs of time, and goods which are obtained through income. For instance, to "produce" children the household needs to allocate time for child care and other resources, such as food and clothing, obtained by sacrificing some income.5 From an economic perspective, the higher the household's income and the lower the costs of obtaining each of these four welfare attributes, the more the household can afford of them. The household in Botswana is assumed to secure its income from three potential sources: farm or household-enterprise production, wage earnings, 4 The Botswana National Migration Survey conducted in 1979 should shed some lLght on this matter. 5 Gary S. Becker, "A Theory of Allocation of Time" Economic Journal, Vol. 75, Sept. 1965; J. Kelvin Lancaster, "A New Approach to Consumer Theory", Journal of Political Economy 74, April 1966. - 23 - and transfers snd remittances from family members.6 In a non-traditional setting, incomes are generated "outside" the household-- in the market-- primarily by selling labor in the wage labor market, and by various forms of lending or renting capital and land in the capital market. For reasons discussed in the previous section, the implied classification of household activities as either market or non-market activities is unrealistic in a traditional setting such as the one in rural Botswana. The two facets of the household's economic behavior, the production of its welfare attributes and income are interwoven, because substantially, or most, income is generated on farms or in household enterprises. Hence, we consider the household in rural Botswana as a relatively self-contained economic entity because it does not buy or sell labor or capital services in what: we might consider a market.7 The household's production processes and! their analytical handling are discussed in the following section. 6 Transfers and remittances from family members are a major source of livelihood in Botswana. From a conceptual view point, we distinguish remittances from transfers, in that the former imply a household decision that one or more of its nmembers work elsewhere, commonly in the mines of South Africa. Transfers are considered gifts from outside the household. 7 Stephen Hymer and Stephen Resnick, "A Model of an Agrarian Economy with Noniagricultural Activities," American Economic Review, Vol. 59, No. 4, Sept. 1969, pp. 493-506; N. Howard Barnum and Lyn Squire, A Model of an Agricultural Household, World Bank Staff Occasional Papers, No. 27, Johns Hopkins Press, Baltimore, 1979; Robert Evenson, "Time Allocation in Rural Philippine Households", American Journal of Agricultural Economics, Vol. 60, No. 2, May 1978; R. Mark Rosenzweig, "Household and Non-household Activities of Youth: Issues of Modeling, Data, and Estimation Strategies", I.L.O. Informal Workshop on Children and Employment, Geneva, 1979, mimeo. - 24 - 2.4 Household Income and Production In this section we consider expLicitly the household's sources of income and two "household production" processes: raising children and educating them. As already mentioned, we consider here three sources of income: production of farm or household enterprise, wage employment, and transfers and remittances. The first source requires time inputs of household members, productive assets such as land, agricultural and other implements, technical know-how and intermediate products such as drought power, seeds, etc., all depending on the particular production process. Earnings depend on the amount of time the household sells in the labor market, and the wage rates that each unit of time commands. A person's wage rate in a given labor market, is known to be relative to his or her level of education and other characteristics such as age - which usually indicates levels of experience and seniority - and gender- which may be a basis for institutional discrimination in the labor market. Remittances should not be differentiated in most respects from wage earnings, but given the special nature of this source of income and the difficulties in measuring it properly compared with other sources of income, it is handled separately to the degree possible. It is clear that it may often be hard to distinguish between remittances and transfers (gifts) in particular, because the definition of a household becomes crucial on this matter. The process of child-care and education are of particular interest to us from an economic perspective. Child care requires the allocation by household members of time which could be used for leisure or to produce - 25 - income, and the sacrifice of consumption by parents. Similarly, schooling consumes children's time, which could be used for work and housecare, as well as other resources. In summary, we wish to look at how the household generates its income by particular sources, a.nd what determines the levels of time inputs to "work" or to income producing activities, and the levels of incomes generated. The particular production processes we shall study because of their particular relevance to issues of unemployment and because of data problems are crop production and wage earnings. In addition, we shall look at levels of fertility and education -as outcomes of the household's (economic) decision- making process. Before turning to the analyses in the following chapters, a few circular, or simultaneous, relationships must be acknowledged, as they present us with some of our basic analytical problems. First of all, when the household members work as; much as they wish, the level of income generated through work and. the level of leisure are co-determined; i.e., while we stipulate that leisure depends upon income, income depends upon leisure. Secondly, in a dynamic framework, the productive assets which determine level of income stem from savirgs which in turn dlepend upon past incomes. Thirdly, household size determines the household's labor endowments and hence its income. That is, over space and time some basic socio-economic outcomes (time allocation, income and holdings of human and non-human capital) are decided upon in a manner in which cause and effect are not easy to establish. Consequently, for both analytical and empirical reasons we assume that particular socio-economic circumstances and phenomena are not "outcomes" of the household decision-making process in what we consider to be the short - 26 - run. Rather, they are the ultimate determinants of the particular outcomes at which we look. These determinants include: household size and composition; levels of human and non-human capital; wage rates household members can command in the labor market; and levels of remittances and transfer incomes. In the chapters which follow we spell out specific assumptions and hypotheses concerning the particular topic studied. - 27 - Chapter 3: The Value and Allocation of Time in Rural Botswana Eva Mueller 3.1. Introduction In rural Botswana, as is true of the agricultural sector in other developing economies, a substantial portion of economic activity takes place outside the so-called formal and monetized sector. As a result a considerable amount of wealth and economic activity is difficult to measure. Prior to the Rural Income Distribution Survey (RIDS) little information was available to the Government of Botswana about levels of income and the process which generates incomme in the rural sector. A major step toward understanding this process and assessing the relation between rural resources and rural activity patterns was the collection of time use data by RIDS. TraditiLonally the economic activilties of rural populations have been studied by collecting data on labor force lparticipation, employment, and unemployment. The time use data extend these measurements by providing information on hours worlced for wages and Ln family enterprises which may be viewed as an indicator oE the extent or duration of labor utilization. We are particularly interested to learn how holdinigs of productive assets (including education) affect time aLlocations to economic work.1 That is, we shall investigate how inequali;ties in resources ownership affect the opportunity of peasant families to engage in productive work. 1 In this chapter the terms market work, economic work, and labor refer to activities which contribute to GNP, including unpaid family labor and work that earns income in kind. Housework is a separate and distinct time use category. - 28 - Since time use data are a relatively new type of information, their usefulness in broadening the analysis of welfare problems should be pointed out. Time can be directly consumed in the form of leisure and it can also be regarded as an input into the generation of income and other utilities such as education and the rearing of children.2 It has been clear all along that labor market activities represent only a single facet of households' and individuals' welfare generating pursuits. Yet key economic problems such as under- and unemployment have been discussed as if only economic work were "productive" of human welfare.3 Against this background, time use data -- providing a complete account of activities -- appear to have major potential. They offer the possibility of extending the scope and definition of "productive activity" beyond that relating to the generation of income. Thus time use data can improve the measurement of human welfare and, when accompanied by information on human and non-human capital, income and household characteristics, can contribute to our understanding of how differentials in levels of welfare, broadly defined, are brought about. This study is distinctive in that it views every member of the household above age 6 as a potential contributor to household welfare. As part of this investigation, we shall examine how the age/sex composition of the household and its size affects its time allocation and its labor supply. In Chapters 6 and 7 inferences are drawn from this analysis regarding the 2 The term income is used to include income in kind. 3 The interest in time use data has been increasing in more developed countries as well, as time allocated to labor market activities has been steadily decreasing. - 29 - opportunity cost of school attendance and the value of children's labor as a motive for high fertility. The framework outlined in Chapter 2 assumes that time allocations are responsive to economic incentives. That is, these allocations are expected to be sensitive to income and price-of-time (or productivity) effects. At the same time we recognize that a culturally determined division of labor by age and sex constrains household choice. A further purpose of this chapter is to clarify the relative role of economic and institutional influences on time use decisions. Five sections follow this introduetion. In the second section we discuss the data and describe time use patiterns in rural Botswana. Then we address some conceptual problems in the thLrd section. Productivity of time and the determinants of itime allocation are analyzed in sections four and five respectively. Conclusions are deferred to the end and are presented in section 6. 3.2. Data and Setting The tinme allocat:ion data pertain to the day prior to the interview and were obtained only in rounds 3, 5, 7, c9 and 11, of the twelve survey rounds. Interviews were spread over all days of the week, including Saturdays and Sundays. All persons in the household, 6 years of age and above, were asked to recall the previous day's activities and the approximate amount of time each activity occupied in chronological order from the time they got up in the morning until they retired at night; mealtime was omitted. The period not covered by the time use study, February through April, falls into the busy season. Thus the time frame of the study leads to some - 30 - understatement of economic activities. Further, some minor activities seem to be underreported in the survey. Any underreporting of work activities probably implies an overstatement of leisure time. On the other hand, "rest stops" during working hours and housework or a leisurely work pace could lead to a considerable overstatement of working hours. Almost all the data presented in this chapter are based on pooled observations for all five visits and thus average out seasonal variations. For the most part the data are presented on an individual rather than a household basis. Non-response was negligible; for about 100 households the time use information was too incomplete to be usable for analysis. This exclusion is equivalent to a non-response rate of 10 percent. The 957 households remaining in the sample contain about 4,600 individuals over 6 years old.4 Although great emphasis was placed on accuracy during the conduct of the survey, the level of detail of the time use data leaves something to be desired. For many people, only one or two major activities are recorded for each day. Some short-duration activities like carrying water are reported by most households, but it is likely that other activities which occupy little time or are of little salience to respondents are under-reported. It follows that the RIDS time use data describe the allocation of time to major activities only. There is no reason to suspect that underreporting varies systematically with population characteristics. Hence, differences in time use between sub-groups of the population may be assumed to reflect real 4 Individuals had to be present at 7 or more of the 12 survey visits to qualify as household members for purposes of the time use analysis. - 31 - differences in behavior, rather than errors in the data. Comparisons of the relative frequencies with which major activities occurred under various conditions also should be valid. Undoubtedly, people in rural Botswana do not keep precise track of time during their daily activities; they merely know that they devoted half of the day to one activity and half to another. In cases where the respondent reported time use in terms of fractions of a day, the interviewers were instructed to assume a day of 12 hours (roughly the time from sunrise to sunset). Therefore the distribution of time between activities probably is more reliable than the absolute amount of time spent on activities. For this reason the descriptive tables show percentage distributions, rather than mean amounts of time spent on various activities. Tables 3.1 and 3.2 present an overiiew of time use within age groups, for men and women respectively. Patterns oE time use vary sharply by age and sex. Animal husbandry is predominantly a male activity, and men are also more involved in wage labor than women. Women spend the largest part of their income earning time on crop cultivation, buit this is not exclusively a female activity. Men participate in certain essenitial operations such as land clearing, plowing and planting. Women aind girls (lo most of the gathering, and they engage to a lesser extent in other economic activities. Their most time-consuming activities are, however, houisework, childcare and fetchling water. In all, adult men spend about 75 percent more time on income earning activities than adult women. Despite ithis, aduLt women still reported on the average nearly 25 Table 3.1 Males: Distribution of Activity Time by Age a A G E Activities 7-9 10-14 15-19 20-29 30-39 40-49 50-59 60+ All Males Percentage Distribution of Total Time Crop Husbandry 2.1% 3.0% 3.5% 5.2% 6.1% 9.1% 7.7% 10.5% 5.4% Animal Husbandry 22.3 28.8 23.9 12.5 15.1 10.6 12.3 9.2 18.7 Wage Labor 0.4 0.4 2.0 12.2 8.1 7.4 5.5 1.9 4.0 Trading, Vending, Processing 0.1 0.1 0.7 0.9 1.0 1.9 1.1 1.6 0.8 Hunting or Gathering 1.2 1.6 1.9 2.0 2.6 3.0 1.9 2.3 2.0 All Income Earning Activities 26.2 33.9 32.0 32.9 32.8 32.1 28.4 25.6 30.8 Repairing, New Building 0.8 0.5 1.6 1.9 2.6 3.3 5.6 3.7 2.1 Fetching Water 1.6 2.3 2.2 1.9 1.7 1.4 0.2 1.1 1.7 Child Care 3.8 1.7 0.9 0.5 0.5 0.1 0.0 0.2 1.2 Housework 2.8 4.4 5.1 5.2 3.0 4.0 2.2 2.4 3.8 All Housekeeping Activities 8.9 8.9 9.8 9.6 7.8 8.7 8.0 7.3 8.7 Schooling 11.1 13.7 9.3 1.1 0.3 0.6 0.3 0.1 6.1 Illness & Health Care 1.5 1.5 2.6 3.0 2.5 2.9 5.3 8.4 3.2 Meetings 0.0 0.0 0.1 0.4 1.3 2.0 2.7 2.9 0.9 Leisure 52.3 42.0 46.2 53.1 55.3 53.6 55.3 55.7 50.3 All Non-Work Activities 53.8 43.5 48.8 56.5 59.1 58.5 63.2 67.0 54.4 All Activities 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% Based on five rounds of interviewing. A varying number of people in each category answered the activity questions in the five rounds. Only daytime activities have been covered; i.e., sleeping at night is excluded. All tables based