*258 *21 World Bank Discussion Papers Africa Technical Department Series How Fast is Fertility Declining in Botswana and Zimbabwe? Duncan Thomas Ityai Muvandi WDPQ58 Ser*. lqqy Recent World Bank Discussion Papers No. 201 Urbanization, Atqricutural Developmnent, and Land Allocation. Dipasis Bhadra and Antonio Salazar P. Brandao No. 202 A-aking M1otherhood Safe. Anne Tinker and Marjoie A. Koblinsky No. 203 Poverty Rediuction in East Asia: Tire Silent Revolution. Frida Johanscil No. 204 AManaging (he Civil Service: Tie Lessons of Reform in Induistrial Countries. Barbara Nunberg No. 205 Designin_g a System if Labor larket Sfatisfics and Inifornation. Robert S. Goldfarb and Arvil V. Adams No. 206 litfornation Technology in World Bank Lending: Increasing the Developtnental Itnpact. Nagy Hanna and Sandor Boyson No. 207 Proceedings of a Cortfereticc on Crrenicy Substituition and Cuirrency Boards. Edited by Nissan Liviatan No. 208 Developinkg Effective Em1ployment Services. David Fretwell and Susan Goldberg No. 209 Evolving Legal Frainei'orks for Private Sector Development in Central and Eastern Europe. Cheryl W. Gray and Associates No. 210 Mfaritne Biotechnology and Developing Countries. Raymond A. Zilinskas and Carl Gustaf Lundin No. 211 Revitalizing Agricultural Researchi in thie Sahel: A Proposed Frameuvrkfor Action. Jan Weijenberg, Josue Dione, Michael Fuchs-Carsch, Adolphe K&e, andJacques Lefort No. 212 Instituitional Optionsfor the Provision of Infrastructure. Christine Kessides No. 213 Thie Co,itribl,tions of Intfruistructirre to Economnic Devclopment: A Reiiew! of Experience and Political imnplicationis. Christine Kessides No. 214 Frotn Mfacroecononiic Correction to Putblic Sector Reform: The Critical Role of Evaluiation. Eduardo Wiesner D. No. 215 Chita: Reforin and Developmient in 1992-93. Peter Harrold and Rajiv Lall No. 216 The Reforn oJ Public Expenditures fir Agriculture. Bonni van Blarcom, Odin Knudsen, and John Nash No. 217 Managing Fishery Resources: Proceedings of a Sytriposimitn Co-Sponsored by the World Bank and Peruviani Ministry of Fisheries held in Lima, Peru,June 1992. Eduardo A. Loayza No. 218 Cooperatives and the Breakuip of Large Mecian ized Farinis: Theoretical Perspectives and Empirical Evidence. Klaus W. Deininger No. 219 Developoient o?f Rural Financial .Miarkets in Sub-Saharati Africa. Sabapathy Thillairajah No. 220 The Marititne Transport Crisis. HansJ. Peters No. 221 Policy-Based Finatice: The Experience of PostwarJapan. The Japanese Development Bank and The Japan Economic Research Institute No. 222 Macroeconotnic Managetnent iti China: Proceedings of a Conference in Dalian,J,ne 1993. Edited by Peter Harrold, E. C. Hwa, and Lou Jiwei No. 223 The Developmetit of the Private Sector in a Smnall Ecorioiny inr Transition: The Case of Mongolia. Hongjoo Hahm No. 224 Toward ati Environmental Strategyfor Asia. Carter Brandon and Ramesh Ramankutty No. 225 "Fortress Europe" arid Othier Myths abont Trade: Policies toward Merchranidise Imports in the EC and Other AMajor Induistrial Ecotiotnies (and 1Vhrat They Meanfor Developintg Colintries). Jean Baneth No. 226 Mongolia: Financing Eduication duiring Econotnic Transitiot. Kin Bing Wu No. 227 Cities without Land Markets: Lessons of tire Failed Socialist Experiment. Alain Bertaud and Bercrand Renaud No. 228 Portfolio Investment in Developing Countries. Edited by Stijn Claessens and Sudarshan Gooptu No. 229 An Assessment of Vuilnerable Groups in AMongolia: Strategiesfor Social Policy Plantittg. Caroline Harper No. 230 Raising the Productivity of Women Farners in Srnb-Saharan Africa. Katrine Saito (Contititued on the inside back cover.) 258 * World Bank Discussion Papers Africa Technical Department Series How Fast is Fertility Declining in Botswana and Zimbabwe? Duncan Thomas Ityai Muvandi The World Bank Washingtoni, D.C. Copyright c 1994 The International 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 September 1994 Discussion Papers present results of country analysis or research that are circulated to encourage discussion and comment within the development commnunity. To present these results with the least possible delay, the typescript of this paper has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. Some sources cited in this paper may be informal documents that are not readily available. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. The World Bank does not guarantee the accuracy of the data included in this publication and accepts no responsibility whatsoever for any consequence of their use. The boundaries, colors, denominations, and other information shown on any map in this volume do not imply on the part of the World Bank Group any judgment on the legal status of any territory or the endorsement or acceptance of such boundaries. The material in this publication is copyrighted. Requests for perrmission to reproduce portions of it should be sent to the Office of the Publisher at the address shown in the copyright notice above. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to copy portions for classroom use is granted through the Copyright Clearance Center, Inc., Suite 910, 222 Rosewood Drive, Danvers, Massachusetts 01923, U.S.A. The complete backlist of publications from the World Bank is shown in the annual Itidex of Piblications, which contains an alphabetical title list (with full ordering inforrnation) and indexes of subjects, authors, and countries and regions. The latest edition is available free of charge from the Distribution Unit, Office of the Publisher, The World Bank, 1818 H Street, NW., Washington, D.C. 20433, U.S.A., or from Publications, The World Bank, 66, avenue d'1ena, 75116 Paris, France. ISSN: 0259-210X Duncan Thomas is a senior economist at RAND Corporation and an associate professor at UCLA. Ityai Muvandi is a researcher at the Center for African Family Studies in Nairobi, Kenya. Library of Congress Cataloging-in-Publication Data Thomas, Duncan. How fast is fertility declining in Botswana and Zimbabwe? / Duncan Thomas, Ityai Muvandi. p. cm. - (World Bank discussion papers, ISSN 0259-210X 258. Africa Technical Department series) ISBN 0-8213-2993-6 1. Fertility, Human-Botswana. 2. Fertility, Human-Zimbabwe. 1. Muvandi, Ityai, 1959- . II. Title. III. Series: World Bank discussion papers; 258. IV. Series: World Bank discussion papers. Africa Technical Department series. HB1073.9.A3T48 1994 304.6'32'096883-dc2O 94-29584 CIP AFRICA TECHNICAL DEPARTMENT SERIES Technical Paper Series No. 122 Dessing, Snpportfor Microenterprises: Lessonsfor Sub-Saharan Africa No. 130 Kiss, editor, Living with I'Vildlife: Wildlife Resource Ianagement with Local Participation in Africa No. 132 Murphy, Casley, and Curry, Farmers' Estimations as a Source of Production Data: Methodological Guidelinesfor Cereals in Africa No. 135 Walshe, Grindle, Nell, and Bachmnann, Dairy Development in Sub-Saharan Africa: A Study of Issues and Optionis No. 141 Riverson, Gaviria, and Thriscutt, Ruiral Roads in Sub-Saliara, Africa: Lessonsfrom World Bank Experience No. 142 Kiss and Meerman, Inteqrated Pest Managemeent anid Africani Agriculture No. 143 Grut, Gray, arid Egli, Forest Pricing atid Concession Policies: Marraging the Hligi Forests oJ West and Central Africa No. 157 Critchley, Reij, and Seznec, iVater HarwestinfJor Plant Production, vol. Il: Case Studies and Cotnclsiotisfor Snb-Saharanr Africa No. 161 Riverson arid Carapetis, Irtennediate Meatis of Transport in Sub-Sairaran Africa: Its Potentialfor Improving Rural Travel and Transport No. 165 Kellaghan and Greancy, Utsin Exa,ninations to Improve Education: A Study int Fouirteeni African Countries No. 179 Speirs and Olsen, Indiienous Integrated Fanning Systems in the Sahel No. 181 Mining Unit, Industry and Energy Division, Strategy for African Mining No. 188 Silverman, Public Sector Decentralizatiorn: Economic Policy and Sector Inivestment Programs No. 194 Saint, Un:iversities itn Africa: Stabilization arid Revitalizationt No. 196 Mabogun:je, Perspective oni Urban Lsitd and Urbatn Aanagenment Policies itn Sub-Saharant Africa No. 197 Zymelman, editor, Assessing Ei'i'ireerinq Education int Siib-Saharan Africa No. 199 Hussi, Murphy, Lindberg, and Brennenian, The Development of Cooperatives and Other Rural Organizations: Tihe Role oflthe World Batik No. 203 Cleaver, A Strategy to Develop Agricutlture in Suib-Saharan Africa and a Focusfor the World Bank No. 208 Bindlish and Evenson, Evaluation of tile Performance of T&V Extetsion in: Kenrya No. 209 Keith, Property Tax: A Practical Vantnal for Ariglopliotie Africa No. 214 Bonfiglioli, Agro-pastoralisin i n Chad as a Strategyfor Sunriival: Ar: Essay otn the Relationship between: Anthropology and Statistics No. 218 Mohan, editor, Bibliography of Publications: Technical Departtneit, Africa Region-Jnly 1987 to December 1992 No. 225 D)ia, A Goverance Approach: to Civil Senice Refornt it: Sub-Sahraran Africa No. 226 Binidlish, Evenson, and Gbetibouo, Evafiatiot: of T&V-Based Extension in Burkina Faso No. 227 Cook, editor, Involintrary Resettlement in Africa: Selected PapersJiotn a Conferetice on: Environment and Settlement Issues it: Africa No. 232 Creightney, Transport arid Ecotiotnic Performance: A Sun'ey of Developing Countries No. 238 Heath, Land Riqhts it: C6te d Ivoire: Sun'ey atid Prospectsfor Project Intenrention No. 250 Rangeley, Thiam, Andersen, and Lyle, Internationral Rieer Basiri Organizations it: Sub-Saharati Africa No. 251 Sharma, Rietbergen, Claude R. Heimo, andJyoti Patel, A Strateyfor the Forest Sector in Sub-Saharan Africa No. 255 Mohan, editor, Biblicgrapihy of Piblications: Tecinrical Department, Africa Region,July 1987 to April 1994 Discussion Paper Series No. 82 Psacharopoulos, Wry Edutcationial Policies Can: Fail: An Oveniew of Selected African Experiences No. 83 Craig, Comparative African Experiences it: Itnpletenting Educatiotial Policies No. 84 Kiros, Implemnentinig Educatiotial Policies it: Ethiopia Discussion Paper Series (continued) No. 85 Eshiwani, Imnplemnenting Educational Policies iti Kenya No. 86 Galabawa, Imnplementinq Educationlal Policies ini Tanzania No. 87 Thelejani, Implementing Educational Policies in Lesotiho No. 88 Magalula, Imnplementiny Educational Policies in Swaziland No. 89 Odact, Inplemnewting Educational Policies in U.'anda No. 90 Aehola, Inplcmentin Educational Policies in Zambija No. 91 Maravanyika, Inplementing Educational Policies in Zimbablwe No. 101 Russell, Jacobscn, and Stanley, International Migration and Dev'elopmnent in Sub-Saiharati Africa, vol. 1: Ov'erview No. 102 Russcll, Jacobscn, and Stanley, International Migration andi Development in Sub-Saharani Africa, vol. 11: Counttry Analyses No. 132 Fuller and Habte, editors, Adjustinq Educational Policies: Consening Resources while Raising School Quality No. 147 Jaeger, The EfJicts of Econonmic Policies on Af.rcan Agricilture: Fromn Past Hann to Fiutuire Hope No. 175 Shanmugaratnamn, Vedeld, Massige, anid Bovin, Resource .IWanagement anid Pastoral Itnstituitiotn Building in the West African Sahel No. 181 Lamboray and Elmendorf, Comwbattin, AIDS and Othler Sexually Transmitted Diseases in Africa: A Review oJft/ie l-l'orld Bank's Agerna ed r Action No. 184 Spurling, Pee, Mkainanga, and Nkwanyana, Agricultural Researc/h in Southlernl Africa: A Frameivork for Action No. 211 Weijenberg, Dion6, Fuchs-Carsch, KCr-, and Lefort. RevZitalizing .4gricuiltuiral Researcd itn thle Sahel: A Proposed Framneworkfor Action No. 219 Thillairajah, Development of Rural Financial .Mfarkets in Sub-Sahlarani Africa No. 230 Saito, Raisiny the Prodnctii'ity of lVnren Farmers ini Sub-Saharan Africa No. 231 Bagehee, Agricultiral Extensiotn in Africa No. 234 Keek, Sharina, and Feder, Population Growth, Shiftinv Cultivation, and Unsuistainable Agricuiltural Development: A Case Study in Aladagascar No. 242 Biggs, Moodv, van Lecuwein, and White, ArJica Can Comripete!: Export Opportunities anid Challenges for Gartnenits and Hotne Products iti the U.S. Market No. 251 Aryeetey, Baah-Nuakoh, Duggleby, Hettige, and Steel, Supply and Demandfor Finarice of Srnall Enterprises itn Ghania No. 252 Pinto and Mrope, Projectizinig the Governance Approachi to Civil Service Reform: An Institujtiotial Environmeent Assessment for Preparing a Sectoral Adjiistmnient Loan in the Gambia CONTENTS Foreword .......................... ............................ vii Abstract ........................... .............................. ix Acknowledgements .................................................. x 1. Introduction .................................................... l 2. Data ..................................... 3 3. The Evidence ................................................... 5 4. The Evidence - Another Look ....................................... Il1 The distribution of education ...................................... 13 Impact of shift in distribution of education on fertility decline .................. 16 Reasons for the shift in the distribution of education ........................ 19 Changes in the determinants of fertility outcomes .................... I .... 22 5. Conclusions .................................................... 27 References ...................................................... 29 Appendix Table 1: Reported Number of Children Ever Born ...... . . . . . . . . . . . . . . . . 31 v TABLES AND FIGURES Table 1: Fertility, Contraceptive Use and Socio-Demographic Characteristics ............. 6 Table 2: Age-Specific Fertility Rates - Zimbabwe .............................. 8 Table 3: Age-Specific Number of Children Ever Born ........................... 12 Table 4: Education Levels of the Cohort of Women Age 25-44 in 1984 ................ 14 Table 5: Fertility Rates Holding Education Distribution Constant .................... 17 Table 6: Determinants of Number of Children Ever Born ......................... 24 Figure 1: Mean years of reported education of males and females in United States Current Population Surveys, March 1980 - March 1990 ................... 20 Figure 2: Mean years of education of females in Taiwan Personal Survey of Income Distribution, 1980 - 1990 ...................................... 20 Appendix Table 1: Reported Number of Children Ever Born ....................... 31 vi FOREWORD The 1986 World Bank Policy Study on Population Growth and Policies in Sub-Saharan Africa underscored the negative consequences of rapid population growth for the region and the need to reduce fertility and mortality to raise the quality of life. The last decade has seen a heightened commitment to the development of national population policies and to the provision of family planning services. While fertility remains high across most of the continent, there are signs that the demographic transition is under way in at least three countries that have achieved relatively higher levels of female schooling and greater access to modern methods of contraception - Botswana, Zimbabwe and Kenya. According to the results of demographic surveys conducted in 1984 and 1988 in each country, there would appear to have been a dramatic drop in fertility in Botswana and Zimbabwe over just four years. This paper on the demographic transition in Southern Africa looks more closely at the results of the two surveys in Botswana and Zimbabwe and finds that fertility has indeed declined, but less than commonly thought. At least part of the decline in fertility between surveys can be attributed to differences in sampling. In particular, women interviewed in the 1988 surveys had higher levels of schooling than the same cohort five years earlier. This is even true of older women who were unlikely to have obtained additional schooling. Since higher education is usually correlated with lower fertility, this would explain part of the observed decline in total fertility rates. There is also evidence that methodological differences in the two surveys could be responsible. This paper is one of several products of the World Bank research project on "The Economic and Policy Determinants of Fertility in Sub-Saharan Africa", sponsored by the Poverty and Human Resources Division of the Africa Technical Department and managed by Martha Ainsworth, principal investigator. It is also part of a broader research effort in the Poverty and Human Resources Division of the Policy Research Department that examines the role of human resources in economic development. We hope that this study will improve the understanding of the demographic situation in Botswana and Zimbabwe and suggest promising areas for future analysis and actions on reducing fertility and raising contraceptive use. Kevin Cleaver Lyn Squire Director Director Africa Technical Department Policy Research Department -vii- ABSTRACT Botswana and Zimbabwe have been acclaimed as being on the vanguard of the demographic transition in sub-Saharan Africa. Key data that are cited to support this claim are the Contraceptive Prevalence Surveys (CPS) and Demographic Health Surveys (DHS) which were conducted in both countries. This paper examines the comparability of these data sources and finds that at least part of the observed decline in aggregate fertility rates in both countries can be attributed to differences in sample composition. In Botswana and Zimbabwe, women of the same cohort are better educated in the second survey relative to the first. Since education and fertility are negatively correlated, this fact explains part - but not all - of the observed fertility decline across the surveys. For example, it accounts for up to half the decline among the cohort of women aged 25 to 34 in 1984 in Zimbabwe. The DHS included a complete birth history whereas the CPS asked only summary questions about the number of children ever born. There is evidence that differences in the structure of the instruments also raise questions about the comparability of the two data sources. ix ACKNOWLEDGEMENTS This paper was sponsored in part by the Regional Study on "The Economic and Policy Determinants of Fertility in Sub-Saharan Africa, " managed by the Human Resources and Poverty Division of the Africa Technical Department. The authors are grateful for financial support from the DHS Small Grants Program funded by the Andrew Mellon Foundation, the World Bank Research Committee, Yale University Center for International Area Studies and the Yale University Economics Department. The paper has benefitted from the comments of Martha Ainsworth, Esther Boohene, Randy Bulatao, Joy de Beyer, Mark Montgomery, Jack Molyneaux, Naomi Rutenberg, Paul Shaw and Jim Smith. The assistance of Trevor Croft and Elizabeth Britten, of IRD, has been invaluable. We are indebted to the Institute for Resource Development, Columbia, Maryland, the Central Statistical Office, Gaborone, Botswana, the Central Statistical Office, Harare, Zimbabwe and the Zimbabwe National Family Planning Council, Harare, Zimbabwe for permitting us to use the data they collected. Nga Vuong provided excellent research assistance. The views in this paper, and any errors, are the responsibility of the authors and do not represent the policies of the World Bank or its Members. x 1 1. INTRODUCTION Fertility in sub-Saharan Africa remains the highest in the world and there is little evidence that the kind of sustained declines in fertility that have been observed in all other developing areas are imminent for most countries on the sub-continent. Some recent evidence, however, has led several researchers and policy makers to argue that the demographic structure of a few societies in the region may be on the verge of dramatic change. In particular, Botswana, Zimbabwe, and sometimes Kenya, are frequently cited as the exceptions to an otherwise dismal picture of stable fertility