on RIDS. Table 3.2 Females: Distribution of Activity Time by Age a ~~~~~~~~~~ ~~~~~~~~A G E Activities 7-9 10-14 15-19 20-29 30-39 40-49 50-59 60+ All Females Percentage Distribution of Total Time Crop Husbandry 2.4% 3.5% 6.0% 8.6% 10.2% 12.8% 13.4% 11.5% 8.0% Animal Husbandry 3.2 3.8 2.1 1.5 1.2 0.9 0.5 0.5 1.9 Wage Labor 0.1 0.8 2.1 2.0 1.1 1.7 0.8 0.1 1.2 Trading, Vending, Processing 0.0 0.5 1.5 1.7 3.0 1.6 1.8 1.2 1.4 Hunting or Gathering 1.6 2.6 2.8 2.5 2.7 2.5 2.8 2.3 2.5 All Income Earning Activities 7.2 11.2 14.4 16.4 18.1 19.5 19.2 15.6 15.0 1 Repairing, New Building 0.5 0.8 2.2 3.2 4.3 5.5 5.8 4.9 3.1 Fetching Water 4.8 6.3 7.7 7.8 7.4 6.4 5.8 4.4 6.5 Child Care 10.5 5.5 3.4 6.2 3.5 1.9 1.5 1.5 4.5 Housework 9.5 15.5 20.8 22.3 19.5 18.7 18.4 13.5 17.8 All Housekeeping Activities 25.3 28.3 34.1 39.5 34.7 32.6 31.5 24.3 32.0 Schooling 14.4 17.4 8.3 1.2 0.3 0.3 0.2 0.1 5.8 Illness & Health Care 1.1 2.0 3.6 4.6 5.6 6.3 6.4 8.6 4.5 Meetings 0.0 0.2 0.4 0.3 0.6 0.9 0.4 0.5 0.4 Leisure 52.0 41.0 39.2 38.0 40.6 40.5 42.3 50.9 42.2 All Non-Work Activities 53.1 43.1 43.1 42.9 46.9 47.7 49.1 59.9 47.1 All Activities 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% a Based on five rounds of interviewing. A varying number of people in each category answered the activity questions in the five rounds. Only daytime activities have been covered; i.e., sleeping at night is excluded. - 34 - percent less leisure than men, due to their involvement in housekeeping and childcare, in addition to economic work.5 Men and women aged 60 and over spend significant amounts of time in income earning activities. On the whole they seem to be a considerable asset to the household, or at least less of a burden than is often supposed. Boys seem to be heavily involved in income earning activities even before they are 10 years old, and at ages 10-14 they spend as much time on income earning activities as adult males. The job of taking care of the small stock in the household falls in large part to the younger boys, while older boys (10 and over) herd and water the cattle. Girls provide some help in the fields and in caring for the smaller animals, but their main contribution consists of child care and housework, to which they devote substantial amounts of time. Girls report slightly more school hours than boys; but over the year children of both sexes allocate at least twice as much time to market and housework as they do to schooling. It appears that children make a substantial economic contribution to the household by working and by relieving women of some domestic duties. A number of studies reporting labor inputs to agriculture in Africa have shown surprisingly low levels of manhours worked per year.6 The Botswana 5 For a more detailed discussion and additional data on the sexual division of labor, see Sherrie Kossoudji and Eva Mueller, "The Economic and Demographic Status of Female Headed Households in Rural Botswana," forthcoming in Economic Development and Cultural Change; Also C. A. Bond, Women's Involvement in Agriculture in Botswana, Gaborone, Botswana, 1974 Lmimeograpned) 6 D. Byerlee and C. Eicher, Rural Employment, Migration and Economic Development, African Rural Employment Paper #1, Michigan State University, East Lansing, MI, 1979; also J. Cleave, Labour in the Development of African Agriculture: The Evidence from Farm Surveys (unpublished Ph.D. dissertation), Stanford University, 1970. - 35 - data for adult males conform to this pattern. Some observers have attributed this phenomenon to the marked seasonality of agricultural work. The seasonality of time use Ln rural Botswana is illustrated in Tables 3.3 and 3.4. Men allocate about 60 percent more time to income earning activities in May than in November, and boys about 100 percent more. The seasonality of time use is even more pronounced for women and girls because crop production is largely "women's work." Time devoted by women to housekeeping activities and child care is curtailed somewhat during the busy season, but girls compensate for this shortfall by devoting extra time to domestic chores. In total, housework time seems to be quite insensitive to the seasonality in labor demand. Consequently, aduilts gain considerable leisure time during the slack work season. Children's time Ls shifted to school attendance in the slack season (vacations coinc:Lde roughLy with the busy work season). Some knowledgeabLe observers believe that Botswana is suffering from a great deal of underemployment. Michael Lipton estimated that "some 45 percent of potential labor time is wasted in idleness, and the productivity of the time at work is very ]ow."7 Quite apart from the accuracy of this estimate, the underemployment question involves a value judgment since there is no norm to indicate how much leisure time is "appropriate." Relatively, adult males and children who do not attend school report more leisure time than do women and children who attend school. 7 Michael Lipton, Employment and Labour Use in Botswana: Final Report, Republic of Botswana Ministry of Finance and Development Planning, Vol. I, Gaborone, Dec. 1978, Table 3.3 Variations in Activity Time by Season for Adults Males Females Visit Visit Visit Visit Visit Visit Visit Visit Visit Visit 3 5 7 9 11 3 5 7 9 11 Activity (May) (July) (Sept) (Nov) (Jan) (May) (July) (Sept) (Nov) (Jan) Crop Husbandry 16.5% 9.6% 2.0% 2.5% 6.7% 24.0% 16.2% 2.0% 1.1% 8.0% Animal Husbandry 13.5 11.0 10.3 11.8 11.9 1.3 1.0 0.7 1.0 1.2 Wage Labor 6.0 7.2 8.5 6.3 7.7 1.0 1.8 1.5 1.3 1.0 Trading, Vending, Processing 9.7 1.2 2.3 1.4 0.7 2.2 1.7 1.9 2.3 1.4 Hunting 2.5 3.2 2.7 2.3 1.1 2.2 3.4 2.5 2.5 2.2 All Income Earning Activities 39.5 32.2 25.9 24.3 28.1 30.7 24.0 8.6 8.1 13.7 Repairing, New Building 2.8 2.6 3.4 4.7 2.7 5.1 2.1 5.6 5.3 4.5 Fetching Water 1.2 1.3 2.0 1.3 0.9 5.0 7.0 7.6 8.0 6.0 Child Care 0.3 0.3 0.3 0.2 0.4 2.7 3.1 4.8 3.6 3.4 Housework 4.0 3.5 3.5 3.1 3.1 17.0 19.0 21.7 20.3 18.3 All Housekeeping Activities 8.3 7.6 9.2 9.4 7.1 29.7 31.3 39.7 37.1 32.2 Schooling 0.5 0.8 0.2 0.8 0.1 0.2 0.7 0.7 0.7 0.3 Illness & Health Care 4.2 5.6 5.9 4.4 4.3 4.3 6.2 7.5 7.3 5.0 Meetings 1.4 2.2 1.2 3.0 1.0 0.4 0.6 0.5 0.8 0.3 Leisure 46.0 51.5 57.6 58.1 59.3 34.6 37.2 43.0 45.9 48.5 All Non-Work Activities 51.7 59.3 64.7 65.5 64.6 39.3 44.0 51.1 54.0 53.8 All Activities 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Table 3.4 Variations in Activity Time by Season for 7-19 Year Olds Males Females Visit Visit Visit Visit Visit Visit Visit Visit Visit Visit 3 5 7 9 11 3 5 7 9 11 Activity (May) (July) (Sept) (Nov) (Jan) (May) (July) (Sept) (Nov) (Jan) Crop Husbandry 7.4% 3.1% 0.5% 0.6% 2.1% 13.4% 3.0% 0.4% 0.5% 2.1% Animal Husbandry 31.4 24.2 21.9 18.0 33.4 4.4 2.1 3.0 1.7 4.6 Wage Labor 0.8 0.6 1.4 0.7 0.6 1.0 0.8 1.2 1.2 1.1 Trading, Vending, Processing 0.4 0.3 0.1 0.2 0.3 0.6 0.7 0.9 0.7 0.6 Hunting or Gathering 2.1 2.2 1.0 0.9 1.4 2.5 2.2 2.4 1.9 3.2 All Income Earning Activities 42.0 30.5 25.0 20.3 37.7 21.8 8.8 8.0 6.0 11.6 Repairing, New Building 1.0 0.9 0.8 0.9 0.6 2.1 0.7 1.2 1.4 0.6 Fetching Water 2.0 2.0 2.4 2.1 1.8 6.4 6.1 7.3 5.9 6.2 Child Care 2.6 2.1 2.4 1.6 1.5 7.4 7.0 6.1 5.3 4.4 Housework 3.8 3.8 4.4 4.6 4.2 16.1 14.6 16.5 14.3 17.7 All Housekeeping Activities 9.5 8.8 10.1 9.2 8.1 32.0 28.4 31.2 26.8 28.9 Schooling 7.1 17.4 10.9 19.2 3.9 4.9 21.2 13.0 21.4 6.6 Illness & Health Care 3.2 1.8 1.4 1.5 1.1 2.9 2.6 1.7 2.0 2.1 Meetings 0.0 0.0 0.0 0.1 0.1 0.1 0.0 0.1 0.5 0.3 Leisure 38.1 41.5 52.7 49.7 49.0 38.3 39.0 46.1 43.2 50.5 All Non-Work Activities 41.3 43.3 54.1 51.3 50.3 41.3 41.6 47.9 45.7 52.9 All Activities 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 - 38 - Tables 3.5 and 3.6 throw some light on the under- and unemployment issue which will be addressed later on in this chapter and volume. These tables compare time use in large and small villages and in Baralong Farms -- the major commercial farming area in Botswana. The data for Baralong Farms show how time is used in an area with relatively high labor demand. Men and boys in Baralong Farms work substantially more hours than men in other places, but boys still show relatively high school attendance. If men and boys in Baralong Farms are not "over worked," people in the other villages might prefer less leisure and more productive work opportunities. All household members, regardless of age and sex, have less leisure in Baralong Farms than elsewhere. Wage labor makes up a small fraction of market work as indicated by the figures in Tables 3.1. and 3.2. This finding is borne out by the data in Table 3.7 which reveals that there is practically no wage labor market for children under 15, and that for young people 15-19 wage employment is rare and of short duration.8 About 20 percent of the women aged 20-59 worked for wages at some time during the survey year, but fewer than 6 percent had jobs which lasted 180 days or more during the year. In comparison, about 30 percent of males aged 20-59 had wage jobs, including about 15 percent with durations in excess of 180 days. The figures in Tables 3.3 and 3.4 show no appreciable seasonality in wage labor. It thus appears that households which do not have sufficient land and cattle to employ all household members productively often 8 This table is based on more extensive data on number of days engaged in wage labor each month, obtained in the 12 monthly interviews concerning household income. - 39 - may not be able to find adequate employmeni: in the wage labor market even during the busy season. Table 3.8 enables us to compare the total time contributed to economic work by adult men, women, and children. A striking fact emerges that, because of the high birth rate and the high out-migration rate, men between the ages of 20 and 59 constitute a mere 16 percent of Botswana's rural population and contribute only 24 percent of all incozme earning time. Women aged 20 to 59 constitute 27 percent of the population and account for 24 percent of income earning time. Boys aged 7 to 19 contribute time in excess of their proportion in the population, while the reverse is true for girls. In all, children and young adults (7 to 19) account for 42 percent of all income earning tiLme. Howover, there is a good deal of diversity between and within age/sex groups in regard to labor productivity and time allocations. In the sections which follow, we shall attenmpt to explain these diverse patterns. 3.3. Some Analytical Problems Before embarking on the analysis of the value or productivity of time and time allocations, we must consider a few conceptual issues.* The first concerns the characteristics of the labor market in rural Botswana. In many rural economies a wage labor market exists, allowing households which are asset-poor to hire out some of their labor to households, business firms, or government agencies which have larger holdings of productive assets. The process of hiring labor out: and hiring in tends to equalize the marginal I am indebted to D. Cherrnichovsky for his contribution to this section. Table 3.5 Males: Percentage Distribution of Activity Time by Age and Location Small Villages Baralong Farms Large Villages Activities ~~~Age Age Age Activities 7-14 Ae 15+ 7-14 Ae 15+ 7-14 Ae 15+ (Percentage Distribution of Total Time) Crop Husbandry 2.6% 6.5% 4.6% 19.3% 2.7% 6.1% Animal Husbandry 28.2 14.7 32.8 16.3 16.2 11.4 Wage Labor 0.4 5.3 0.0 2.7 0.6 10.7 Trading, Processing, Vending 0.1 0.9 0.0 1.1 0.3 2.2 Hunting or Gathering 1.4 2.4 1.0 1.1 1.7 1.9 All Income Earning Activities 32.6 29.8 38.4 40.5 21.5 32.2 Repairing, New Building 0.6 3.0 0.0 3.5 1.0 2.6 Fetching Water 1.8 1.4 1.4 2.2 3.4 1.9 Child Care 2.4 0.4 4.0 0.7 2.2 0.6 Housework 3.5 3.3 2.0 2.7 6.0 6.4 0 All Housekeeping Activities 8.3 8.0 7.5 9.1 12.6 11.4 Schooling 11.9 1.9 16.1 1.9 16.9 4.0 Illness & Health Care 1.5 4.1 3.0 3.5 1.2 4.6 Meetings 0.0 1.5 0.0 4.7 0.0 0.7 Leisure 45.7 54.7 34.9 40.4 47.8 47.0 All Non-Work Activities 47.2 60.2 38.0 48.6 49.0 57.3 All Activities 100.0 100.0 100.0 100.0 100.0 100.0 Table 3.6 Females: Percentage Distribution of Activity Time by Age and Location Small Villages Baralong Farms Large Villages Age Age Age Activities 7-14 15+ 7-14 15+ 7-14 15+ (Percentage Distribution of Total Time) Crop Husbandry 3.2% 10.5% 1.8% 11.0% 2.7% 6.6% Animal Husbandry 4.0 1.2 4.1 2.1 1.8 1.1 Wage Labor 0.5 0.9 0.0 0.8 0.5 4.3 Trading, Processing, Vending 0.4 1.9 0.0 0.9 0.0 1.6 Hunting or Gathering 2.3 2.7 2.1 2.0 1.9 2.2 All Income Earning Activities 10.0 17.2 8.0 16.8 6.8 15.9 Repairing, New Building 0.7 4.0 0.3 5.1 0.9 3.8 Fetching Water 6.0 7.1 3.3 6.1 4.8 5.4 Child Care 7.6 3.3 9.9 4.2 5.8 4.3 Housework 13.5 19.4 12.4 23.0 12.2 19.4 All Housekeeping Activities 27.9 33.9 25.9 38.5 23.6 33.0 Schooling 14.8 1.5 32.7 3.2 20.4 3.8 Illness & Health Care 1.7 5.1 2.3 8.5 1.5 7.2 Meetings 0.0 0.5 1.5 3.5 0.0 0.2 Leisure 45.1 41.7 29.6 29.4 47.6 39.9 All Non-Work Activities 46.9 47.4 33.4 41.5 49.0 47.3 All Activities 100.0 100.0 100.0 100.0 100.0 100.0 Table 3.7 Number of Days Engaged in Wage Labor Per Year by Age and Sex Age Days Worked for Wages 7-19 10-14 15-19 20-29 30-39 40-49 50-59 60+ Total Males 0 99.3% 96.0% 86.4% 68.3% 71.6% 70.8% 68.7% 83.6% 83.7% 1-29 0.0 1.4 4.3 5.3 6.9 7.1 7.0 4.7 3.9 30-99 0.7 2.1 7.0 6.1 5.5 6.0 6.2 4.5 4.4 100-179 0.0 0.2 1.2 3.4 2.9 2.4 3.2 1.0 1.5 180-259 0.0 0.1 0.8 6.9 2.9 4.5 3.2 1.2 2.1 260+ 0.0 0.1 0.3 10.0 10.2 9.2 11.6 4.9 4.5 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Females 0 99.8 95.3 87.8 81.8 78.8 79.3 82.4 92.9 87.3 1-29 0.2 0.8 3.2 4.7 8.2 7.8 8.8 3.9 4.3 30-99 0.0 2.2 6.0 6.8 7.0 8.4 6.8 1.8 4.9 100-179 0.0 1.5 2.3 1.4 2.2 1.3 0.4 1.2 1.4 180-259 0.0 0.2 0.3 2.4 1.2 0.2 0.9 0.0 0.8 260+ 0.0 0.0 0.5 2.9 2.6 3.0 0.6 0.3 1.3 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 - 43 - Table 3.8 Comparison of the Contribution of Adults and Children to Economic Work Population Above Age 6 "Economic" Time (Percentage Distribution) Males 7-19 21.5% 30.0% 20-59 16.2 24.1 60 and over 5.0 6.1 Females 7-19 24.6 12.1 20-59 27.4 23.8 60 and over 5.3 3.9 Total 100.0 100.0 - 44 - productivity of a given kind of labor throughout the rural economy. As we have seen, opportunities for wage labor are quite limited in Botswana, particularly for women and children. Only 11 percent of the work time of all members of the labor force is devoted to wage work and there is little seasonal variation in wage labor time. Data (not shown here) indicate that about half of the people who engage in wage labor are government employees, sales and construction workers, mechanics, tanners, drivers, teachers, nurses, and the like, leaving only a small group which performs agricultural labor. Lucas (Chapter 4) finds that only one percent of crop-producing households reported any expenditure on wage labor. This figure does not include expenditures on plowing services reported by 17 percent of such households, often female headed. A household which does not buy or sell labor may be called an autarchical household.9 The concept of an autarchical farm operation accords with a good deal of evidence from farm management studies in LDC showing that labor intensity and output per acre are inversely related to size of landholdings.10 Lucas (Chapter 4) finds the same to be true for crop production in Botswana. Autarchical modes of production imply that asset-poor households often use family labor to a point where marginal returns fron work are very low rather than hiring out their surplus labor to households with more resources who could use that labor more productively. This is not the place to analyze the reasons for the existence of autarchical working 9 Mark R. Rosenzweig, "Household and Non-Household Activities of Youths: Issues of Modelling, Data and Estimation Strategies," ILO Informal Workshop on Children and Employment, Geneva, 1979 (mimeographed). 