rates in the region. (See, amongst others, Mhloyi 1988; World Bank 1989; Freedman and Blanc 1992; van de Walle and Foster 1990; Ainsworth, Beegle and Nyamete 1994). Evidence for dramatic fertility decline in Botswana and Zimbabwe appears to be drawn from Censuses and two recent nation-wide demographic surveys conducted in each country, the first as part of the Contraceptive Prevalence Survey (CPS) programme and the second in the Demographic and Health Survey (DHS) programme. Since it is well known that Census data are potentially subject to important biases for the estimation of fertility trends, it would seem that the weight of the evidence used to draw inferences regarding fertility trends in Botswana and Zimbabwe comes from the two recent demographic surveys.' Indeed, these surveys, which have been have been conducted in a large number of countries, have made a substantial contribution to the knowledge base regarding fertility, child survival, contraception, and maternal and child health throughout the world. In this paper, the CPS and DHS data from Botswana and Zimbabwe are re- examined. There has been some discussion of puzzles that have appeared in tabulations of these data (see van de Walle and Foster 1990, for example) although it was only recently that all four datasets were released to the public domain so that they may be compared at the micro-level. This turns out to be a useful comparison. There is some evidence that at least part of the observed decline in aggregate fertility rates in both Botswana and Zimbabwe can be attributed to differences in the sample composition. It has been pointed out, for example, that the Zimbabwe CPS tended to under-represent women at either end of the age distribution (World Bank 1989) and there is evidence that the DHS suffered from the same problem. There also appears to be substantial under-reporting of fertility by older women in the second Botswana survey (van de Walle and Foster 1990). Evidence is presented here for both Botswana and Zimbabwe which indicates that relative to the CPS, women in the DHS tend to be better educated. This does not simply reflect the fact that younger women have more education than older cohorts: women of 1. This is especially true in Zimbabwe where data from two Censuses are usually cited; the 1969 Cenisus is generally recognized as being of limited value to policy makers in post-Independence Zimbabwe. There is also some debate regarding the quality of the 1982 Census which was conducted soon after Independence. 2 the same cohort are better educated in the second survey, relative to the first. One of the few facts that social scientists agree on is that there is typically an inverse correlation between education and fertility. Part of the observed decline in fertility can apparently be accounted for by the shift in the distribution of education of the same cohort of women across the two surveys. The results presented below also suggest that differences in the design of the fertility questions in the CPS and DHS may further contaminate inferences drawn from aggregate statistics. While there is certainly evidence for some decline in fertility in both Botswana and Zimbabwe, the evidence for dramatic decline, based on these data, is rather less clear. 3 2. DATA As part of the Contraceptive Prevalence Survey programme, the first wave of the Botswana Family Health Survev-I (BFHS-I) was conducted in 1984 by the Central Statistics Office in collaboration with the Institute for Resource Development (IRD).2 Drawing a nation-wide sample of 3,064 women aged 15 through 49, the survey collected information on fertility, contraception. child survival and some socio-demographic characteristics of the woman (Manyeneng and others 1985). Four years later, in 1988, the same agencies collaborated in the collection of the second wave of the Botswana Family Health Survey-Il (BFHS-II). which formed part of the worldwide Demographic and Health Surveys. This survey was both more extensive (with a sample size of 4,368) and also considerably more comprehensive. In addition to the information collected in BFHS-1, the second wave collected detailed information on maternal and child health, breastfeeding and contraceptive histories (Central Statistical Office 1989). Two similar surveys were conducted in Zimbabwe during the same period. The CPS, called the Zimbabwe Reproductive Health Survey (ZRHS), was carried out in 1984 by the Zimbabwe National Family Planning Council in collaboration with IRD and covered 2.574 women (Zimbabwe National Family Planning Council 1985). In 1988, the Central Statistical Office and IRD implemented the Zimbabhve Demographic and Health Survey (ZDHS) which had a substantially larger sample of 4,201 women (Central Statistical Office 1989). Although in most respects the 1984 and 1988 surveys in each country are broadly comparable. there is a key difference between them that may be important for a study of fertility. In the 1984 surveys, each woman was asked about children to whom she had given birth (living at home and away) and those that died. In the 1988 surveys, however, after eliciting that information, the enumerator asked each woman to provide a complete birth history covering every child. If there was a discrepancy between the number of birth history entries and the original tally of children born, the enumerator was instructed to reconcile the difference.' This caveat notwithstanding, these data offer an unique opportunity to examine the dynamics of fertility change in Botswana and Zimbabwe. In particular, they afford the researcher the luxury of being able to cross-check estimates for consistency without having to make strong assumptions about the underlying data generating process. There can be little doubt that the availability of these surveys has already had a substantial impact on the understanding of demographic processes in Southern Africa; as additional 2. IRD/Macro Systems was named Westinghouse Public and Applied Systems at the time. 3. Unfortunately, on the data tapes used for this study. we are unable to identify whether such discrepancies arose or how they were resolved. An examination of these data may be a valuable exercise in and of itself. 4 data become available at the micro-level, further analyses along these lines will presumably add to this understanding. The aim of this paper is quite simple: to determine the extent to which the observed dramatic decline in fertility in Botswana and Zimbabwe reflects reality as opposed to differences across the surveys. 5 3. THE EVIDENCE Table I presents some summary statistics regarding fertility and sample composition from the four surveys: the 1984 ZRHS, the 1988 ZDHS, the 1984 BFHS-I and the 1988 BFHS-I1.4 The story they tell is fairly well known. The total fertility rate (TFR) in Zimbabwe was estimated to be about 6.5 in 1984 and declined by one child to 5.5 by 1988. The average woman had borne 3.4 children according to the 1984 survey and almost one half a child less, 2.95, in the 1988 survey. Completed fertility, as measured by the number of children ever born to women aged 45 to 49, fell from 7.5 in 1984 to 6.9 in 1988, suggesting that much of the decline in fertility has been concentrated among younger women. These declines in fertility have occurred in the context of rising child survival rates as well as increasing knowledge and use of modern contraceptives (see Table 1). Knowledge of modern methods is virtually universal in Zimbabwe today. The usage rate among all (married and unmarried) women has risen by almost 20 percent, from 23 percent in 1984 to 27 percent in 1988. These increases are probably a reflection of both successful family planning and public health programs (Boohene and Dow 1987), as well as broader social changes that have taken place in the country since Independence in 1980.5 Massive spending on social services by the government is surely partly responsible for this change;6 Zimbabwe has registered very impressive successes in increasing access to both the health care system and public education. For example, between 1980 and 1986. primary school enrollment ratios rose by over 40 percent. The vast majority of the people of Zimbabwe were excluded from secondary schools prior to Independence; at that time, only 8 percent of eligible children were enrolled. These enrollments have increased by almost six fold, to 46 percent by 1986. The number of tertiary students has tripled during the same period and, unlike many developing countries, the vast majority (over 70 percent) of these students are in education, science and teacher training (UNESCO 1986). These increases are also reflected in the demographic surveys; in 1984, the average woman had 4.9 years of schooling whereas by 1988 she had over six years. 4. The BFHS I and It as well as the ZRHS estimates are all weighted to account for different sarnpling proportions (of rural and urban households) included in the surveys; the ZDHS is a proportional probability sample. As a matter of fact, the essence of the results are independent of whether or not weights are applied. 5. Zimbabwe is often cited as having a very successful family planning programme. Family planning services were first introduced in Zimbabwe in 1953 and they have been, at least since 1966, an integrated component of the public health system. In 1984, famnily planning services were reorganized and the Zimbabwe National Family Planning Council was formed as a parastatal operating under the Ministry of Health. 6. In 1989, government spending as a proportion of GNP was high (40 percent). Almost one quarter of public spending went to education, while the health sector received 7.6 percent of the public budget. 6 TABLE 1: FER77LITY, CONTRACEPTIVE USE AND SOCIO-DEMOGRAPHIC CHARACTERIS77CS Means and [standard errors]. 