10 See for example, World Bank, World Development Report, 1980, Oxford University Press, New York, 1980, p. 42. - 45 - arrangements. They may reflect a reluctance on the part of small farmers to hire out family labor, high transport costs, a view on the part of larger farmers that hired labor is less desirable than family labor, or an institutionally determined agricultural wage which is too high to clear the labor market.11 The point is that, to the extent that autarchical modes of production prevail, the household's access to productive assets and know-how rather than a market wage determine the value of time. This study will therefore differ from previous research in that it will attempt to derive at least a rough estimate of the marginal productivity of work performed by relating time iniputs into the production process to income generated by this work, rather than by assuming that prevailing wage rates measure the opportunity cost of time for all members of the households and at all times of the year. The marginal productivity of labor is of interest not only because it influences time allocations but also because it may help us to evaluate how adequately the rural labor supply is utilized. There is no doubt general agreement that work having zero marginal productivity would signify surplus labor.12 However, zero marginal productivity is a polar case and should rarely be found in practice, except perhaps where somebody works for training, to keep busy, to keep up social appearances, to keep company - as may be the 11 For a further analysis; of these issues see A. Berry and R. H. Sabot, "Labour Market Performance in Developing Countries: A Survey," World Development, Vol. 6, 1978, especially pp. 1222-28. 12 For a discussion of low or zero marginal productivity of labor in a peasant economy see W. A. Lewis, "Economic Development with Unlimited Supplies of Labour", Manchester School, May 1954; J. C. Fei and G. Ranis, Development and Labor Surplus Economy, Richard D. Irwim Inc., Homewood, Ill., 1964. - 46 - case among some children in Botswana -- or where certain activity-shariag arraagements have become imstitutiomalized. Laborers doing work of very low productivity also may be regarded as beiag uader-utilized but there is ao generally accepted criterioia for deciding how low productivity has to be to sigaify under-utilization. The amouats of leisure time reported by differeat groups ia differeat socio-ecomomic strata may be a partial iadicator of underemploymeat. Im the analysis which follows we shall look for evideace of both low margimal productivity of labor amd loag hours of leisure time as iadicators of possible uaderemploymeat. This may emable us to determime whether labor umder- utilizatiom exists, but it will not allow us to quantify its extent. Because time use data measure only work duratiom without regard to work iatensity, it is particularly important to view loag leisure hours and low productivity as two related facets of labor under-utilization whem amalyziag time use data. Self-employed people who have little work to do may work more slowly or rest more oftem than those who are pressed to accomplish a great deal. In that case the productivity of work time should iadicate uader-utilizatioa, evea if duration does aot. Based on the theoretical framework pioneered by Becker aad Willis, an individual's allocation of time to economic work may be viewed as a function of the household's unearmed income, his own labor productivity and that of other family members, together with some relevatt control variables.13 13 Gary S. Becker, "A Theory of Allocation of Time," Economic Journal, Vol. 75, Sept. 1965, pp. 493-517; Robert J. Willis, "A New Approach to the Economic Theory of Fertility Behavior," Journal of Political Economy, March/April 1973, Part II, pp. 514-64. - 47 - - Increases in labor produ,ztivity should indiuce more economic work at the expense of other activities through a price effect. At the same time increases in labor productivity should reduce work time via an income effect. The lat'ter expectation derives from the assumption that income increases raise the demand for leisure and such other "normal" commodities such as child schooling and child rearing at the expense of economic work. At best, we can observe the net impact of labor productivity through both channels of causation. This problem limits the capacity of economic theory and analysis to predict time allocations. Labor productivLty can be viewed as a predetermined variable only when generated by a compestitive labor market in which households buy and sell labor whenever marginal productivities deviate from the market wage. In a setting where autarchicalt modes of production predominate, as in the case of rural Botswana, a person's productivity depends largely on the complementary resources available in the household enterprise such as education (know-how), land, cattle, tools, etc. Moreover, the marginal productivity of labor is not observable. We shall therefore proceed in two, steps. First, we shall calculate a household enterprise proiuction function and shall derive estimates of the marginal producl:ivity of household labor, categorized by age and sex, from that function. Thus in step one we shall obtain information on the magnitude of marginal labor productivity and on the way in which time inputs, human and non-human capital and certain control variables affect labor productivity. These results have a bearing on the underemployment issue. Second]Ly, we shall estimate the allocation of time by individuals as a function of the variables considered to 'be predetermined in our analysis -- - 48 - education of adults, productive assets, and demographic characteristics of the household. In this analysis education and productive assets must be viewed as proxies for productivity, i.e., as having price and income effects. 3.4 Productivity of Time For estimation purposes income of household x (Ix) is considered net of transfers. It is assumed to be generated by a production function where age- sex-specific time T c(a, s) , human capital (Ed) and physical assets (A) are inputs, in addition to a set of controls (RV). We thus propose to estimate: ln I e a + b In Tc(a,s) + clnEd + dlnA + e RV + e. (3.1) x x x x x The time inputs of households are here condensed into three categories: total working time of males aged 15 and above, of females aged 15 and above, and of children 7-14 (regardless of sex). Children 15-19, a relatively small group, are combined with adults since their efficiency in production should come close to that of adults. Also, the time use analysis will show some substitutability between the economic work time of young boys and girls, between men and older boys, and between women and older girls. The regressions were also run with children divided by sex and with separate categories for time inputs by children 15-19. Instances where these more detailed breakdowns provide additional (or different) information will be reported in the text. Before turning to the results, some limitations of the productivity estimates must be pointed out. First, income and time allocation are codetermined, but we do not have enough explanatory variables to identify a - 49 - simultaneous equation system. The simultaneity bias may led to some underestimation of the time input coefficients. Second, land is not a major constraint on production in Botswana although quality of land may be, but we do not have a measure of land quality. Third, time use data based on a sample of only 5 days in the year do not have a Uigh degree of precision.14 For these reasons the estimated coefficients must be viewed as indicators of orders of magnitude. Since we are interested in productivity estimates for households which exhibit substantial variations in earnings opportunities, a few interactions are considered. We use first only nonwage or self employment income as a dependent variable, and exclude time spent on wage labor from the analysis. Then we add wages as part of income and wage labor as part of the time inputs and reestimate the same function. We also differentiate between households who own cattle and households who do not. A final distinction is made between households with at least one male and one female age 15-64 and households with no adult males (these constitute 29 percent of all households). The various estimates of relationship (3.1) should enable us to assess the relative productivity of time inputs of the various categories of household members under different economic and demographic circumstances. Table 3.9 presents the regression coefficients for all households which contain at least one woman and one man of prime working age with self- employment income as dependent variable. The equation for all households 14 Another problem is that Botswana has a strong tradition of mutual assistance among relatives in economic work. Thus some reported economic work may have been performed for other households. It is unlikely that such practices affect our productivity estimates to an appreciable degree since working for others in most cases involves some reciprocity. Table 3.9 Determinants of Household Income (ln Y) for Households with at Least One Male and One Female Aged 15-64, Wages and Wage Time Excluded Households Households l 1 Househelds 3 with No Cattle with Cattle Independent Variables b Beta t b Beta t b Beta t Constant .69 .92 .68 ln cattle .18 .56 17.68 .60 .67 16.74 ln smaller animals .06 .16 4.72 .06 .21 2.94 .03 .07 1.70 ln land .02 .03 .83 .08 .13 1.99 -.01 -.01 -.31 ln work time children 7-14 .01 .02 .73 -.00 -.00 -.05 .02 .08 2.18 males 15 or over .04 .09 2.65 .05 .17 2.47 .04 .10 2.56 females 15 or over .00 .00 .10 -.00 -.01 -.09 -.00 -.01 -.27 In education* .09 .07 2.21 .08 .07 1.11 .00 .00 .04 Place**: Baralong Farms .62 .10 3.38 1.04 .15 2.48 .50 .11 2.93 Large villages -.05 -.01 -.48 -.06 -.03 -.37 -.05 -.02 -.47 ln age of head -.03 -.01 -.31 -.11 -.04 -.68 -.04 -.01 -.32 R2 .52 .16 .53 N 607 204 367 Mean 5.92 5.03 6.43 Standard Deviation 1.14 .96 .90 lb is the estimated regression coefficient. 2Beta is the standardized regression coefficient. 3t is the t-statistic. **Dummy variable;small villages are the omitted category. *Highest education in the household. - 51 - explains 53 percent of the variance in income, for cattle owning households 52 percent, and for households without cattle only 16 percent. The differences between the R2's reflect the significance of cattle ownership in the determination of agricultural income in Botswana. The regression coefficients (b) represent the elasticities of income with respect to the inputs into the production function. Presumably a 10 percent increase in cattle ownership in cattle owning households would increase self-eTnployment income in these households by 6 percent. However, because of unavoidable shortcomings of the data and method, such precise inferences are not warranted. We can say ithat for all households and for cattle owning households, the value of catt:le is the major determinant of income. Ownership of smaller animals also has a significant effect on income, especially in households without cattle. Land seems to be important only in households without cattle. Education enhances self-employment income in households without cattle, but not to any great extent. Residents of Baralong Farms, the commercial farming area in the sample, appear to have an advantage in generating irLcome even among the self-employed; all other things equal, they have higher incomes than residents of other areas. As expected, adult male labor is a significant factor in the production process under all observed circumstances. Children's labor makes a significant contribution in cattle owning households only. Women's labor time does not prove to be statistically significant in explaining variation in household income. Table 3.10 presents coefficients with wages included in income, and wage work in working time. The importance of the education variable is greatly increased in this version. This result indicates that, while Table 3.10 Determinants of Household Income (ln) for Households with at Least One Male and One Female Aged 15-64, Wages and Wage Time Included Households Households All HouseVolds 3 with No Cattle with Cattle Independent Variables bl Beta t b Beta t b Beta t Constant .75 .90 .70 ln cattle .15 .49 14.25 .54 .62 15.31 ln animals .02 .05 1.47 .01 .03 .40 .01 .02 .35 ln land -.03 -.04 -1.32 .04 .06 .96 -.05 -.09 -2.24 ln work time Children 7-14 .00 .01 .22 -.02 -.06 -.96 .02 .09 2.30 Males 15 or over .07 .15 4.44 .08 .24 3.77 .06 .13 3.27 Females 15 or over -.02 -.03 -1.09 -.01 -.03 -.52 -.02 -.04 -1.14 ln education* .23 .20 5.98 .26 .24 3.81 .11 .11 2.85 Place**: Baralong Farms .61 .11 3.37 .99 .14 2.38 .48 .11 2.87 Large villages .31 .11 3.30 .36 .15 2.19 .23 .10 2.29 ln age of head -.10 -.03 -.98 -.16 -.06 -1.00 -.13 -.04 -1.05 R2 .44 .20 .52 N 607 240 367 Mean 6.19 5.50 6.58 Standard Deviation 1.05 .97 .88 is the estimated regression coefficient. 'Beta is the standardized regression coefficient. 3t is the t-statistic. **Dummy variable; small villages are the omitted category. *Highest education in the household. - 53 - education has a moderate impact on earnings in self-employment, it has a major effect on the opportunities for wage work as well as the level of wages, This particular iSSUE- is discussed in detail in Chapter 5. Location in a large village (as compared with a small village) offers no advantages for self-employment but improves access to wage work, especially the better paying wage jobs. Animals and land are somewhat reduced in importance as source of income when wage work is taken into account. The contribution of male labor to income rises substantially when wages and time worked for wages are included in the data; but this is not the case for women's and children's labor. This finding is in accord with earlier evidence showing that men do more wage labor than women, and that children do practically no wage labor. The time inputs of children in cattle owning households remain significant factors in the generation of household income; the time inputs of women and of children in households without cattle remain insignificant. Tables 3.9 and 3.10 included only households which contained a prime working age male (15-64). Table 3.11 presents a similar analysis for female headed households without prime working age males. Most of the resu:Lts parallel those appearing in Tables 3.9 and 3.10. The greatest difference appears in the estimated contribution of time inputs of adult romen to household income. In female headed households the economic workc time of women has a significant effect on household income, especially when wrage work is included. The work contribution of children is insignificant for the most part, probably because female headed households own fewer animals than households with male heads. Education has a smaller positive impact in female headed households than in others because female wage - 54 - Table 3.11 Determinants of Household Income (ln Y) in Female Headed Households (with no male aged 15-64) Wages and Wage Wages and Wage Time Excl4ded 3 Time Included Independent Variables b Beta t b Beta t Constant .67 .68 ln cattle .18 .53 8.88 .16 .49 8.20 ln animals .08 .28 4.40 .08 .29 4.49 ln land -.02 -.03 -.52 -.06 -.10 -1.72 ln work time Children 7-14 .02 .05 .93 -.02 -.07 -1.13 Males 15 and over Females 15 and over .04 .09 1.66 .07 .17 3.27 ln education* .03 .02 .42 .17 .16 2.91 Place**: Baralong Farms .68 .05 1.02 1.10 .09 1.77 Large villages .18 .06 1.12 .49 -.20 3.33 ln age of head -.39 -.12 -2.30 -.58 -3.82 R2 .55 .53 N 200 200 Mean 5.02 5.29 Standard Deviation 1.11 1.02 lb is the estimated regression coefficient. 2Beta is the standardized regression coefficient. 