1. ZIMBABWE: Reproductive and Demographic and Health Surveys ZRHS ZDHS 1984 1988 Fertility Total # children ever born 3.40 [.06] 2.95 [.05] Completed fertility (45-49) 7.46 6.87 Total fertility rate 6.5 5.5 Child survival rate 90.3 92.0 Contraceptives: % of all women who Know of modern method 81.1 95.4 Currently use modern method 22.8 27.2 Socio-demographic characteristics Age 28.02 [.18] 27.82 [.15] Years of education 4.92 [.07] 6.06 [.06] Sample size 2574 4201 Ever had child 2014 3005 2. BOTSWANA.: Family Health Surveys BFHS-I BFHS-II 1984 1988 Fertility Total children ever born 3.05 [.05] 2.58 [.04] Completed fertility (45-49) 6.85 5.75 Total fertility rate 6.5 5.0 Child survival rate 90.8 93.9 Contraceptives: % of all women who Know of modern method 65.5 95.1 Currently use modern method 16.1 28.9 So cio-demographic characteristics Age 28.28 [.17] 27.68 [.141 Years of education 4.40 [.07] 5.48 [.061 Sample size 3064 4368 Ever had child 2414 3279 7 Taking a longer perspective on fertility, however. there appear to be some puzzles. According to Census data, the total fertility rate in Zimbabwe declined from 6.7 to 5.6 between 1969 and 1982; it then rose to 6.5 in 1984 and had fallen back to the 1982 level by 1988. Apart from young women (aged 15 to 19), age-specific fertility rates in 1982 and 1988 are very close, whereas those in 1969 are substantially higher for every age group. This would be consistent with a decline in fertility during the 1970s and a levelling off in the 1980s. Estimated age-specific fertility rates based on the 1984 survey, however, are rather different. They are higher than the 1982 and 1988 estimates for all women under 40 and are, in fact. 40 percent higher for women aged 35 to 39. For older women, the reverse is true. Based on the 1984 survey, estimated fertility rates of women aged 45 to 49 were only one third of the comparable estimates in both the 1988 and 1982 datasets. It is generally thought that Census data tend to underestimate fertility as women are inclined to fail to recall births several years ago and, especially, those that involved early mortality. There are, by now, many potential indirect methods which seek to adjust fertility (and mortality) data to account for recall error. The World Bank (1989) has applied the Brass (1968) P/F method to evaluate the internal consistency of the four data sources and argues that there was indeed substantial under-reporting of fertility in the two Censuses and thus calculated adjusted fertility rates (reported in Table 2). According to these adjusted data, the "national total fertility rate was close to 8 in the late 1960s, ... it then fell to around 7 by 1981/82, to around 6.5 by 1983/84 and to an average of 5.7 around 1986"' (World Bank 1989: see, also. Mhloyi 1988). Researchers and policy makers have, therefore, come to the conclusion that there is evidence for significant fertility decline in Zimbabwe during the 1980s. Turning next to Botswana, according to the 1981 Census, the total fertility rate was about 7.1; it had declined to 6.5 by 1984 (according to the first wave of the BFHS) and collapsed to 5.0 by 1988 (BFHS 11). If these data reflect reality, then fertility has declinied by over 25 percent in only 7 years and much of this decline is also concentrated in the later part of the eighties. Furthermore, as in Zimbabwe, the number of children born to the average woman declined by almost hali a child (from 3. 1 to 2.6 children) and the completed fertility rate dropped by over a child, from 6.9 to 5.8, suggesting that the reductions in fertility are also concentrated among younger women. For a discussion of and economic explanation for fertility change in Botswana. see Rutenberg and Diamond (1993). The Botswana Maternal Child Health/Family Planning Unit was formed in 1973 within the Ministry of Health, but it has been during the 1980s that public investment in family planning has grown most rapidly - especially since 1984, when the first BFH survey was fielded. The evidence suggests these investments have had a high rate of 7. The authors do not discuss the rather surprising fact that among old wonmen, even according to the adjusted numbers, fertility decliined between 1982 and 1984 but then rose by 1988. 8 TABLE 2: AGE-SPECIFIC FERTILITY RATES - ZIMBABWE Census 1969 Census 1982 ZRHS 1984 ZDHS 1988 Age Unadj Adj Unadj Adj Unadj Adj Unadj Adj 15-19 .08 .12 .09 .13 .13 .16 .10 .13 20-24 .27 .35 .26 .33 .29 .30 .25 .26 25-29 .30 .37 .25 .32 .30 .30 .25 .26 30-34 .26 .31 .23 .38 .26 .26 .22 .22 35-39 .22 .26 .17 .21 .22 .21 .16 .16 40-44 .15 .16 .09 .11 .09 .08 .09 .09 45-49 .07 .08 .04 .04 .11 .01 .04 .03 TFR 6.7 8.2 5.6 7.1 6.5 6.5 5.7 5.5 Sources: Central Statistical Office 1985; Johansson 1989; Zimbabwe National Family Planning Council 1985; Zimbabwe Central Statistical Office 1989; World Bank 1989. All adjusted numbers are drawn from World Bank 1989, Table 11. 1969 and 1982 adjustments based on P/F ratios; 1984 and 1988 are adjusted for true age-group. return. Knowledge of contraceptives has increased by 50 percent: in 1984 about two- thirds of women knew about modern methods and by 1988 that proportion had increased to 95 percent. As in Zimbabwe, knowledge of modern methods had become virtually universal by 1988. Use of these methods rose even faster, almost doubling during the four years between the two surveys from 16 percent to 29 percent. School enrollment rates also rose during this period' and this is reflected in the two surveys: the average woman had 4.4 years of schooling in 1984 and 5.4 in 1988, placing her slightly below the average Zimbabwean woman. All of this evidence suggests that fertility has indeed declined dramatically in both countries during the four years between the two surveys. And several authors have heralded the onset of the (long-awaited) demographic transition in sub-Saharan (or at least Southern) Africa. Yet the final reports for both the ZDHS and BFHS II recommend caution in taking these declines at their face value (Botswana Central Statistics Office 1989; Zimbabwe Central Statistical Office 1989). Several reasons have been cited which suggest that following their advice would be prudent. Recall that the 1984 surveys collected summary information on the number of children ever born whereas the 1988 surveys first asked each woman the number of children ever born and then obtained a birth history on each child. One might expect the 8. Secondary school enrollments rose by over a quarter from 25 percent in 1984 to 32 percent in 1988. In 1989, the share of the public budget spent on education and health (20 percent and 5.5 percent) was slightly less than in Zimbabwe but government spending accounted for a larger share of GNP (50 percent). 9 birth history method to result in less under-reporting but it has been argued that in fact, in Africa, birth histories tend to result in lower estimates of fertility (Government of Kenya 1989). Exactly why this should be so is not at all clear (van de Walle and Foster 1990); it is plausible that high parity women suffer from fatigue and so truncate their birth histories (and then reduce the total number of children born) or they may just get confused as they enumerate each child. It has also been noted that there is some slippage in the birth histories as respondents (or enumerators) appeared to mis-classify children as being older than 5 and thus not complete the child health module (Rutstein and Bicego 1990; Arnold 1992). Furthermore, both the demographic surveys collected information on only woman aged 15 to 49 in each household. One might expect that young and old women would be mis-classified as being outside the admissible age range in order to reduce the interviewer's workload (Arnold 1992). There is some sense that this is indeed a problem in the Zimbabwe surveys since the age distributions of women indicate considerable under-representation by young women' and also in the Botswana DHS. The next section takes another look at the evidence to determine whether there might be other indications in the data that suggest prudence in inferring a time-series pattern from the two surveys. 9. Whereas one quarter of women between 15 and 49 were aged 15 to 19 according to the 1982 Census and 1988 ZDHS, this age group accounted for only 20 percent of women in the 1984 ZRHS. The 1987 Intercensal Demographic Survey (Central Statistics Office, 1991) collected information on some 29,000 women aged 12 and above: these data should not be subject to end- point problems at 15 or 49. According to the 1987 data, 27 percent of the 15 to 49 year old women were aged between 15 and 19 suggesting that young women were mis-classified in the ZDHS as well. I1 4. THE EVIDENCE - ANOTHER LOOK Table 3 presents the age-specific number of children reported to have been ever born to women in Zimbabwe (in the upper panel) and Botswana (in the lower panel). Column 1 is based on the 1984 survey and column 3 on the 1988 survey. According to these numbers, there have been significant declines in fertility in Zimbabwe in every age group with the largest declines being in the early ages. Similarly, in Botswana, significant reductions in fertility are registered for every age group, apart from the youngest (among whom fertility has not changed significantly and is slightly higher in the second survey although this may simply reflect sampling variation). Based on these data, one might also infer that women tend to complete their fertility around age 40 in both countries since the number of children ever born to women aged 35-39 in 1984 (5.36 in Botswana) is only slightly less than the average number born to women aged 40-44 in 1988 (5.43 in Botswana; the comparable numbers in Zimbabwe are 6.2 and 6.4, respectively). This inference may, however, be misleading since reported fertility for the next cohort of women (aged 40-44 in 1984) actually declined