3t is the t-statistic. *Highest education in the household. ** Dummy variable; small villages are the omitted category. - 55 - rates are less responsive to educational attainment than males' wage rates (as discussed in Chapter 5), and opportunities to obtain wage jobs also are more limited for women. As noted earlier, Baralong Farms do not offer the same economic opportunities in commercial crop production to women as to men. Also there is a stronger negative relation between the age of females heading households and lhousehold income than between the age of males heading households and 'household income. A number of alternative forms of the production function were calculated, including a linear form with and without interactions between hours worked and assets, separate equations for high, middle and low income households, difEerent age divisions, household composition groups, and the like. Such experimentation is justified as long as the results are viewed as a kind of sensitivity analysis. The findings regarding the relation of time inputs to income were found to be quite robust. Hours of adult male labor always have a significant positive relation to income. Child labor inputs are significant in cattle owning households only. When boys' and girls' working hours are treated as two separate inputs, only the work contributions of boys in cattle owning households appear statistically significant. The regression coefficient for inputs of working time by adult women is significant only in female headed households.15 15 In a similar study in rural Niger, based on much more detailed time use data for each agricultural operation and each field, translog production equations were estimated with child labor inputs included. The results indicate that the effect of child labor (8-14 years old) upon agricultural output levels was not significantly different from zero. However, female labor was infinitely substitutable for male labor and had a positive marginal productivity. See Randall Stuart Thomas-Peterhans, The Stratification of the Production and Marketing of Livestock in South- eastern Niger in the Department of Zinder (unpublished Ph.D. dissertation, University of Michigan, 1982, Chapter IV). - 56 - Marginal productivities of work time were calculated from the regressions in Tables 3.9-3.11. They may be expressed in pula per day, assuming a 6-hour work day on the average. On this basis the marginal productivity of male labor comes to .18-.22 Pula per day, and the marginal productivity of female labor is slightly lower. These figures can be compared with daily wage rates to give the reader some perspective on the orders of magnitude involved. As shown In Chapter 5, at the time of the survey the average daily wage in rural areas was 1.26 Pula for men without education and 1.50 Pula for those with 1-4 years of schooling; it was .54 Pula for women with 0-4 years of education. Thus the marginal productivities are for the most part much lower than the daily wage.16 This is not surprising, considering the very limited opportunities for wage work and the widespread poverty. At the time of the survey the average daily income per person In rural areas was about .35 Pula. Thus it is plausible that men would find it worth while to pursue work in their own enterprises which at the margin yielded the equivalent of only .18-.22 Pula a day. To summarize this section, it appears that non-human capltal and to a lesser extent human capital are critical in the determination of productivity levels. In asset-poor households people are forced to undertake work of very low productivity. Particularly, women and children in such households seem to pursue income-earnings activities up to a point where the marginal return to their labor is close to zero. Because men have the option to migrate and male labor is relatively scarce, the marginal productivity of male labor is higher, even in asset-poor households. 16 The same result was obtained in Niger. See Thomas-Peterhans, ibid., Chapter IV. - 57 - 3.5. Determinants of Time Use The determinants of time use were estimated by means of regression analysis. The dependent variable for the analysis of time allocation is tij, the amount of time spent by individual i on activity j. Four categories of activities are considered: (1) economic work, (2) housework and child care, (3) schooling for age groups 7-19, and (4) leisure. Housework and child care are combined since they are often concurrent activities. The unit of measurement for time allocation is total number of minutes spent on each activity category summed over the five days. Wage labor is measured, however, in days per year for reasons to be explained later. The independent, exogeneous variables include human and non-human assets, and demographic characteristics of the household. These variables are entered additively into the estimating equation: t - a + bEdi+ cEdx + dAx + eIux +fRVx c (3.2) By estimating the linear model and using the same explanatory variables for each of the four time uses, one can canpare the regression coefficients across equations to observe possible substitutions between one time use and another. 17 As noted earlier, asset holdings act in the regressions as proxies for both earned income and price effects. The physical assets (Ax) included 17 Total activity time doeta not add to the same number of hours in each household, since eating and sleeping times were omitted. However, there is a fairly narrow dispersion around a mean of 12 hours. - 58 - are value of cattle, value of small animals, and area of land cultivated.18 Data on equipment owned were collected, but they are too highly correlated with other assets to be useful in the analysis. The ownership of human capital is represented in this section by two measures: (1) the education of the person whose time allocations is being analyzed (Edi) and (2) the highest education among adults in the household (Edx). Transfer income (Iux) are the only directly measured source of income that is entered into the equation as a predetermined variable. Household size is represented by a series of variables reflecting the size of each age/sex group. This permits different kinds of household members to influence time allocation differently. The remaining independent variables control for age and village characteristics. Separate equations are estimated for males and females 20-64. Children 7-19 also are divided by sex since the sexual division of labor leads boys and girls to pursue quite different activities. Young people aged 15-19 are here grouped with children rather than adults since, like children (1) they do very little wage labor, and (2) they still show a fair amount of school attendance. Initially separate regressions were run for age groups 7-14 and 15-19 but findings were quite similar. Hence for the sake of brevity, the younger and older children of each sex are combined, with a control for age among the explanatory variables. The results for adult men are presented in Table 3.12, for adult women in Table 3.13, and for boys and girls in Tables 3.14 and 3.15, 18 The data on land holdings are weak; for about one-third of households it cannot be determined whether they had no land or whether they had land but the area was unknown. Further, no information is available on land quality or value. It is quite possible that larger holdings are of poorer quality than smaller holdings. Table 3.12 Determinants of Individual Time Use, Males Aged 20-64 Nulaber of Days Worked Economic Time Other for We es Than Wage Work Housework Leisure Explanatory Variables bL Betal t3 b Beta t b Beta t b Beta t Constant 157.07 115.87 128.24 1898.85 Value of Cattle (in P)a 50-999 -22.86 -.09 -2.41 165.06 .09 2.37 36.22 .03 .81 47.46 .02 .51 11300-2499 -18.02 -.07 -1.87 215.80 .13 3.04 2.42 .00 .05 -144.50 -.08 -1.76 2500+ -41.50 -.14 -3.61 169.63 .08 2.00 13.19 .01 .24 137.59 .05 1.23 Value of Other Animals (in p)aa 1-49 -51.10 -.20 -4.00 140.58 .08 1.50 -92.80 -.09 -1.55 79.82 .04 .64 50-199 -68.73 -.31 -5.38 341.60 .23 3.o3 jO.24 .03 .51 -62.71 -.03 -.50 200+ -64.54 -.26 -4.62 349.61 .20 3.40 -19.23 -.02 -.29 -111.45 -.05 -.82 Land (Acres of Holdings)aaa 2-9 -:33.79 -.16 -3.93 167.10 .11 2.49 94.66 .10 2.22 84.63 .04 .96 10-19 -'50.50 -.19 -4.80 37.67 .02 .49 149.11 .13 3.02 -30.69 -.01 -.30 20+ -44.10 -.16 -3.98 94.01 .03 1.15 195.79 .17 3.77 -291.64 -.12 -2.70 Place- Baralong Farms -12.98 -.02 -.75 401.40 .11 3.15 -131.88 -.06 -1.62 -253.27 -.06 -1.50 Large Villages :13.86 .08 2.56 -45.69 -.02 -.67 145.34 .12 3.33 -245.42 -.10 -:2.71 Own Education 1.57 .12 3.75 -6.46 -.07 -2.11 -.83 -.02 -.42 -3.55 -.03 -.87 Highest Education 3.47 .23 6.55 -18.01 -.11 -2.94 -5.61 -.06 -1.43 -6.46 -.03 -.80 Age -.04 -.00 -.02 5.18 .09 .38 3.99 .12 .46 21.89 .31 1.22 Age Squared -.004 -.04 -.20 .00 .00 .02 -.06 -.16 -.62 -.30 -.34 -1.37 Baby Present -1.37 -.01 -.19 31.61 .02 .59 75.39 .08 2.21 -58.26 -.01 -.82 Household Members (no.) Boys 7 -14 5.68 .05 1.64 -60.34 -.08 -2.36 -12.97 -.03 -.80 61.86 .07 1.83 Girls 7-14 -.84 -.01 -.25 20.41 .03 .83 -24.42 -.06 -1.56 -10.68 -.01 --.33 Boys 15-19 *-4.88 -.03 -.86 27.67 .02 .64 30.59 .04 1.11 -75.90 -.05 -1.33 Girls 15-19 -.98 -.01 -.17 108.68 .09 2.57 -32.81 -.05 -1.22 7.08 .00 .13 Men 20-64 -1.83 -.02 -.50 -10.52 -.01 -.39 44.29 .10 2.56 4.28 .00 .12 Wom,en 20-64 -5.85 -.04 -1.68 49.92 .07 1.95 -20.07 -.05 -1.23 -71.61 -.08 -2.11 Men 65+ -40.71 -.11 -3.48 306.96 .13 3.54 -75.22 -.05 -1.37 9.42 .00 .08 Women 65+: -.93 -.00 -.10 14.81 .01 .21 -69.85 -.05 -1.54 -52.63 -.02 -.56 Sex c,f Head -31.74 -.11 -3.37 83.09 .05 1.20 54.41 .05 1.23 -75.77 -.03 -.83 Transfer Income -.07 -.07 -2.06 -.22 -.03 -.85 -.05 -.01 -.28 .84 .09 2.43 B,2 .27 .16 .09 .10 H1 846 846 846 846 Mean Time 50.27 836.74 312.65 1978.65 Standard Deviation [07.99 740.33 454.02 944.70 1. b is the estimated regression coefficieat. 2. Beta is the standardized regression coefficient. 3. t is the t-statistic. a. Categorical variables omitted category under P 50. aa. Categorical variables omitted category is 0. aaa. Categorical variables omitted category is 0-1. - 60 - Table 3.13 Determinants of Individual Time Use, Females Aged 20-64 Number of Days Worked Economic Time Other for Wages Than Wage Work Housework Leisure Explanatory Variables b' Betas t3 b Beta t b Beta t b Beta t Constant 21.13 -286.30 1734.71 1551.29 Value of Cattle (in P) 50-999 -12.41 -.07 -2.50 18.54 .01 .47 -19.75 -.01 -.38 -62.28 -.03 -1.11 1000-2499 -12.48 -.07 -2.28 52.86 .04 1.22 -41.50 -.02 -.73 62.52 .03 1.01 2500+ -16.78 -.08 -2.39 59.73 .04 1.08 36.76 .02 .50 44.77 .02 .57 Value of Other Animals (in P) 1-49 -5.29 -.03 -.75 41.87 .03 .75 -10.95 -.01 -.15 40.69 .02 .51 50-199 -10.56 -.07 -1.47 7.89 .01 .14 65.60 .04 .88 105.85 .06 1.31 200+ -18.31 -.11 -2.28 103.08 .08 1.62 -1.39 -.00 -.02 89.99 .05 .99 Land (Acres of Holdings) 2-9 -13.91 -.09 -2.84 72.11 .06 1.86 -19.22 -.01 -.38 107.08 .06 1.94 10-19 -14.07 -.08 -2.49 96.71 .07 2.16 140.37 .07 2.38 -154.71 -.08 -2.43 20+ -11.20 -.06 -1.78 77.74 .05 1.56 192.87 .10 2.94 -289.61 -.13 -4.08 Place- Baralong Farms 12.99 .03 1.10 -136.24 -.04 -1.46 18.01 .00 .15 -276.20 -.06 -2.08 Large Villages 10.22 .05 1.91 -187.92 -.12 -4.43 -9.86 -.00 -.18 -59.15 -.03 -.98 Own Education .36 .04 1.63 -.41 -.01 -.23 -2.43 -.03 -1.05 -1.80 -.02 -.72 Highest Education 2.66 .18 5.73 -3.56 -.03 -.97 -20.43 -.13 -4.22 2.14 .01 .41 Age .84 .15 .81 32.29 .75 3.97 -14.48 -.25 -1.35 -.47 -.01 -.04 Age Squared -.01 -.22 -1.16 -.37 -.68 3.62 .06 .08 .43 .06 .08 .45 Baby Present -1.79 -.01 -.45 -57.49 -.05 -1.81 101.87 .07 2.43 -82.64 -.05 -1.83 Household hembers (no.) Boys 7 -14 -3.02 -.04 -1.48 -22.06 -.04 -1.37 55.63 .08 2.62 -53.43 -.07 -2.33 Girls 7-14 4.78 .08 2.74 20.85 .04 1.51 25.78 .04 1.42 -65.50 -.09 -3.33 Boys 15-19 .12 .00 .04 2.42 .00 .09 38.55 .03 1.12 -3.46 -.00 -.09 Girls 15-19 -7.21 -.07 -2.37 65.01 .08 2.70 -101.18 -.09 -3.19 63.62 .05 1.86 Men 20-64 -2.91 -.04 -1.24 12.22 .02 .66 -13.71 -.02 -.56 15.89 .02 .60 Women 20-64 -.56 -.01 -.29 69.69 .14 4.59 -32.42 -.05 -1.62 -40.68 -.06 -1.88 Men 65+ -3.80 -.02 -.67 124.44 .04 2.76 -40.41 -.02 -.68 -9.76 -.00 -.15 Women 65+ 4.07 .02 .80 11.48 .01 .29 75.74 .04 1.44 -50.30 -.02 -.88 Sex of Head 14.02 .10 2.89 44.69 .04 1.17 9.15 .01 .18 -75.49 -.05 -1.38 Transfer Income .01 .01 .39 -.15 -.03 -1.00 .11 .01 .53 .44 .06 2.04 R2 .11 .08 .09 .09 N 1370 1370 1370 1370 Mean Time 18.03 595.62 1268.25 1464.69 Standard Deviation 70.80 551.58 730.30 791.48 1. b is the estimated regression coefficient. 2. Beta is the standardized regression coefficient. 3. t is the t-statistic. - 61 - Table 3.14 Detenminants of Individual Time Use, Boys Aged 7-19 School Economic Work Housework Leisure Explanatory Variables b' Beta' t' b Beta t b Beta t b Beta t Constant -376.54 -1333.38 788.86 4549.55 Value of Cattle (in P) 50-999 -53.74 -0.04 -1.1 222.16 0.12 3.4 -85.35 -0.08 -2.1 -37.57 -0.02 -.6 1000-2499 52.18 0.04 1.0 245.79 0.13 3.6 -34.09 -0.03 -.8 -109.16 -0.06 -1.6 2500+ -0.50 -0.00 -.01 359.19 0.15 4.0 -41.49 -0.03 -.7 -186.18 -0.08 -2.0 Value of Other Animals (in P) 1-49 -134.57 -0.09 -1.8 366.85 0.19 3.7 -140.22 -0.13 -2.3 7.98 0.004 .1 50-199 -158.15 -0.12 -2.1 481.19 0.28 4.9 -247.88 -0.25 -4.2 16.51 0.01 .2 200+ -311.55 -0.21 -3.8 458.39 0.23 4.3 -316.08 -0.28 -4.8 223.00 0.11 2.1 Land (Acres of Holdings) 2-9 -129.91 -0.10 -2.6 2i6.13 0.12 3.2 -62.99 -0.06 -1.5 11.28 0.01 .2 10-19 132.77 0.08 2.2 20.36 0.01 .3 -18.95 -0.01 -.4 -162.80 -0.07 -2.0 20+ -82.19 -0.05 -1.4 137.64 0.07 1.8 66.86 0.06 1.5 -335.83 -0.17 -4.4 Place- Baralong Farms 79.15 0.02 .7 166.38 0.04 1.1 31.29 0.01 .4 -402.48 -0.09 -2.7 Large Villages 115.69 0.07 2.2 -158.35 -0.07 -2.3 104.23 0.08 2.5 -89.64 -0.04 -1.3 Own Education 17.75 0.18 5.8 -13.03 -0.10 -3.3 0.63 0.01 .3 -4.35 -0.03 -1.1 Highest Education 23.09 0.17 5.0 -33.70 -0.18 -5.5 11.96 0.11 3.2 -3.15 -0.02 -.5 Age 139.21 0.74 3.4 298.30 1.19 5.5 -35.81 -0.25 -1.1 -425.87 -1.74 -7.7 Age Squared -6.47 -0.88 -4.0 -10.73 -1.09 -5.0 1.55 0.28 1.2 16.40 1.70 7.5 Baby Present 33.37 0.03 .9 9.77 0.01 .2 20.51 0.02 .7 -57.55 -0.03 -1.1 Houseahold Members (no.) Boys 7 -19 50.34 0.09 2.9 -43.39 -0.06 -1.9 -8.55 -0.02 -.6 5.71 0.01 .2 Girls 7-19 19.55 0.04 1.2 33.05 0.05 1.6 -23.85 -0.06 -1.9 -42.13 -0.06 -2.0 Men 20-64 21.39 0.03 .9 -21.04 -0.02 -.7 10.69 0.02 .5 -16.50 -0.02 -.5 Women 20-64 8.30 0.01 .4 85.58 0.10 3.3 -16.76 -0.04 -1.1 -79.80 -0.10 -3.1 Men 65+ -177.14 -0.09 -3.0 220.16 0.09 2.8 -121.40 -0.09 -2.5 85.38 0.03 1.1 Women 65+ 91.38 0.05 1.8 13.85 0.01 .2 -37.66 -0.03 -.9 -75.13 -0.03 -1.1 Sex of Head 8.75 0.01 .2 -121.73 -0.07 -1.9 -22.07 -0.02 -.6 152.21 0.08 2.3 Transfer Income -0,005 -0.09 -2.7 0.004 0.05 1.8 0.001 0.03 .9 0.000 0.005 .2 112 00.19 0.21 0.09 0.15 N 100l 1001 1001 1001 Mean Time 419 1122 343 1653 Standard Deviation 63ti 849 482 832 1. b is the estimated regression coefficient. 2. Beta is the standaLrdized regression coefficient. 3. t is the t-statistic. -62 - Table 3.15 Determinants of Individual Time Use, Girls Aed 7-19 School Economic Work Housework Leisure Explanatory Variables b' Betas t3 b Beta t b Beta t b Beta t Constant -8D8.51 -63.40 684.68 3515.58 Value of Cattle (in P) 50-999 -14.85 -0.01 -.3 -95.15 -0.08 -2.4 139.09 0.08 2.3 57.23 0.03 .9 1000-2499 79.41 0.05 1.5 -40.52 -0.03 -.9 8.67 0.00 .1 50.83 0.03 .7 25004- 77.62 0.04 1.2 -180.32 -0.12 -3.2 162.36 0.07 1.9 63.00 0.03 .7 Value of Other Animals (in P) 1-49 -23.78 -0.01 -.3 -22.79 -0.02 -.4 -45.84 -0.03 -.5 157.60 0.09 1.7 50-199 9.76 0.01 .1 111.12 0.11 1.8 -31.62 -0.02 -.3 71.84 0.04 .7 200+ -70.65 -0.05 -.9 178.32 0.15 2.7 -168.34 -0.10 -1.7 188.59 0.10 1.8 Land (Acres of Holdings) 2-9 32.96 0.02 .7 -6.77 -0.01 -.2 -38.56 -0.02 -.6 -5.65 -0.00 -.1 10-19 151.35 0.09 2.8 -103.71 -0.08 -2.3 107.32 0.06 1.6 -171.34 -0.08 -2.4 20+ 69.06 0.04 1.2 -25.06 -0.02 -.5 69.74 0.04 1.0 -213.84 -0.10 -2.9 Place- Baralong Farms 458.07 0.12 4.3 -154.39 -0.05 -1.7 21.33 0.00 .2 -511.31 -0.11 -3.6 Large Villages 157.36 0.09 3.2 -102.28 -0.08 -2.5 -135.55 -0.07 -2.2 46.81 0.02 .7 Own Education 8.17 0.09 3.2 1.88 0.03 .9 -1.50 -0.01 -.5 -5.40 -0.05 -1.6 Higheat Education 22.90 0.18 5.5 -5.46 -0.06 -1.6 -20.69 -0.14 -3.9 -3.73 -0.02 -.7 Age 226.82 1.31 6.2 65.94 0.48 2.2 -10.44 -0.05 -.2 -256.02 -1.15 -5.3 Age Squared -10.33 -1.52 -7.3 -1.45 -0.27 -1.2 2.05 0.25 1.2 8.25 0.95 4.4 Baby Present -15.14 -0.01 -.4 -10.93 -0.01 -.4 89.95 0.06 1.9 -73.11 -0.04 -1.5 Household Members (no.) Boys 7 -19 27.27 0.05 1.9 -42.44 -0.11 -3.5 39.87 0.07 2.2 -36.32 -0.05 -1.9 Girls 7-19 -15.78 -0.03 -1.2 -18.55 -0.05 -1.7 29.39 0.05 1.8 5.52 0.01 .3 Men 20-64 -57.99 -0.09 -2.7 55.17 0.10 3.1 4.53 0.01 .2 -14.66 -0.02 -.5 Women 20-64 34.47 0.06 1.9 -15.58 -0.03 -1.0 25.64 0.04 1.1 -49.60 -0.07 -2.0 Men 65+ -139.54 -0.08 -2.5 -2.61 -0.00 -.1 38.34 0.02 .5 136.15 0.06 1.8 Women 65+ 36.32 0.21 .8 75.90 0.06 1.9 -227.65 -0.12 -3.9 82.28 0.04 1.3 Sex of Head -30.22 -0.02 -.7 -9.83 -0.01 -.3 173.73 0.11 3.0 -119.91 -0.07 -2.0 Transfer Income 0.00 0.01 .2 -O.OU -O.ul -.4 0.OU 0.04 1.4 0.00 -0.01 -.3 R2 0.20 0.11 0.10 0.13 N 1146 1146 1146 1146 Mean Time 492 394 1058 1551 Standard Deviation 634 505 759 813 1. b is the estimated regression coefficient. 2. Beta is the standardized regression coefficient. 3. t is the t-statistic. - 63 - respectively. School tin.e is not shown in the first two tables since adults rarely go to school in rural Botswana; at most they receive some religious or agricultural instruction. Mean education time is only 24 minutes for men and 23 minutes for women over a span of five days. For adults, economic work is divided into two components, wage work and non-wage economic time, because we expect the independent variables to have quite different relations to these two sub-categories. Wage time is measured by the more complete series on wage labor (obtained in 12, rather than 5 rounds) and refers to days worked during the entire year instead of minutes worked on the five sample days. Since children rarely work for wages, the wage equations are omitted in Tables 3.14 and 3.15. The education coefficients show that the more educated males spend more time in wages labor and less in self-employment than males with less education. This result is consistent with our earlier finding that education enhances productivity in wage employment much more than productivity in self- employment. The effect on adult females is similar, except that women reduce housework in favor of wage employment. In some cases servants may help with the housework in the most educated households. For adults of both sex, leisure time is not significantly influenced by education. We may infer that the negative effect of education on leisure via higher labor productivity and the positive effect via hLgher income roughly offset one another. Boys who spend more time in school devote less time to economic work, while girls do letss housework; children of both sexes also reduce their leisure time due to schooling. The findings that in the more educated - 64 - households boys do more housework probably implies that boys substitute to some extent for their sisters who also go to school.19 In total, children's economic work is more significantly reduced by school attendance than are housework or leisure, although the latter activities are curtailed appreciably. The higher the level of livestock ownership, the more time adult males devote to farming and the less time they devote to wage employment. Large landholdings induce only a small increase in male time spent on farm activities, since crop cultivation is largely "women's work." In the case of women, total economic work requires more time in households with large landholdings than in those with smaller holdings. Landholdings and large herds of smaller animals reduce wage time. The finding that large landholdings also are associated with larger time inputs by women into housework is puzzling, especially since the same relationship is evident for men. We suspect that some borderline economic work, such as crop processing and garden cultivation, is misclassified as housework. Leisure time of males is quite insensitive to holdings of productive assets, as it is to education. Again we infer that price-of-time and income effects offset one another. Leisure time of women, on the other hand, is reduced on balance when landholdings are relatively large. As predicted by the earlier productivity analysis, cattle ownership has a very strong positive effect on the economic time inputs of boys 15-19 years of age, somewhat weaker for the younger boys (not shown separately in 19 Tables 3.14 and 3.15 show somewhat higher mean school hours for girls than for boys. See also Ch. 6. - 65 - Table 3.14). The economic work time of younger boys is positively and significantly correlated with holdings of smaller animals, which are typically herded by them. For boys the productivity effect of animal ownership clearly outweighs the income effect. The economic work time of girls also is positively related to ownership of smaller animals, for which they sometimes help to care. Since girls do not herd cattle, only the negative income effect of cattle ownership on work time is evident. However, in cattle owning households girls do more housework, thereby relieving their brothers from domestic obligations. The reLation between landholdings and children's work is weaker and less consistent in sign than is the case for animal holdings. In the case of girls, larger landholdings seem to be associated with more housework, presumably freeing the mother for crop cultivation. In this indirect way, large landholdings enhance the economic role of girls and shorten their leisure hours. Schooling time of boys is curtailed very significantly by ownership of smaller animals, while large cattle holdings signify high income and hence a stronger positive income effect on schooling. Since the marginal productivity of girls in economic work is relatively low, it is not surprising that assets tendi to be positively reLated to school attendance (via the income effect), although this -relationship is not very robust. Transfer incomees are normally fassociated with a pure income effect. This is not to be expected in Botswana since a substantial part of transfer income comes frcm adult maLes who are away for work in South Africa or the towns of Botswana. Thus there may be some selectivity bias in that households with low labor requirements are more likely to have outmigrants. On the other - 66 - hand, a "displacement effect" may occur whereby all other things being equal, family members who are left behind must substitute in economic or household work for the absent member. The estimated coefficients indicate that, as the level of transfer incomes rises, economic work by adult men as well as women decreases and leisure increases. By contrast, the displacement seems to affect boys; as transfer incomes rise, boys do more economic work, apparently at the expense of schooling. We turn now to the variables which reflect the household's demo- graphic characteristics. Age is measured in the equations by actual age and the square of its value. For adult males we cannot detect any age pattern in time allocations from our cross-sectional data. In contrast women's economic work increases up to about age 42, as child rearing gradually occupies less time. After 42, aging reduces market work and household work and increases leisure. Not surprisingly children's economic time increases with age and then levels off, while the opposite is true for leisure time. School time peaks around age 11-12. A baby in the household is expected to put added demands on the mother's housework time (which includes childcare) and that of other house- hold members who can assist her. Indeed, the amount of time allocated to housework by each woman aged 20-64 is raised modestly when a baby is present, while leisure and economic work are reduced. However, as we just noted, an additional effect of childrearing on women's time allocations seems to be captured by the age variable. Girls likewise spend more time on housework and have less leisure when there is a baby in the household. Men's leisure time is curtailed somewhat under these circumstances, while housework time and economic work time rise (the latter not significantly). - 67 - In femlale headed households women spend more time on economic work than do women in male hetaded households, largely at the expense of leisure. Female headed hLouseholds have lower incames than male headed households and, as we saw earlier, the lack of male labor in these households raises the marginal productivity of women's work. To some extent women substitute for the missing male labor. These circumstances help to explain the longer working hours of women in these households. Table 3.13 also shows that female headedness is associated with more wage labor by women, probably a consequence of the low level of resources for self-employment (perhaps not fully captured by our asset measures). Boys in female headed households do less economic work and have more leisure than boys in male headed households, in full consistency with the relatively limited opportunities for self-employment in these households. Girls, on the other hand, do more housework and have less leisure time, apparently substituting for ,their mothers. Household composition effects may be measured by the number of house- hold members in each age/sex category or by a single variable representing household size. These measures were used Ln alternate equations, but only results based on the more detailed measure are shown in the tables. Regardless of measure, three findings stand out: (1) The larger the household, the less leisure time its members have. This result is consistent for men and women aged 20-64; and children 7-19 of both sexes and is statistically significant for each age/sex category. It suggests that in larger households, which tend to have more dependent members (young children and old people), have to work harder to provide sufficient income and household services. It also suggests that after controlling for assets, there are no important economies of scale in household size. (2) The larger the - 68 - household, the more time each man, woman, and boy devotes to work in the family enterprise, but not to wage work. This finding is closely related to the previous one and probably has the same explanation. It also reinforces other indications that wage work is not readily available to households which (may need it. (3) The larger the household, the greater is the specialization ,'f children's functions along sex lines. In the larger households, boys show more school attendance and more economic work but less household work while girls do more housework and less economic work. In the smaller households, where the labor supply from children may not be well-balanced by sex, there is more flexibility in the sexual divisions of labor. The variables which represent number of members in each age/sex group provide some further insights into substitutability between household members. For example, when there are girls 15-19 in the household, they share the housework with their mothers. The consequent reduction in housework per adult woman frees up time for more economic work. On the other hand, children under 15, young males 15-19, and to a lesser extent, elderly people increase domestic obligations per woman. Boys are substitutes for each other and, to some degree, for men. The reduction in market work per boy which occurs as the number of boys and adult males in the household increases is balanced by an increase in school time. Apparently the demand for boys' labor, although substantial, is limited by the household's asset holdings so that families with more than one boy may send one of them to school while one or two others care for animals and perform other chores. Economic work per girl also decreases as the number of siblings increases leaving each girl more time for housework and schooling. This finding confirms the view that there is a limited demand for economic - 69 - work by children, and that girls are substitutes in economic work for each other and for boys. As regards location, men and women in larger villages appear to have more opportunitiLes for wage labor than their counterparts in smaller villages. Men report more agricultural work in Baralong Farms than elsewhere at the expense of household time and leisure. Place of residence also has some impact on the time allocation of children. School attendance is greater in Baralong Farms and larger villages than in smaller places for both boys and girls. In the Baralong Farms area boys (like men) are heavily involved in agricultural work. As a result of this work together with relatively high school attendance, leisure hours are substantially shortened. In the larger villages there seems to be less economic work available for children than in the smaller ones, so that more frequent school attendance does not impinge significantly on leisure time. 3.6. Summary and Conclusions Time us,e data are an interesting new source of information about household behavior. At the beginning of thLs paper a number of issues were identified on which time use data might throw some light. These issues concern income generation, employment, and the contribution of women and children to household welfare. Findings related to these topics are summarized in this section, along with some conclusions regarding the potential value of time use data. Time use! data arte of value first of' all for descriptive purposes. As we saw, they tell a good deal more than conventional employment statistics do about the uses which peopl.e make of their time in rural areas, about the - 70 - various income earning activities by which they piece together a living, and about variations in activity patterns by age and sex. By comparing time use with income data, one can learn how time devoted to various economic activities compares with the income earned from these activities. Our time use data also provide information on the extent of seasonality in farm work; such information can seldom be derived from conventional employment statistics. Further, the descriptive data show how leisure, housework, and schooling compete with economic work for household time. The multivariate analysis of time use data presents some rather serious methodological problems which stem from the codetermination, on the one hand of productivity, time allocation, and incomes, and on the other, total time allocated to particular activities by the household and the division of labor among its members. This problem is compounded when the household is close to an autarchic economic unit as appears to be the case for many households in rural Botswana. We attempted to circumvent these problems by looking separately at the productivity of time in conjunction with household income and at the determinants of time use by different groups of individuals and under different economic circumstances. On a theoretical level, our findings are consistent with a number of hypotheses set forth earlier in this chapter. These hypotheses imply that time allocations are influenced by economic incentives, i.e., income and productivity effects. To be sure, our data did not permit a statistical separation of income and productivity effects, which usually operate on time use in opposite directions. Yet, in a number of instances where one effect could be assumed to be weak relative to the other, the stronger effect showed the expected impact on time allocations. - 71 - The anLalysis of time allocations clearly shows that time devoted to economic work, comprising self-employment and wage-employment, responds positively to the household's human and non-human capital. The more productive capital the hLousehold has, the more economic work its members, particularly its male children, perform. Thus the productivity effect of capital on economic work outweighs its income effect. And even though asset- poor households are forced by their low income to engage in work of low productivity, there is no evidence of a backward bending supply curve of labor, i.e., of the income effect outweighing the productivity effect. At the same time we found strong evidence that in rural Botswana time allocations are constrained by a culturally determined division of labor by age and sex. In order to adhere to this division of labor, households may adjust their mix of assets in accordance wlth the available labor supply; for example, households without adult males seldom raise cattle. Also, household composition is quite fluid in rural Botswana and may be adjusted to match asset holdings. Such chianges over time are known to occur but could not be documented on the basis of our cross-sectional data set. However, we did see evidence of some flexibility in the sexual division of labor. When households have an unusual demand for the economic work time of a particular kind of labor, say adull: females, other household members will substitute at the margin for that person in housework and economic work. Likewise, when a household lacks a particular category of latbor, say adult males, other household members will porform some of the work that would normally be assigned to the missing category of labor., In sume, the RIDS data show that time allocationsi are subject to traditional norms and at the same time are responsive to scoe degree to income and productivity effects. - 72 - Our estimates of the marginal productivity of work time can only be viewed as approximations since data quality and statistical procedures are subject to a number of reservations which have been discussed at some length. The robustness of the conclusions nevertheless suggests that our major findings are valid. In Botswana the marginal productivity of work time in rural self-employment is very low. People with small holdings of productive assets may be forced by their poverty to pursue some work which adds only minimally to income. They may also slow their work pace in accord with the available time. The marginal productivity of some time inputs by children are close to zero (although average productivity is no doubt positive). The productivity of adult male labor, of women's labor in female headed households, and of children's labor in cattle raising households is positive and significant at the margin, although quite low. This general result has important implications for employment policy even if the calculated regression coefficients andImarginal productivities are not precise. Labor underutilization is a major issue in Botswana. It is difficult to define the concept of surplus labor, especially in the case of women and children. We have used a purely operational concept, viewing as "surplus" any labor which does not make a statistically significant marginal contribution to household income. We have also compared the amounts of leisure people have under various economic and demographic conditions. Our findings are consistent with Lipton's conclusion that there is substantial surplus labor in the rural areas of Botswana. The incidence of under-employment varies, however, by age, sex, and asset position. The inference that there is underutilized adult male labor rests on several findings. First, we saw that reported leisure time for adult males is - 73 - higher than the leisure time reported for uromen and for children 10-19; adult males have as much leisure as children 7-9 years of age. Second, adult males work 60 percent longer during the busy season than during the slack season, and almost the entire seasonal differential in working time is balanced by leisure. Third, adult males work about a third longer in the commercial Baralong Farms area than in other places. And finally, we saw that the marginal productivity of male labor is much lower than their market wage, suggesting that some men are forced to undertake work which contributes only very modestly to household income. This may be the case particularly in the slack season and for men in asset-poor households. It must be recognized that the advantaged position of adult males with respect to ,Leisure may represent one of several facets of male privilege in the Botswana culture. Therefore one muslt be cautious in inferring that male labor is underutilized. Still it is unlikely that cultural norms regarding time use will survive if they are grossly inefficient economically.20 Of course, the hl'gh male migration rate out of rural areas also suggests that there is a surplus of male labor. However, we do not find a surplus of male labor in the Lewis or Fei and Ranis sense, where outmigration leaves rural production unaffected. The opportunity cost of men's time in rural Botswana may be quite substantial during the busy season and seems to be positive, though low, at other times. Moreover, the absence of male labor handicaps the income earning effort of female headed households, as has been shown elsewhere.21 20 This presumption is discussed by Gunnar Myrdal, Asian Drama. 