during the four years between the surveys, from 6.3 to 5.8 in Botswana and from 7.0 to 6.9 in Zimbabwe.'" Pointing out this fact in Botswana, van de Walle and Foster (1990) attribute it to misreporting and suggest it reflects differences in the methods used to collect fertility data in the two sets of surveys." Given the fact that the anomaly arises in both countries, this seems an appealing hypothesis. The study next exploits the birth history information in the 1988 surveys. The number of children each woman reported she had borne as of 1984 is calculated and her age is set to the level it would have been in 1984. These estimates of her fertility by 1984, based on the 1988 data, are reported in column 2 of Table 3 (labelled ZDHS 1984 and BFHS-2 1984 for Zimbabwe and Botswana respectively). Columns I and 2 are, therefore, estimates of the same thing: the average number of children born to a woman of a particular age (group) as of 1984. In an ideal world, they would be identical (apart from sampling variation). On average, for women aged 15 through 44 (in 1984),12 fertility is significantly lower in the 1988 survey relative to the 1984 survey in both Zimbabwe and Botswana. For example, in Zimbabwe, the 1984 survey estimates fertility in 1984 to be 3.1 children per woman, but only 2.8 at that time according to the 1988 survey. This difference accounts for 80 percent of the observed decline in the number of children ever born over the period 1984 to 1988. In Botswana, the difference between the estimates based on cohort adjustment (in column 2) is (slightly) greater than the difference between the observed number of children ever born in the two surveys and this is due to much 10. This decline is significant in Botswana but not in Zimbabwe. 11. Recall summary data were collected in the 1984 surveys and birth histories collected in the 1988 surveys. 12. Since women over 49 were not included in the survey in 1988, ages in 1984 are truncated at 45. 12 TABLE 3: AGE-SPECIFIC NUMBER OF CHILDREN EVER BORN (1) (2) (3) 1. ZIMBABWE: Reproductive and Demographic and Health Surveys Survey: ZRHS 1984 ZDHS 1984 ZDHS 1988 Age 15-49 3.396 [.06] 2.953 [.05] 15-44 3.120 [.061 2.756 [.05] 2.662 [.041 15-19 0.303 [.03] 0.224 [.02] 0.188 [.01] 20-24 1.649 [.051 1.469 [.05] 1.299 [.04] 25-29 3.205 [.08] 3.088 [.07] 2.894 [.06] 30-34 4.630 [.11] 4.403 [.09] 4.346 [.09] 35-39 6.219 [.151 5.734 [.14] 5.537 [.11] 40-44 7.037 [.211 6.715 1.19] 6.399 [.17] 45-49 7.464 [.26] . 6.872 [.20] Sample size 2574 3312 4201 2. BOTSWANA: Family Health Surveys Survey BFHS-1 1984 BFHS-2 1984 BFHS-2 1988 Age 15-49 3.054 [.05] . 2.581 [.04] 15-44 2.809 [.05] 2.330 [.041 2.387 [.04] 15-19 0.256 [.02] 0.204 1.01] 0.261 [.02] 20-24 1.444 [.041 1.412 [.03] 1.166 [.03] 25-29 2.870 [.06] 2.679 [.061 2.546 [.05] 30-34 4.164 [.09] 4.160 [.09] 3.698 [.07] 35-39 5.362 [.151 4.737 [.141 5.088 [.111 40-44 6.259 [.211 5.622 [.19] 5.425 [.18] 45-49 6.845 [.28] 5.752 [.21] Sample size 3064 3593 4368 Note: [Standard errors in parentheses]. First and third columns based on 1984 CPS and 1988 DHS resp. Middle column based on 1988 DHS evaluated at 1984; fertility calculated using birth history information. 13 lower estimates (based on the 1988 data) among women aged 15 through 19 and 35 though 39 in 1984. There are several possible explanations for these differences. Recall that methodology differed across the two surveys and there is some evidence that young and old women were under-represented in the samples. Perhaps the most obvious candidate is recall error and, indeed, it is standard practice in demographic studies to treat this sort of difference as an indicator of recall error. (In essence, this is a key idea underlying the P/F method for adjusting retrospective fertility data with a single survey as suggested by Brass 1968).3 But might there be some other reasons for the anomalies in the estimates of fertility based on these surveys? The distribution of education It has been pointed out above that during the 1980s, Zimbabwe, and to a lesser extent Botswana, enjoyed spectacular growth in the educational attainment of its youth. This growth is reflected in the fact that the average woman in 1988 reported one more year of schooling than her counterpart in 1984. Even in the context of the rapid social change that took place in these countries, this represents a large increase. This sub- section, attempts to determine the extent to which it reflects changes in the underlying populations rather than sampling differences. Because education levels have increased over time in both Botswana and Zimbabwe, it is not possible to disentangle population from sample differences by comparing the education of women of the same age. Instead, to isolate the differences in education between the samples, it is important to compare the same cohorts of women. Table 4 summarizes education levels reported by women age 25 to 44 in 1984. The first column is based on the 1984 surveys; the second column is based on reported education of the same cohort of women in the 1988 surveys, at which time they would have been age 29 to 48. The third column reports the difference between reported schooling in the second and first survey for this cohort of women. If there are no sampling differences between the pairs of surveys then the difference should be zero. 13. Little is known about the reliability of recall data that involves remembering dates (as in the birth histories recorded in 1988). Becker and Mabmud (1984) attempt to validate retrospective birth history data for Matlab by matching them with vital sistics. They find very few births are missed (around 5 percent) and these tend to be non-live births. There is, however, a general tendency for women to place the event too far back; for example, among women aged 30-39, the reported fertility rate for the previous four years was 4 percent lower than the actual rate (using a backward questionnaire) and 6 percent lower for women aged 40-49. This evidence would suggest that numbers in the second column of Table 3 should be higher than in the first column. Recall also that in the DHS there is some evidence that childrnn aged 5 (born in 1983) were misclassified as born before then although this fact should not affect fertility as of 1984 (reported in 1988). 14 TABLE 4: EDUCATION LEVELS OF THE COHORT OF WOMEN AGE 25-44 IN 1984 1. ZIMBABWE Surve-: , . ZRHS ZDHS Dfterence -. {ZRHS - ZDHS Average number of years of education 4.36 . 4.68. .0.32 10.08] j10.081 10.12] Percentage of women No education 22.9 21.2 -1.7 11.11 11.01 11.51 Completed primary 30.0 34.0 4.0 school/more 11 21 11.11 11.7] *Attended secondary 10.5 13.5 3.0 school 10.81 10.81 [1.2] Completed more than 4.4 66 2.2 Form 2 10.6] [0.61 10.8] 2. BOTSWANA Surt-ev. BFHSI BFH-52 Difference BFHS2-BFHSI Average number of years of education 3.53 4.02 0.49 10.061 [0.05] 10.131 Percentage of women No education 37.3 36.5 -0.8 [1.2] [1.1] [1]7] Completed primary 27.1 30.2 3.1 school/more [1.11 [1.01 [1.6] Attended secondary 10.5 15.3 4.8 school 10.81 [0.8] [1.2] Completed more than l 8 6.6 4.8 Form 3 10.31 [0.61 10.7] Note: Standard errors in parentheses. 15 In both surveys. each woman reports the level of schooling (primary, secondary and tertiary) she attained as well as the specific grade completed (or passed) within that level. These data have been converted to years of education and average attainment is presented in the first row of each panel. The fraction of women who completed particular levels is reported in the rest of eadh panel.4 On average, the same cohort of women in the DHS reports more years of schooling than those in the CPS. This difference is one-third of a year in Zimbabwe and half a year in Botswana and, in both cases, this difference is significant. In Botswana, for example, the t-statistic on the difference is 3.8. Where. within the education distribution, are these differences concentrated'? Examining the grouped data in the remaining rows of Table 4, slightly fewer women report no schooling in the DHS than in the CPS. But this discrepancy is not significant. However, the probability that a woman reports herself as having completed primary school is significantly greater in the DHS. Similarly, a significantly higher proportion of DHS women report having attended secondary school. (The t-statistics on the differences are 4.0 in Botswana and 2.5 in Zimbabwe.) This inference also carries through to the proportions reporting completion of Form 3 in Botswana and Form 2 in Zimbabwe.1 14. A very sniall fraction of women reported they had attended primary school but could not recall the exact grade; we assume they have not completed primary schooling and, when calculatiing years of schoolinig, assume they completed three at the primary level. Since less than I percent of women failed to report an exact grade, the effect of varying this assumption on our estimates is trivial. An assumption also has to be made about the years of schooling completed in previous levels but because exactly the same assumnptions are made in processing both the 1984 and 1988 surveys, this is unlikely to generate spurious discrepancies in education levels between the surveys. Furthermore, this concern is only likely to be important for women who complete 'O' or 'A' Level and continiue to a tertiary education institution. They account for a tiny fraction of the samiple. For example, in both the Zimbabwe CPS and DHS, only about 1.5 % of all women report more than Form 4 schooling (when 'O' Level examinations are often written). Obviously, data processinig differenices of one year or even two in these women's education levels will not have any substantial impact on the average for the whole sample. Furthermore, inferences based oni levels of schooling, rather than years, will be unaffected by these assumptions. The 1984 survey asks about higlhest grade completed whereas the 1988 survey ask about highest grade passed. The bulk of the population does not complete primary school: for them there are no tortnal examiniations that need to be passed in order to progress to the next grade. We thus view the differenices in questions as being largely one of sermantics for most of the population. The distinictioni between passinig and completing is relevant only in the case of public examinations, suclh as 'O' Level and 'A' Level, which are taken toward the end of secondary school. The estinmates of sample completioni rates in Table 4 are for levels below 'O' Level. 