21 Kossoudji and Mueller, op. cit. - 74 - While out-migration removes much of the surplus of male labor from the countryside, the migration rate of women is only about one-fourth that of men. The sexual division of labor and child rearing obligation prevent women's labor from being a close substitute for male labor. We found no evidence in numerous formulations of the productivity analysis that the marginal productivity of women is positive and significant except in female headed households. In the slack season women work only about a third as much as in the busy season; yet their housework time barely increases at all. It appears that even in the busy season women are not so hard-pressed that they are forced to cut corners on housework and child care. If both housework and market work are taken into account women work about 25 percent longer than men. Yet, considering the segmentation of the labor market and the low productivity tasks assigned to women, the labor of women may be even less effectively utilized than the labor of men. About 70 percent of the increase in women's home time associated with the presence of a baby is balanced by a reduction in leisure, while the curtailment of women's economic work and girls' leisure is modest. Given this finding together with the finding that in male headed households the - productivity of women's work is not significant at the margin, it would appear that in rural Botswana the opportunity cost of women's time is not a weighty deterrent to fertility. This inference is consistent with stated desires for a large number of children. Since this issue will be discussed further in Chapter 7, it is sufficient to add one qualification here. Babies need to be cared for in the busy as well as the slack season, so that there are bound to - 75 - be occasional periods in some households when child care interferes with rewarding economiLc work. Children, especially boys, tend to report long working hours in rural Botswana. No doubt, children's work makes a significant contribution to household welfare, but our data suggest that more child labor is available than can be used productively. Evidence for this conclusion is provided by the insignifican1: marginaiL productivity of children's labor in households without cattle (about 45 percent of all housieholds) and the absence of a market for child wage labor. In Baralong Farms, where adult males are engaged in commercial crop cultivation, boys work muich longer hours in animal husbandry than in other villages with out reducing their schooling, suggesting that there may be excess Leisure in other places. Another piece of evidence may be found in the negative partial correlations between the work time of children and the number o, children in the household. Further, if adult female labor is used in marginally productire work, women could probably do with less assistance from daughters in housework and child care. One may then wonder why children work such long hours and often do not attend schooL. Part of the explanation may lie in measurement problems. Time use data do not take account of the intensity of work. If some children intersperse their work with play, reported working hours may be inflated. Another part of 1the explamation may lie in social customs regarding work sharing, parents' desire to train and socialize children for adult responsibilities, and possibly the value parents attach to their own leisure. Finally, we must remember that there are great variations among households in the economic worktlme of children and that these are related to asset holdings. Quite cliaarly, households which are well endowed with - 76 - productive assets benefit substantially from the labor of their children, while this is much less true for poor households. This finding has important implications for the process of income determination and for fertility decisions. We shall see in Chapter 7 that well endowed households do indeed have more children. The conclusions drawn here are derived from a data base which has serious limitations. Since no comparable information is available for other African countries, our conclusions for Botswana are worth setting forth even though they are admittedly tentative. Further time use studies in Botswana or other African countries are needed to test their validity and generality. Future time use studies should, first of all, attempt to collect data for more than five days during the year, so as to enhance the accuracy of the time use data. Personal observation of a small sample would be an interesting complement to a larger time use study which relies on recall of recent events. Observations which could throw light on the intensity of work effort would be particularly valuable. The analytical potential of the Botswana time use study was greatly enhanced by the very detailed income information which was simultaneously collected, and this should be continued in future studies. Further, the analysis of time use data requires a larger number of explanatory variables. In particular, it would be desirable to treat household structure and household assets as endogenous variables. For this purpose one would need to collect some information on the economic and demographic history of households, on availability of relatives with whom a large household could be formed, on relatives who have migrated and on expectations regarding their return, and on the origin of asset holdings -- migrant savings or transfers, inheritance, bride price, gifts, etc. - 77 - Chapter 4: The DistrLbution and Efficiency of Crop Production in Tribal Areas of Botswana Robert E.B. Lucas 4.1 Introduction Crop farming is not a very major source of income in Botswana; its value was originally estimated by the Rural Income Distribution Survey (RIDS) at about 8.3 million Pula or some four percent of the GDP. 1 Such crop growing as exists is almost entirely rainfed and therefore much affected by periodic droughts. Nevertheless, the potential importance of crop farming to Botswana has been emphasiLzed both in the Lipton report and in the Government's initial commitmetnt (albeit on a small scale) to the Arable Land Development Programme.2 Moreover, crop production per crop farming household is reported to vary enormously by locality in RIDS (p. 4.31) as reproduced in Table 4.1: Table 4.1: Mean Primary Income from Crops per Households with Crops (Pula) Small Villages 76 Large Villages 140 Barolong Farms 1,004 Freehold Farms 3,883 1 Rural Income Distribution Survey in Botswana, 1974-5; Republic of Botswana. 2 Michael Lipton, Employment and Labor Use in Botswana: Final Report, Vol. I, Republic of Botsiwana, Ministry of Finance and Development Planning, Gaborone, 1978. - 78 - Thus, household differences in crop production can potentially play a significant role in determining overall income inequality. The present paper, for these reasons, examines inter-family variations in crop production within tribal (non-freehold) areas of Botswana. In particular, three major lines of analysis are pursued. First, a study of the distribution of inputs into crop farming across families includes: (1) the allocation of tribal arable lands to households of various types; (2) purchases of current inputs such as ploughing services and fertilizer; (3) ownership of crop farming equipment; and (4) the allocation of family labor time to crop growing. The next section, then analyzes productivity in this arable sector and some of its determinants. Finally, some issues and estimates on the question of profitability are presented. 4.2 The Distribution of Inputs A. Land The seminal, though dated, work on land tenure in Botswana is clearly that of I. Schapera, Native Land Tenure in the Bechuanaland Protectorate, (Lovedale Press, 1943). All land in the former "Native Reserves" essentially belonged to the chief and tribe occupying the areas. Each married man was entitled to sufficient arable land to grow crops for his family. Thus, each household was given land according to its size, men with bigger families receiving larger portions. Such land could be inherited by the man's children or given or loaned to others, but not sold or rented. Inheritance was not confined to the male line, for a father was required to set aside a field for each of his daughters on marriage. This field should then be cultivated by - 79 - the woman and her husband, but pass to one of their daughters upon the mother's death. Additional land could not be cleared without permission from the chief or his surrogate, so although extensive open land may have existed it was not "free" for the taking. Indeed, In some areas, even by 1943, overcrowding of some arable land was already reported. If a man left his land temporarily, he could retain the land upon returning. But, land abandoned permanently automatically reverted to the chief. In a more recent study of land tenure in Kgatleng, Roberts (S. Roberts, "Arable Land Disputes and Administrative Change in the Kgatleng," S.S.R.C. Conference on Land Tenure in Botswana, University of Manchester, England, March 1978) reports two very impori:ant shifts in this traditional system. First, the centralized system of tribal land allotment caused such a backlog of claims that the practice of self-allotment became prevalent. Self- allotment amounted to individuals finding plots for themselves without resort to the tribal authorities, and such self allotments were generally condoned provided it did rLot result: in friction with other claimants to the land. In particular, much of this self allotment encroached upon former grazing lands. The second important change reported by Roberts is in the transfer of lands. TraditiorLally, trLnsfers occurred in the form of loans and inheritance within the sub-waLrd of the tribe. But as self-allotment proceeded, lands were no longer blocked together in subwards and transfers increased outside of the "extended family". Such transfers increasingly carried a price (often a beast for one field), at first illegally but progressively more openly. Thus, in more recent times, purchases of arable lands seem to have become much more common. ConversaLtions with District Officers and anthropologists working in - 80 - Botswana suggest, however, that overt purchase remains rare except for cleared land in which case payment is nominally for the work of clearing. It is interesting to note that despite the very high land to population ratio, at least cleared arable land is certainly not "free". To what extent this is attributable to limited availability of (good or cleared) cultivable land, to limited administrative assignment of lands to arable purposes, or to economic pressures for alternative land use in cattle grazing cannot be resolved here. Rather, this sub-section proceeds to an examination of the net outcome of these mechanisms of land allotment by 1974/75. RIDS asked each household in the survey how much land they "possessed", although these data were not reported in the main data file. Lands assigned to all members of the interviewed household were included, no matter whether these lands were ploughed or not. The quantity of land was self-reported by the household, giving rise to certain inherent problems in the units of measurement. Almost all households reporting land in RIDS did so in "acres", but these are African acres which are not a fixed area of measurement but refer rather to number of plough turns. As a measure of true area, the numbers reported obviously then contain a certain degree of pure randomness or error. But it also seems plausible the reported "acres" correlate to some extent with actual area. Since there seems no inherent reason to expect any particular direction of bias in the random portion, average errors should cancel out in looking across households with different characteristics. Thus, in this sub-section factors found to correlate with reported "acres" will be taken to be correlated with the underlying true area component. - 81 - Table 4.2 presents weighted averages of reported lands area by income class and sex oE the household head, as well as the percentage of households reporting lands. Only households actually reporting lands area are included in the average "'acreage" figures since it is unclear whether the remaining families have no land or simply failed to report lands. (Lands data are unavailable for the samples from Kanye and Etsha). Some 74 percent of rural households report having lands, and the average area is about 8 'acres". On the whole, female headed households are estimated to have 35 percent less lands than male headed households, among households with lands. rhis significant difference might be attributable to at least two types of efi-ect. On the one hand, families may tend to be female-headed because they have less lands and in particular men from households with relative]!y little land may tend to migrate within Botswana or to the South African minets. On the other hand, a household may have less land because it is female headed. With increased difficulty of ploughing for households lacking men, and generally lower incomes pushing for liquidation of assets, female-headed houiseholds may be more willing to sell off any lands they initially a,cquire. Separation of these two effects involves some tricky questions of timing, and cannot easily be disentangled from a one-shot household survey. Table 4.2 also disaggregates these results by income class of the household. Income class is defined by gross available household income (see RIDS for a precise definition) in Pula per year, divided by household size measured in adult equivalents. (Number of adults age 15 or over plus half the number of chLildren present on average over the 12 monthly interviews of the - 82 - Table 4.2 Lands Area by Income Class of Household and Sex of Head All households Male headed Female headed households households F A P F A P F A P Household income per adult equivalent (Pula per year) _ . _. <50 8.1 6.6 70.9 5.6 6.8 82.0 12.2 6.4 62.8 51-100 26.8 6.4 77.4 21.8 8.5 87.6 34.7 4.4 67.0 101-150 20.1 8.7 74.5 20.6 9.4 80.1 19.3 7.3 64.9 151-200 16.1 8.3 80.7 17.7 8.8 86.1 13.6 7.0 69.2 201-250 8.2 8.5 83.3 9.4 12.3 81.7 6.4 3.7 87.0 251-350 9.0 9.7 72.1 11.4 10.0 69.9 5.2 9.1 80.1 350+- 11.7 11.7 50.1 13.6 12.6 54.0 8.7 9.2 40.6 Overall 100 8.2 73.6 100 9.6 78.4 100 6.2 66.1 F = percentage of households in income class A - average area in "acres" P = percentage of households reporting lands area - 83 - survey. Persons of unknown age are assumed to have an equal probability of being an adult or child)'. The lowest income households, those with below 50 Pula per adult equivalent in I'able 4.2 average only 6.6 "acres" amongst land holders, and the highest income group has the largest average holding. Moreover, the lowest income group has the second lowest proportion of households reporting any lands. On the other hand, the highest income category has much the lowest fraction of households reporting any lands. Thus, the upper middle income households (151-350 Pula) average less lands given they possess, but possess so much more often than the highest income group as to average more overall if non-reporting iB interpreted as zero lands. It seems the very richest can afford or are assigned more lands if they so choose, but more often elect to focus their attentions on other forms of income earning, such as cattle. In those panels of Table 4.2 differentiating by sex of household head, the reason for so dloing should be apparent. Whereas 27 percent of male headed households receive incomes below 101) Pula per adult equivalent, the corresponding proportion of female headed households is 47 percent. The greater incidence of poverty, the lack of male heads for ploughing, and the sex difference in traditional land allotment procedures all warrant separate examination of female headed households. 