15. For Zimbabwe, these estimates can be compared with the distribution of education of the same cohorts of women as recorded in the 1987 Intercensal Demographic Survey (ICDS). Relative to this (large) samulple, 5-year cohiorts of women aged 25 through 49 as of 1984 uniformly report around a fifth of a year less schooling on average in the ZRHS and the difference for this age group is significanit. Of course, women under 25 are also better educated in the 1987 survey: for 16 In sum, women in the 1988 surveys report significantly more education than those women of exactly the same cohort in the 1984 surveys in both Botswana and Zimbabwe. A good part of the higher average number of years of education in the 1988 survey reflects a shift in the education distribution from primary to secondary school. Whether this is because the women in the 1988 survey were better educated than those in 1984 or whether they simply reported themselves as being better educated, we cannot tell.16 What one can say, however, is that simple comparisons of aggregates based on these data may be quite misleading. Indeed, given the fact that education and fertility tend to be negatively correlated (see below), this suggests that it would be prudent to evaluate the evidence regarding dramatic decline in fertility in both countries with this caveat in mind. Impact of shift in distribution of education on fertility decline Recall that, according to the CPS and DHS, among women age 15 to 49, between 1984 and 1988, the number of children ever born declined from 3.40 to 2.95 in Zimbabwe and from 3.05 to 2.58 in Botswana. How much of this observed decline can be attributed to sampling differences in the survey? This question can be answered by exploiting the micro-data in the surveys to calculate predicted fertility rates while holding the underlying distribution of education for the population constant. These estimates are reported in Table 5. example, women aged 15-19 report almost a year more schooling in 1987. Young women in the 1988 ZDHS report significantly more schooling than those in the 1987 survey: women aged 15-19 (in 1988) report a half year more schooling in the 1988 survey. This difference is also significant and seems like a very large nunmber for a single year suggesting that perhaps it reflects differences in the samples and not just additional schooling. For women aged 25 through 49, in 1988, educational attainment reported in the 1987 and 1988 surveys are virtually identical (at about 4.8 years of schooling); it differs by about .04 years which is half the standard error on the difference (0.08). For these older women, the 1987 ICDS and 1988 ZDHS education distributions are consistent although this does not imply that they are necessarily correct. 16. A National Literacy Programme (NLP) was established in the late 1970s in Botswana. A 1987 evaluation of the programme indicated it was very successful with 81 percent of those tested attaining literacy levels associated with Standard 4. (UNICEF, 1989). Since the questions in the BFHS asked about the highest grade passed (BFHS-1) or completed (BFHS-2) at school, it is unclear that the NLP should have any impact on reported years of completed schooling in the surveys. Nevertheless, there is evidence for both cohorts of women of a shift from reporting 1 to 3 years of schooling to reporting 4 to 5 years of schooling. However, for our purposes, the key shifts in the education distribution are in secondary schooling and the NLP cannot possibly explain that fact. Furthermore, recall that, on average, half a year more of schooling is reported by Batswana women surveyed in BFHS-2 compared with those in BFHS-1. In BFHS-1, half the women age 25 to 44 were literate in 1984; thus, to account for the differences in education, every single one of the other half would have had to participate in the literacy program between 1984 and 1988 and would have had to report themselves as having completed an extra year of schooling. This implies that by 1988, among women age 25 to 44, illiteracy would have been wiped out in Botswana. 17 TABLE 5: FERTILITY RATES HOLDING EDUCATION DISTRIBUTION CONSTANT Age Cohort: 25-34 35-44 1. ZIMBABWE Fertility rate based on (1) (2) (3) (4) (1) (2) (3) (4) survey dated: % % 1984 1988 Diff unexpl 1984 1988 Diff unexpl Education distribution: As reported 3.85 3.57 0.28 . 6.58 5.89 0.69 Based on 1984 survey 3.85 3.62 0.23 20 6.58 5.96 0.62 10 Based on 1988 survey 3.70 3.57 0.13 52 6.49 5.89 0.60 14 Based on 1987 ICDS 3.73 3.55 0.18 36 6.46 5.89 0.60 14 2. BOTSWANA Fertility rate based on (1) (2) (3) (4) (1) (2) (3) (4) survey dated: % % 1984 1988 Diff cunexpl 1984 1988 Diff unexpl Education distribution: As reported 3.43 3.07 0.35 . 5.76 5.17 0.59 Based on 1984 survey 3.43 3.10 0.33 7 5.76 5.33 0.43 27 Based on 1988 survey 3.42 3.07 0.35 0 5.64 5.17 0.47 20 Note: % unexpl is percentage of observed fertility decline that is left unexplained when the education distribution is held constant. The first row presents the average number of children ever born as reported by women of the same age in the 1984 survey (in column 1) and in the 1988 survey (in column 2). The differences, in column 3, is the observed fertility decline for this age group of women. In Zimbabwe, the decline is seven-tenths of a child for women age 35 to 44.'7 Appendix Table I reports the number of children ever born to women, stratifying on both age and education. With these fertility rates, one can calculate the average number of children ever born to women as reported in each survey, using the distribution of education for each cohort implied by the 1984 and 1988 surveys. 17. Without an education history, it is necessary to restrict attentioni to women who would have completed their education by 1984 (and who are also followed in 1988): thus, the evidence is examined for women age 25 to 44, stratified into two ten year age groups. 18 To illustrate the procedure, consider women age 25-34. The education specific fertility rates based on the 1984 survey (Appendix Table I) are multiplied by the education distribution in that survey. The sum of these products. which is reported fertility for these women, is displayed in row 2. column I of Table 5. Next, take the education distribution for women age 21-30 in 1984 (who would have been 25-34 in 1988) and multiply those numbers by reported education specific fertility rates as reported by 25-34 year old women in the 1988 survey. The sum of those products is an estimate of the fertility this cohort of women would have reported in 1988 had the education distribution remained constant across the two surveys (at the level reported in the 1984 survey while adjusting for the fact that women in different cohorts are being compared). This number is displayed in row 2. column 2 of Table 5. The difterence between these numbers is an estimate of fertility decline for this cohort of women, holding the education distribution constant. Comparing this estimate with the decline as reported in the two surveys (row 1, column 3) tells us how much of that reported decline can be explained by the change in the education level of the same cohort of women across the two surveys. This is the percentage reported in column 4; if none of the observed decline can be attributed to changing sample composition, then this proportion should be zero. The same exercise is repeated using the education distribution in the 1988 survey (row 3) and, for Zimbabwe, the 1987 Intercensal Demographic Survey (row 4).1 In Zimbabwe, among women aged 25-34, over half the observed decline in fertility across the two surveys can be attributed to differences in the distribution of education of this cohort assuming the distributions implied by the 1988 survey; 20 percent remains unaccounted for when the 1984 distribution is adopted and 37 percent given the 1987 distribution. Among women aged 35 to 44, between 10 and 14 percent of the observed decline is due to the shift in the distribution of schooling. In Botswana, the observed decline in fertility for 25 to 34 year old women is robust to changing the education distribution. Among older women (35-44), however, between 20 and 27 percent of the observed fertility decline can be attributed to the sample composition. There is no doubt that there has been a decline in fertility in both Zimbabwe and Botswana. What is in doubt, however, is the rate of decline. Estimates based on aggregate data may be too high as they can, in part, be explained by the fact that relative to the 1984 survey, the 1988 survey gathered information from women who are, or report themselves as being, better educated. Without knowledge of the shape of the true distribution of education at each survey date, it is very hard to determine the magnitude (and significance) of fertility change in either country. 18. It is important to note that tests of differences in the education distribution must be based on comparing women of the same cohort in the two surveys. To assess the impact of these differences on fertility decline.women in the same age group must be compared. If the samples are identical, then a comparison of reported fertility of women (as of 1984) for the same cohort of women could provide information about respondent recall error. This issue is taken up below. 