1Hot only do female headed households average smaller areas of land overall, 6.2 acres as opposed to 9.6, and have a lower reporting rate of lands possession, but even within each income class female heads avetrage lesEt land area and in most cases lower chances of reported possession. The reason for lower land holdings among female headed households is apparently not merely because they have lower incomes to afford possession and operation, but it is quite plausible they tend to be lower income partly because they receive less land. - 84 - In the traditional system, land assignment is partly according to family size. To the extent arable land may today be bought and sold, one might also expect family size to influence land holdings through availability of labor to till the land and through pressures of demand for food. Figure 4.1 plots piecewise linear regression estimates of lands area against the number of adults (age 15+) in the household multiplied by the number of months present within the household during the year of the RIDS survey. 3 Both among male-headed and female-headed households there is a general tendency for those with greater adult presence to display larger areas of lands reported. However, given number of adults present, the female-headed households average considerably smaller amounts of land within each category. No matter whether the traditional or a more market-oriented system rules, ability to plough is likely to influence land allocation. For the most part ploughing is a man's task, though not exclusively. Absence of men can be supplemented by hiring of ploughing services, or by reliance on other members of the subward, but availability of adult men for ploughing within the family is likely to prove cheaper and certainly more convenient. The latter is particularly true given the importance of ploughing immediately after the first rains. Taking the ploughing season as extending approximately from November through February, the piece-wise linear regressions depicted in Figure 4.2 therefore look at the number of men actually present in the 3 Persons reported "elsewhere", "died", or "location unknown", are not included. Individuals of unknown age are counted as 50 percent in the adult category, 50 percent in the child category. See Table A.1 in the appendix. - 85 - Figure 4.1. 16 ' "Aces" Lands against adults present. / Combined Male head 14 12 / / ,' - -- - -Female head 10-- 8~~~~~~~~~~~~~~~~~~~ 6~~~~~~~~~~~~~~~~ 4 / _ fi . Adults by months present 12 36 60 84 108 - 86 - Figure 4.2. 18 "Acres" Lands against men present for ploughing. 16 Male head 14 / Combined // 12 / 10/ - -- / ,- -/ -- - Female head /,_/ 8 6 Men 0 1 2 3 4 - 87 - household during this time. (The regression results are tabulated in A.2 in the appendix). From Figure 4.2, it is clear that both within male-headed and female- headed households, the area of land rises with the number of men present. However, it is not absence of men in the female-headed household which reduces the allocation of land, for given the number of men in every instance female- headed households average less land. It is especially note-worthy that female-headed households with no man present during ploughing average some 5 "acres", but in male-headed households where the head and all other adult males are absen,t the average is about 8 "acres". Thus, it seems that it is not own ploughing capacityj which reduces land allocation to the female-headed household. Figure 4.3 explores also the associLation between cattle ownership and land allotment, depicting the piecewise regression results from Table A.3 of the appendix. In terms of technology, there is little reason to expect auy association between these variables beyond ownership of sufficient cattle (about 8 ploughing beasts, which may require a total herd of 20 animals) to do one's ploughing. But the association beyond this level is clear, land - assignment increases steadily with cattle owmed. In the traditional system, this could only reflect undue influence of the big cattle owners in land allotment through the tribal system. But with the emergence of a market for arable lands, as described by Roberts (1978, op.cit.), an alternative explanation exists: obviously, the wealthy cattle owners are better able to afford lands as a general part of their portfolio of assets. In addition, crop farming in Botswana is a very risky business as rains vary from year to year, and the wealthy would be better able to undertake this type of risk on a - 88 - Figure 4.3. "Acres" Lands against cattle. 18 - ,'/'~~~~~~~~~ 16H 14 Male >- / /'~~~~~~~~~~~~~~~~~ / / ,'~~~~~~~~~~~~~~~~~~ / X Combined , 12 X ,' ~~~~~~Female head 16 8 /' 6 Cattle 0 20 40 60 to - 89 - larger scale than necessa.ry to feed their own family. It is particularly interesting to note in Figure 4.3 that female-headed households with larger numbers of cattle have at least as much arable land as male-headed equivalent households. It seems the lesser possession of land of female-headed households is nolt so much a question of ploughing ability, but partly a question of lower incomes and capacity for affording lands, partly a question of whatever influence in the tribal system iLs correlated with cattle ownership. So far, the analysis has been confined to examination of land holdings against variables; each taken one at a time. However, Table 4.3 presents regression estimaLtes of the following equation: Qn (LAND) = b + b FEM + b2 CATL + b3 MEN + b4 RAIN + b5 FAM (4.1) Estimation is by means of "ordinary" least squares and standard errors of coefficients are presented in parentheses directly beneath each parameter estimate. Each observation is one household possessing land. The variables included in the analysis are family size (including children as half) FAM, number of men present during ploughing MEN, number of cattle owned CATL, average rainfall RAINi, and whether this is a female-headed household FEM. These included variables are defined more precisely in the appendix. Even given the values of other variables held constant, it remains true that reported lands increase with family size. On average, an increase in family size by 12 adult month equivalents raises lands held by some 8 percent. But the composition of the household also matters, since given family size and the other variables held fixed, each extra adult male present - 90 - Table 4.3 Multivariate Analysis of Lands Area Dependent variable: Logarithm of lands ("acres") Constant 0.577 (0.260) Female headed household -0.153 (0.089) Number of cattle owned 0.0017 (0.0008) Number of men present during 0.088 ploughing season (0.039) Average annual rainfall 0.0024 (0.0005) Family size (adult month 0.0067 equivalents) (0.0016) Number of observations 544 Standard Error 0.95 R2 0.16 - 91 - during ploughing season raises land allotment by 9 percent. It might be noted at this juncture that the direction of causality is indeterminate from these results, of course. In particular it may either be true that households with more men availalble acquire more land, or t'hat men'either remain or indeed join households with more fields. Given lhousehold size, men present, and the other factors held fixed, ownership of caittle remains positively associated with arable holdings: each 10 extra cattle are associated with a 2 percent increase in lands area on average. Also, given the other elements, rainfall is positively associated with "acreage". Indeed, the effect is very large -- an increase in 100 mms. of typical annual rain is associated with a 27 percent increase in reported acres" on average. Finally, given these attributes of households and their locations, it is estimated that female-headed households average 14 percent less land than comparable male-headed households. This effect is so much less than the simple differential reported earlier as to be statistically, insignificantly different from no effect at all. At least a good part of the observed lower lands areas held by households headed by women is explicable in terms of their cattle ownershilp, location and other factors entering this multivariate view. B. Purchased Iniputs Next in this study of crop inputs, let us turn to items purchased over the crop cycle. Ta'ble 4.4 presents an analysis of such expenditures by households for four types of expenditure: fertilizer, chemicals, ploughing services and petrol for tractors. 4 In this table: P indicates (weighted) 4 The examinatLon of household expenditures on wage labor for crop production is omitted from this discussion. The quality of data collected on this does not appear to warrant close examination, showing high variability across interview teams in an unreasonable fashion. - 92 - Table 4.4 The Distribution of Expenditures on Crop Husbandry Ploughing Petrol for Fertilizer Chemicals Service tractor ____ ___ ___ ___ P V_ P V* P__V P V Overall 1.4 0.7 0.8 0.0 17.3 45.8 4.0 14.0 Sex of household head Male 1.7 0.9 0.6 0.0 13.5 18.3 3.6 8.3 Female 0.8 0.2 1.1 0.1 24.9 101.0 4.7 25.2 Number of adults males 0 0. 0. 2.7 0.2 25.6 91.5 7.0 52.7 present during 1 0.7 0.4 0.5 0.0 18.8 63.6 3.1 7.2 ploughing season 2,3 2.2 0.9 0.3 0.0 15.3 19.4 3.6 6.5 4- 3.5 1.9 1.1 0.0 5.7 4.0 4.9 15.5 Number of cattle owned 0 0.3 0.4 0.3 0.0 17.7 60.5 0.7 5.3 1-8 0.4 0.4 0. 0. 18.5 31.0 4.9 28.6 9-20 1.7 1.1 0. 0. 17.7 55.5 4.4 12.3 21-80 5.9 1.2 3.5 0.2 6.5 7.4 9.3 17.3 81+ 0. 0. 8.8 0.4 61.0 72.5 12.1 6.7 "Acres" of lands 1. 0. 0. 0. 0. 35.2 405.0 0. 0. 2-3 0. 0. 0. 0. 15.9 53.9 7.7 53.9 4-6 0. 0. 0. 0. 13.9 13.7 1.0 1.8 7-10- 0.7 0.8 0. 0. 16.4 29.1 2.0 1.9 11-15 3.6 0.9 0. 0. 15.3 9.4 5.8 15.0 16-30 2.0 0.9 2.5 0.2 17.1 9.5 8.2 21.2 31-100 7.1 3.2 4.7 0.1 17.S 8.5 4.6 18.7 100+ 0. 0. 0. 0. 29.3 0.5 6.1 0.2 Average annual rainfal 300-400 0. 0. 0. 0. 14.7 37.5 5.6 5.4 (mas.) 401-600 1.2 0.7 0.9 0.1 16.S 49.1 4.3 16,8 601+ 5.7 1.4 0. 0. 17.4 5.1 0. 0. unknown 2.1 0.9 0. 0. 28.0 46.5 2.1 0.3 Village size 0-500 0,8 0.2 1.6 0.1 17.9 17.2 5.6 13.8 501-1000 0.7 0.3 0. 0. 21.5 130.9 1.1 0.8 1001-5000 1.8 0.8 0. 0. 14.4 25.9 3.3 23.2 5001+ 0.5 0.2 0. 0. 14.2 35.0 4.0 5.6 Household income per <50 0. 0. 0. 0. 20.6 34.7 0. 0. adult equivalent 51-100 0. 0. 0.4 0.0 21.0 82.0 0.5 0.1 (Pula per year) 101-150 0.9 0.9 0. 0. 11.5 15.2 3.4 7.8 151-200 0.8 0.3 1.0 0.0 14.3 13.6 1.6 1.1 201-250 3.3 1.3 0. 0. 23.6 127.8 8.2 66.0 251-350 2.3 0.4 5.2 0.4 18.4 19.8 10.1 37.0 351+ 8.2 3.7 0. 0. 115.4 14.7 16.5 38.0 Barolong Farms 0.8 8.1 5.4 0.1 18.9 14.3 10.8 39.2 P = percentage of households reporting some expenditure V = theba per "acre" spent, including nonspending households - 93 - percentage of households reporting some expenditures of each type among households reporting crops grown and lands available. V indicates value of the expenditure measured in Thebe (Pula/100) per reported "acre" of land. Lookiqg across the overall figures in Table 4.4 it is immediately obvious that with the exception of ploughing services, purchased inputs are a rarity. For example, onLy 4 percent of households report any expenditure on petrol for a tractor, and even smaller proportions report expenditures on other items except ploughing services. On the latter, the average household spends about 45 Thebe per "acre", this expenditure being made by 17 percent of crop-producing households. On average, a higher proportion of female-headed households purchase inputs and also spend more per acre on these inputs, with the exception of fertilizer. Particularly interesting is the great difference in purchases of ploughing services by female-headed households as compared to male-headed. Twenty-five perc:ent of households headed by women growing crops report expenditures on ploughing services at an average rate among purchasing and non-purchasing households of one Pula per "'acre". It seems that lack of cattle for ploughing anLd perhaps; of men in the female-headed households is compensated by purchasing; ploughing servicets. This argument would be supported by the observed decline, in Table 4.4, both in proportion of households hiring ploughing services and in expenditure per "acre" as the number of men present during ploughing season rises. Twenty-six percent of crop-growing households with no males present hire ploughing services, but only 6 percent of households with four or more men. Notice also, that although fertilizer inputs are rare, they increase in frequency and amount with the number of men present. Ihis feature presumably reflects the fact that - 94 - households with more men can afford fertilizer (out of wage earnings of men in the mines or elsewhere). The modern farming inputs -- fertilizer, chemicals and petrol for tractors - all tend to increase in frequency of use as number of cattle owned rises, presumably partly reflecting ease of financing. However, since lands area also tends to rise with number of cattle, it is less clearly true that expenditure on these items per acre rises with number of cattle. Purchases of ploughing services show no clear pattern with regard to cattle ownership. Note though, that 61 percent of households owning more than 80 cattle report hiring ploughing services: presumably their own cattle are too far removed from the lands for use in ploughing or else larger scale cattle breeders, focusing on beef production, would rather hire tractor ploughing than depreciate their cattle as draft power. Fertilizers and chemicals are found not to be used at all among farmers with small holdings. On the whole, these expenditures tend to rise with holdings though surprisingly those households with more than 100 "acres" report no such purchases. Frequency of purchases of ploughing services shows no distinct pattern among smaller compared to larger lands. Largely because of this, one finds expenditures on ploughing services per "acre" being much the highest for those with only one acre, and declining fairly steadily from there. Households with just one acre spend on average 4.05 Pula on ploughing services, but those with more than 100 "acres" spend less than one Thebe per "acre". Either there are very rapidly declining costs per acre charged for ploughing services, or those with large holdings are better equipped to do their own ploughing -- a point to which we shall return. - 95 - The use of fertilizer tends to be greater in areas with more rain in a typical year. On the other hand expenditures on petrol for a tractor are higher in the dry areas which seems rather surprising. The other purchased inputs display no particular pattern with regard to amount of rainfall. It is also true that expenditures display no obvious pattern across the various sizes of village. The lowest income households report no purchases of any of the modern crop inputs. Indeed, from Table 4.4 the consequences of a positive income elasticity of demand or of higher income households' access