19 Reasons for the shift in the distribution of education There are several reasons why one might observe increases in reported education for the same cohort of women across each pair of surveys, even in the absence of a change in actual educational attainment. First, women may mis-report their age: in order for cohort-specific education rates to rise in a context of increasing education over time, younger women would have to report themselves as being older than their actual age. Furthermore, in order to affect comparisons of the pairs of data sources, the propensity to over-report one's age would also have to be increasing over time. Finally, this study pointed out above that the age distributions in the pairs of surveys seem to be consistent with each other although it does appear that young and old women were mis-classified so that they were excluded from the survey; this should not have affected the women in the 25 to 40 age range -the women on whom we have focussed. Age mis-reporting therefore seems like an unlikely candidate to explain the observed differences in education. As women get older, however, they may simply report themselves as being better educated. Once again, in order for this to explain the preceding results, this propensity to mis-report would have to be an increasing function of time. That does not seem too unreasonable in a society with dramatically rising educational attainment as older women may seek to conform more closely to the average for the country. It has been remarked earlier that in Zimbabwe, where there is an additional source of information on education by cohort, there is very little difference between cohort specific reported education in the 1988 DHS and the 1987 Intercensal Demographic Survey. But what is the evidence in other countries? Unfortunately, as far as the authors are aware, there are no nation-wide regularly repeated cross-section data sources available for any African country and so they rely on data from Taiwan and the United States. The Current Population Survey is collected annually in the United States; using the March waves from 1980 through 1990, the authors have calculated the reported years of education for white men and women born in the year 1940 and in 1945; they are displayed in Figure 1. For men, there is no evidence that reported education has risen for these cohorts and for women there is some tendency towards upward drift although the biggest gap is only about a 3 percent increase in reported schooling which is not close to the magnitude of the change observed in Botswana or Zimbabwe. '9 Perhaps in a society that is changing more rapidly, regression towards the mean will be more apparent. Figure 2 presents reported education for women in Taiwan using a similar data source, the Personal Survey of Income Distribution, for the same period. Sample sizes are smaller than in the US data and so we examine women born between 1940 and 1944 in addition to the 1945 to 1949 cohort. There is no evidence 19. The cell sizes lie between 700 and 1100 and the standard errors are all very close to 0.1. In a bivariate regressioii of educationi on time, the slope is zero for both cohorts of men; for women, the estimated drift is 0.03 years of schooling in each survey year. 20 Figure I Mean years of reported education of males and females in United States Current Population Surveys, March 1980 - March 1990 Males Females 13.4 13.1 Birth year: 19.15 13.2 ,- - , 13.2 13 13 Blrtri year: 1915 32.6 ~~~~~~~~~~~~~~12.8------ a- 1 : \ . 12."B1""'1' 12.1 eirth year: 194C 12.1 Birtn year: 1940 12.2 12.2 12 12 ___ 60 8t1 2 63 61 65 66 87 e8 8g 90 ao et 82 B3 84 65 86 87 86 89 90 Survey year Survey year Figure 2 Mean years of education of females in Taiwan Personal Survey of Income Distribution, 1980 - 1990 .Females 7.4 7.2 . 7 Birth year..: 19JO-J4 6.8 airth years: 1945-4-A 6./ 6.4 60 81 62 63 84 85 06 87 86 e6 9o Survey year 21 of consistent upward drift in reported education in these data; if anything the trend appears to be downward.20 In the US and Taiwanese data, there does not appear to be a general tendency for reported education to rise with age, at least to the extent observed in the Botswana and Zimbabwe data. We turn next to data from Kenya where demographers have argued there is some evidence for dramatic decline in fertility. As in Botswana and Zimbabwe, a CPS and DHS were conducted in 1984 and 1988 respectively but, as noted by Blacker (1994), there may be reasons to be skeptical that these surveys are directly comparable. Based on estimated parities, he concludes the Kenyan CPS is an outlier and should be rejected. He does not, however, compare the distributions of education of women in the two surveys. It turns out that women of the same cohort tend to be better educated in the Kenyan DHS relative to the CPS. For example, among those aged 25 to 34 in 1984, the average woman reported completing 3.7 years of schooling in the CPS but 4.5 years in the 1988 DHS; for the 35 to 44 age group, reported years of education was 2.0 and 2.5 respectively. As in Botswana and Zimbabwe, the differences tend to be concentrated among the better educated: for example, in the 25 to 34 year group, 15 percent report completing Form 2 or more in 1984 but in 1988 this proportion is 20 percent. Once again part, but not all, of the difference in fertility decline can be attributed to these sampling differences. Apparently, the Botswana and Zimbabwe cases are not unique. Rather than changes in reporting behavior by women, it may be that the differences in the surveys reflect changes in the underlying population. It seems unlikely that mortality alone could explain all the differences: not only would this imply extremely high mortality rates but also differentials between the least educated and better educated for which there is no evidence. Net migration abroad is also far too small to account for the observed differences. Alternatively, there may be shifts in the population that are not adequately captured by the sampling scheme. In both Botswana and Zimbabwe, the sampling frames of the CPS and DHS are based on the 1981 and 1982 Censuses, respectively. If, as the frame ages, the more mobile are less likely to be included in the samples, then given the overwhelming evidence that the better educated are more mobile, one would expect education of the same cohort to be lower in the second survey. Yet exactly the opposite is seen. Furthermore, net migration tends to be out of the rural sector into urban areas and so the education differentials should be negative in the rural areas but positive in the urban areas. There is no evidence that this is true, either in Botswana or Zimbabwe. For example, among women in Zimbabwe aged 25 to 34 in 1984, reported education of the average urban woman in the earlier survey was 6.3 years and 7.1 in the later survey; among rural woman, reported education was 3.2 and 20. The authors are grateful to Chris Paxson who calculated the numbers underlying the figure. For a description of the data see Deaton and Paxson (1993) and Republic of China (1989). It is possible that the downward trend reflects in-migration of women from the mainland. 22 4.2 years in the two surveys. The lower levels of cohort specific fertility, as of 1984, in the later survey also persist in both the rural and urban sectors. Stratification on region leads to the same conclusion in terms of both education and fertility.2' In order to update the Census based sampling frame, for all four surveys a new listing of households was drawn up in each cluster and then households were randomly drawn from those listings. One possible explanation for the education differentials across the pairs of surveys might be that the drawing was not random, or that the listings were not complete, with better constructed dwellings (where the better educated live) being more likely to be included in the samples.22 If this is a widespread problem, then as frames are updated, there should be discrete changes in the underlying population. Furthermore, the education drift should be common across cohorts. Figures 1 and 2 provide no evidence for this in either Taiwan or the United States: in fact, in each figure the education changes for the pairs of cohorts tend to be inversely correlated with one another. Unfortunately, one cannot determine whether an explanation along these lines can explain the differences between the pairs of surveys in Botswana and Zimbabwe. Apparently, there is no simple explanation for the differences in the sample characteristics across the pairs of surveys in Botswana and Zimbabwe. The study does find, however, a similar pattern in the Kenyan CPS and DHS -but not in national samples for the United States or Taiwan. Explaining these differences is an important and complex issue, especially in the context of rapidly changing socio-economic environments. Changes in the determinants of fertility outcomes The study has argued that comparing levels of fertility outcomes in the 1984 and 1988 surveys is not straightforward. Furthermore, indirect methods that fail to take into account the differences in the sample compositions may lead to misleading inferences regarding the dynamics of fertility change in Botswana and Zimbabwe. Demographers are also interested in the determinants of fertility outcomes and, in particular, the relationship between education and the number of children ever born to a women. If differences across the pairs of surveys do not reflect changes in reporting behavior, but rather sample composition differences, then it should be possible to trace out changes in the effects of education on fertility during the 1980s. This is because sample composition differences should not bias estimated education 21. Among women aged 25 to 34 in 1984, in Mashonaland their average reported education was 3.9 years in 1984 and 4.4 years in 1988; in Matabeleland, the difference is larger (3.4 to 4.7) and in the rest of Zimbabwe it is smaller (5.3 to 5.7). It would be imprudent to make much of these inter-regional differences since they are based on small samples and the standard errors are quite large, being around 0.4 for Matabeleland and 0.15 for the other two regions. 