to fertilizer, tractor use and even chemicals is apparent. This immediately raises questions as to the distr:Lbutional consequences of programs to subsidize and promote further use of such inputs, in Botswana as in most LDCs, unless some mechanism is found to enhance use also by lower income households. On the other hand, all strata hire ploughing services with no clear pattern of variability in extent or expenditure per "acre" across income classes. Finally, in the last horizontal panel of Table 4.4 separate figures are reported for the Barolong Farms region. The exceptionally high usage of fertilizer, chemicals and petrol for tractors is immediately obvious. Quite why this region behaves so differently with respect to inputs is beyond the scope of thLs study, though we may note at least the geographical location of Barolong on the South African border and the long history of permanent crop farms in this area as opFiosed to seasonal lands dwellings in other tribal areas. C. Farm Equipment The stock of equipment held by each household is evaluated in RIDS by taking an inventory of each type of equipment separately and evaluating this - 96 - list at given prices in the office after the survey. Although this method has the obvious disadvantage of ignoring the age structure of the equipment, in a subsistence economy where records of purchase date and price are simply not kept it probably provides not an unreasonable approximation. Adding together the items of equipment associated with crop husbandry, 5 Table 4.5 shows the distribution of this equipment by value. The overall average value of equipment is 176 Pula, but it is apparent that male- headed households are better equipped for crop farming, averaging about 72 percent more equipment by value per crop producing household as compared to female-headed households. Also, the amount of equipment rises sharply, though slightly less than in proportion, as the number of men present during the ploughing season increases. This observation may generally be attributed to three plausible types of effects: (i) the savings necessary for equipment accumulation may largely be funded out of earnings of men from the mines, the amount of funds increasing with the number of men returned. (ii) Men may tend to join or not to leave those households which are better equipped. (iii) If equipment closely complements typical male crop activities - such as ploughing -- households with more men would have more incentive to invest in crop related equipment. Amount of crop equipment is also positively associated with cattle owned, presumably reflecting a general wealth effect. Although households with only one "acre" of land are least well-equipped as would be expected, there is no clear-cut tendency for increase in endowment of equipment until 5 The items of equipment included are: tractor, plough, harrow, planter sledge, handmill, wagon or cart, granary and cultivator. - 97 - Table 4.5 The Distribution of Farm Equipment by Value (Pula) Overall 176.06 Sex of household head Male 204.45 Female 118.97 Number of adult males present 0 51.10 during ploughing season 1 144.40 2-3 1 209.24 4+ 377.70 Number of cattle owned 0 100.72 1-8 118.30 9-20 214.05 2]-80 355.55 81+ 511.30 "Acres" of lands 1 89.36 2-3 128.38 4-6 118.08 7-10 153.22 11-15 116.61 16-30 273.08 31-100 390.78 101+ 388.52 Average annual rainfall (mms.) 300-400 105.10 401-600 179.82 601+ 271.27 unknown 117.18 Village size <500 144.13 501-1000 132.28 1001-5000 168.93 5001+ 235.50 Household- mncome ~per adult <50 79.39 equivalent. D X-100 71.75 (Pula per year) oi-lso 714.75 iOl-1-I50 143.04 151-200 222.71 201-250 197.21 251-350 277.96 351 511.60 Barolong 640.54 - 98 - lands area exceeds 15 acres. Among the households with larger areas of land, there is an obvious tendency to be better equipped for working this land. It is also estimated that crop equipment available is greater in areas with heavier "normal" rains. Households in areas with more than 600 mms. report more than 2.5 times the value of crop equipment possessed by those in the 300-400 mms. belt. Whether this is a consequence of greater wealth in the rainier areas or of higher contribution to production of equipment in such areas is not discernible from this simple uninvariate analysis. The larger villages are better equipped for crop farming, per crop producing household. The value of equipment for a typical crop-growing household in the largest villages is about 63 percent higher than in the smallest villages. And again in the last panel of Table 4.5 the outstanding degree of capital intensity of farming in Barolong is obvious. Indeed the average value of equipment per crop farming household in Barolong exceeds even that among the highest income group of households, despite a fairly sharp and steady rise in equipment per household across income classes. Table 4.6 displays the percentage of crop-growing households possessing at least one of each of the crop equipment types distinguished. Quite obviously ploughs are most commonly owned, though double row ploughs are comparatively rare. The equipment types associated with more modern farming techniques - planters, cultivators, harrows, tractors -- were also not common as of 1974-75, though they are certainly increasing. We have seen, however, that access to tractors may be far more common than ownership, through hiring of ploughing services. D. Family Labor RIDS asked each individual in the survey their hourly activities on - 99 - Table 4.6 Frequency of Possession of Equipment Types Percentage of crop-growing households possessing at least one PlougL: single-row 59.2 doublo-row 10.0 Sledge 28.8 Granary 25.9 Wagon or cart 10.7 Planter 5.6 Cultivrator 4.1 HandmJLl 3.4 Harrow 2.9 Tractor 0.9 Irrigation 0.0 - 100 - the day prior to interview for five of the twelve visits to each household. The classification scheme adopted for activities in RIDS includes "working for others - for wage, salary or goods", and "crop husbandry - all activities connected with growing crops". Given the inclusion of wage work as an activity, one might expect the crop husbandry activity to refer to working on one's own land. But Table 4.7 shows the fraction of time reported working on crops occurring among persons in households with no reported crop output. This ranges from 11 percent for female adults to 19 percent for male adults. Several potential explanations might be postulated: i) The survey might, of course, have misreported outputs, activities or both, but the scale involved seems to remove this as the major cause (one hopes!). ii) Crop failure is not uncommon in Botswana, and the observed time inputs with zero output may be such instances, but the harvest in the year of the survey was not bad. iii) The phasing of RIDS was very unfortunate from the perspective of crop studies. Most of the crops reported are those from the 1973-74 season, harvested in March through June of 1974 - the first four visits of the survey. Thus, only time inputs in this early period refer to the actual harvest recorded, the remainder going towards the 1974-75 harvest, only the early part of which is included. Any study of crop farming from this study must therefore assume some consistency of inputs by households across the crop years, but time inputs could have been observed in 1974-75 even though no crop was grown in 1973- 74. - 101 - Table 4.7 Percentage of Time-Spent on Crops by Non-producing Households CbLildren Male 15.6 Female 16.2 Adult Male 18.9 Female 10.6 - 102 - iv) Tribal affiliations in Botswana are characterized by understood commitments to help one another - especially within the sub-ward -- without payment of anything likely to be deemed a wage. Thus, time spent on crops, even though not for wages or goods, may nonetheless actually be work performed on lands and crops of other households. With these reservations in mind, Table 4.8 proceeds to examine time reported spent on crop husbandry but only among households actually reporting some output. The figures in Table 4.8 are estimates of the total numbers of hours spent in 12 months by all individuals in the household falling within a child/adult, male/female class. 6 Essentially, the idea is to gain some sense of the total time input by family members of a particular type. It is clear that adult women provide most of the labor time on crops. For a typical crop-producing family, the adult female members contribute some 610 hours a year working on crops. In fact, the women put in many more hours in total than the children and men combined. The girls are estimated to provide the next largest input of time at 176 hours a year, followed by men and then boys. In female-headed households, total time input into crop husbandry by boys and girls exceeds that in male-headed households, as does time input by women. To some extent this is presumably compensating for the lower time 6 The estimate is actually obtained by multiplying the average number of hours reported worked yesterday on crops per visit in which activities are recorded for the household, by 30 days per month and 12 months per year. Note that it is not unreasonable to count 30 days a month since observations include activities on all seven days at random. This also assumes the five, intermittent months in which activities are reported are an unbiased set, though in fact the very busiest month of December is omitted. - 103 - Table 4. 8 Family Labor Inputs (Hours per Year) Child Adult Male Female Male Female Overall 67 176 70 610 Sex of household head Male 60 154 83 586 Female 82 220 43 658 Adults by months present 0-11 16 84 25 355 12-23 71 107 30 375 24-47 57 192 48 528 48-9S 75 182 106 812 96+ 183 260 319 830 Number of adult males present 0 96 201 9 628 during ploughing season 1 46 155 37 539 2,3 74 186 104 689 4+ 85 188 170 .569 Number of cattle owned 0 62 167 55 653 1-8 67 171 58 557 9-20 69 177 70 549 21-80 73 193 136 631 81 65 242 36 1017 "Acres" of lands 1 55 256 20 815 2-3 41 79 52 464 4-6 67 167 7S 525 7-10 64 182 76 637 11-15 34 145 35 485 16-30 88' 197 70 810 31-100 91 217 154 564 100+ 243 426 33 744 Average annual rainfall (mms. 300-400 37 141 19 567 401-600 73 185 79 644 601+ 62 206 80 624 unknown 25 74 5 245 Village size <500 66 224 88 7.46 501-1000 43 151 60 485 1001-5000 63 140 35 552 5000+ 136 176 111 615 Household income per <50 108 297 109 734 adult equivalent 51-100 65 181 44 643 (Pula per year) 101-150 80 150 48 571 151-200 47 164 129 582 201-2SO 59 126 76 631 251-350 82 265 37 633 351+ 40 100 83 492 - 104 - provided by men within the female-headed households, but to what degree it is a result of children and the women taking up tasks done by men in the male- headed households, (where male input is seen in Table 4.8 to be higher), cannot be discerned from these data. That women and children are to some extent substituting for men seems, however, likely since their generally higher contributions are on smaller areas of land on average. But it is also true they are working with less equipment which would tend to require a greater input to achieve a given performance. Increases in family size, as measured by adults multiplied by months present, are accompanied by greater inputs of time to crop husbandry by all four family member groups. Of course, it was also observed in Table 4.3 that larger families have more land. Together, these suggest either that people attach themselves to households having more lands, or that households of greater size tend to acquire more land, such that extra crop work is performed by each group as the household is bigger. 7 At least in this sense, extra household members are not redundant to crop production. 8 Total input by adult males is found, in Table 4.8, to increase as the number of men present during the ploughing season increases. Indeed, it is estimated that each additional man contributes a greater number of hours over the year. Yet the reported hours per man per year are very low, even if they are only approximately correct. It is then difficult to imagine that other 7 Some care must be taken, for an increase in hours performed by each group may result in a total fall in hours worked as the composition of the household can also shift. 8 Even so, it is in principle feasible that people could be removed from the rural sector and even from crop production without loss of crop output. - 105 - demands on men's time, in the rural areas, are so pressing the fourth man's time in crop farming could not easily be made up by the other three if the fourth went to the mines. This would only be true if the work of all men were of necessity concentrated in a very short spell, for example with all of them working full time at ploughing after the rains. But RIDS actually shows the time of men spent on crop husbandry surprisingly widely spread over the crop cycle, (though no data are reported for December, the main ploughing month), Note also there seems to be no simple correlation between number of men present and time spent by other groups on crop care. If correct, this would tend to indicate one of two things: (i) Either there is not a strict division of tasks with children, for example, performing bird-scaring, women harvesting and men ploughing so that extra time spent by men may not be in ploughing; (ii) or there is not a fixed relationship between tasks - twice as much ploughing does nol: imply twice as much bird-scaring. No doubt the truth is some combination of these. Combined with the results comparing male and f-emale-headed households, -lt would however seem, within the confines of this mninvariate framework that male adult time input is unlikely to be a complement rather than a substitute for inputs by other household members. That time spent in crop husbandry does not31decline with number of cattle owned is somewhat unexpected, though of course this may be misleading in being a uninvariate view. With more cattle, one might have anticipated the time of boys and perhaps girls to be drawn out of crops and substituted into cattle herding, but perhaps the absolute time commitment to the former is so small as to obviate substitution. Also, bigger cattle owners have larger areas of land. On the other hand, the bigger cattle owners were also found to - 106 - be better equipped for crop farming, and perhaps the larger lands areas are simply farmed using more equipment with no more labor. More surprising is the discovery of no correlation between reported time in crop activity and area of lands. Other things equal this would lead us to anticipate lower yields on the larger holdings. But larger lands areas are associated with more equipment and fertilizer, and the net outcome in terms of yields must await investigation in the next section of this paper. Family labor time spent on crops is less for each of the groups in the more arid areas. This is consistent with the findings of smaller lands, less equipment and less fertilizer in such areas. The disincentives to farming in arid zones obviously dominate the need for greater contributions to achieve a given output for feeding one's family. Finally, within each sex/maturity category in Table 4.8 the greatest amount of time spent on crop activities tends to be in the lowest income class. This finding suggests that if marginal changes reflect the average a policy to stimulate crop production would, ceteris paribus, tend to absorb more labor from the lower income classes. 4.3 Productivity and Some Determinants of Production A. Output and Land Productivity The total value of crops produced over 12 months per crop-producing household is broken down in Table 4.9 by type of household. On average crop- producing households generate 134 Pula worth of crops in the tribal areas. Dividing by reported "acres", this amount to a yield of 17.61 Pula per "acre" overall. - 107 - Table 4.9 Distribution of Output and Land Productivity Pula Pula/"acre" Overall 134.36 17.61 Sex of household head Nale 162.87 17.43 Female 77.02 17.98 Adults by months present 0-11 35.04 9.10 12-23 58.56 19.59 24-47 113.11 17.56 48-95 193.23 17.60 96+ 211.94 13.39 Number of adult males present 0 57.06 15.92 during ploughing season 1 103.13 19.10 2,3 149.68 15.52 4+ 334.84 22.40 Number of cattle owned 0 109.59 16.17 1-8 92.70 15.17 9-20 151.28 18.45 21-80 222.66 24.73 81+ 246.97 13.78 "Acres" of lands 1 86.03 86.03 2,3 59.48 27.08 4-6 75.28 15.58 7-10 88.47 10.37 II-15 111.36 8.63 16-30 175.33 8.20 31-100 517.71 9.38 101+ 256.60 1.56 Average annual rainfall (mms.) 300-400 73.77 8.03 401-600 143.15 18.59 601+ 147.36 4.89 unknown 74.60 24.31 Village size