22. Experience from other surveys suggests there is some evidence that when enumerators "randomly" pick households in a community, they are least likely to visit the poorest. 23 effects and so it may be reasonable to assume that will be comparable across the surveys at least within cohorts. Next, this assumption is examined in order to delve a little more deeply into potential sources of differences between the pairs of surveys. Fertility estimates for women in 1984 based on the birth histories recorded in 1988 are reported in the middle columns of each group in the second panel of Appendix Table 1. These are exactly the same estimates as presented in Table 3 except now they are stratified on both age and education. One might expect that as women recall beyond the last five years, their memories dim and they fail to enumerate all the births. Estimates of fertility should, therefore, be lower in the middle column (recall for the period prior to 1984 but reported in 1988) than the first column (recalled in 1984). The patterns in Zimbabwe are quite intriguing. Consider first women aged 25 through 34. Those at the bottom of the education distribution certainly do report fewer births.73 But women who have at least completed primary schooling report more births.24 A remarkably similar pattern emerges for older women: those with no education report one child less on average (a significant difference) and it is only women who have more than Form 2 schooling that report more births. Why would better educated women systematically report more births in the 1988 survey?2- We can only speculate on this question. It may be that better educated women tend to place events further back than those with little education. This seems unlikely and the evidence in Becker and Mahmud's (1984) Bangladesh validation study suggests no relationship between education and extent of back-casting. A more likely candidate, perhaps, lies in differences in the survey design: in the second survey women were asked to complete an entire birth history, whereas in the first survey they reported only the number of children born (alive and dead, by gender). It may be that the birth histories prompted better educated women to recall more events or they may have reported children born after 1984 as being born before then.26 On the other hand, it may be that less educated women failed to recall details 23. For women with some primary education this difference is significant. 24. Women who have completed Form 2 remember a significantly larger number of births prior to 1984 in the seconid survey relative to the women interviewed in the first survey. 25. The fact that respondent recall error is related to education means that comparisons of fertility as of 1984 based on the 1984 and 1988 surveys will confound respondent error and changes in sample composition. 26. There is evideiice that women (or enumerators) tended to classify children aged 5 at the time of the survey as being older (presumably to avoid having to complete the child module of the survey). It is not clear why better educated women would have reported children who were aged more than 4 in 1988 as being younger than that. 24 TABLE 6: DETERMINANTS OF NUMBER OF CHILDREN EVER BORN 1. ZIMBABWE Age group 15-24 25-34 35-44 Survey ZRHS ZDHS ZDHS ZRHS ZDHS ZDHS ZRHS ZDHS ZDHS Date 1984 1984 1988 1984 1984 1988 1984 1984 1988 Maternal education: Pre-school 0.043 -0.295 0.130 0.450 -0.451 -0.390 0.096 0.741 0.256 [0.35] [2.58] [1.02] [1.611 [2.00] [1.83] [0.22] [1.93] [0.71] Some primary -0.153 -0.308 -0.297 -0.249 -0.342 -0.464 -0.393 0.526 0.147 [1.62] [4.11] [3.51] [1.45] [2.31] [3.39] [1.37] [1.89] [0.60] Completed primary -0.224 -0.418 -0.527 -0.455 -0.460 -0.646 -0.881 -0.065 -0.237 [2.22] [5.30] [6.26] [2.32] [2.84] [4.40] [2.33] 10.171 [0.80] Form 2 -0.306 -0.688 -0.652 -1.015 -0.573 -0.807 -0.881 -0.754 -0.517 [2.85] [7.97] [7.67] [3.66] [2.58] [3.99] [1.21] [1.28] [1.21] > Form 2 -0.681 -0.866 -1.081 -1.851 -1.806 -1.707 -3.106 -0.818 -1.818 [6.04] [11.051 [12.95] [5.62] [7.67] [8.84] [4.61] [1.45] [3.96] # obs 1071 1584 1861 862 1096 1268 486 632 782 R-squared 0.49 0.47 0.51 0.26 0.23 0.29 0.15 0.10 0.11 F(educ) 11.1 31.3 71.4 9.8 12.1 16.2 5.0 2.7 4.4 F(all covs) 128.1 175.0 244.4 37.2 41.6 65.6 10.6 8.6 12.3 2. BOTSWANA Age group 15-24 25-34 35-44 Survey BFHSI BFHS2 BFHS2 BFHSI BFHS2 BFHS2 BFHSI BFHS2 BFHS2 Date 1984 1984 1988 1984 1984 1988 1984 1984 1988 Maternal education: Pre-school -0.127 0.048 -0.237 0.247 0.362 0.364 0.771 -0.005 -0.372 [0.85] [0.34] [1.54] [1.07] [1.75] [1.69] [2.31] [0.021 [1.17] Some primary -0.198 -0.036 0.057 0.132 0.186 0.178 0.216 0.184 0.001 [2.80] [0.64] [0.92] [0.97] [1.47] [1.59] [0.74] [0.69] [0.01] Completed primary -0.142 -0.211 -0.127 -0.131 -0.304 -0.308 -0.627 -0.171 -0.521 [2.36] [4.33] [2.29] [0.94] [2.34] [2.93] [1.34] [0.36] [1.83] Form 1-3 -0.390 -0.296 -0.261 -0.899 -0.703 -0.826 -0.757 -0.576 -0.869 [5.74] [5.78] [4.61] [5.16] [4.30] [6.85] [1.16] [1.09] [2.08] > Form 3 -0.689 -0.682 -0.704 -1.209 -1.445 -1.438 -0.769 -1.955 -2.000 [5.581 [9.72] [8.46] 13.41] [7.22] [8.94] [0.68] [4.02] [5.24] # obs 1337 1881 1895 1023 1178 1524 553 534 728 R-squared 0.45 0.49 0.41 0.22 0.27 0.25 0.06 0.08 0.09 F(all educ) 10.8 24.3 24.3 9.2 17.4 28.9 2.3 4.1 6.7 F(all covs) 135.0 228.9 162.2 35.3 55.3 62.3 4.5 5.4 8.7 Note: See Table 3. [t statistics] in parentheses. F() is F statistic for joint significance. Regressions include controls for mother's age and location. 25 on every live birth and so revised their number of children ever born downwards. Without more evidence, we cannot distinguish these hypotheses. In order to understand the implications of these differences for inferences regarding the changing effects of education on fertility in Southern Africa, Table 6 presents the data slightly differently. The estimated effect on the number or children ever born to a woman of her highest level of educational attainment is reported for each age cohort (controlling for age and whether the woman lives in an urban area). The first column of each panel is based on the 1984 survey data, the third column on the 1988 data and the middle column is based on the 1988 survey back cast to reflect age, fertility and education as of 1984. If the only differences between the pairs of surveys are due to sample composition, then the estimated education effects in the first two columns of each panel should be the same. They are not. Generally speaking, for all age groups and both surveys, fertility and education tend to be negatively correlated but the magnitude and significance of this correlation varies dramatically between the surveys. For example, taking women aged 35 to 44, according to the 1984 survey, the fertility of women who have completed primary school or more is significantly lower than women with no schooling. By 1988, this difference is significant only for women with at least Form 2 education. Furthermore, the effect of education on fertility has apparently been considerably reduced during these four years.27 Comparing these estimates with those based on the ZDHS dated as of 1984 suggests that both of these inferences may be wrong. According to these data, women with at least Form 2 schooling had no fewer births than those with no schooling in 1984 but by 1988 there was a dramatic increase in the impact of education. Furthermore, based on the 1988 data, in 1984 women with less then seven years of schooling (had not completed primary school) had borne significantly more children than women of the same cohort with no education. Thus, comparing the second and third columns, we might infer that the impact of education on reducing fertility has actually increased during the four years between the two surveys. In Botswana, the patterns are broadly similar, although perhaps not quite as stark. Returning to Appendix Table 1, women at the bottom of the education distribution tend to recall fewer births2x and the better educated are inclined to recall more births (prior to 1984) in the second survey. The 7 per cent of women at the top of the education distribution (greater than Form 3) are an anomaly as they report far fewer children born prior to 1988 than in 1984; indeed, women aged 35 to 44 in the second survey report one fewer children born by the first survey date. The same point is abundantly clear in Table 6. Consider women aged 35-44 with at least Form 3 27. For examnple, women with at least Form 2 education had 3.1 children fewer than women with 11o education in 1984 but only 1.8 children fewer in 1988. 28. Apart from those with pre-school education in the 25 to 34 age group although this difference is certainly niot siginificant. 26 schooling: comparing the 1984 and 1988 estimates, the impact of schooling on reducing fertility has risen dramatically; comparing the estimates based on BFHS-2 in 1988 and 1984, all of this observed change appears to be due to sample differences. The evidence indicates that not only might comparisons of levels of fertility outcomes based on the 1984 and 1988 data be misleading but also that comparisons of their determinants are also complicated by differences between the pairs of surveys. It also suggests that these differences do not reflect solely sample composition but that there is additional (unobserved) heterogeneity. Understanding the extent to which the levels and correlates of fertility have changed in Botswana and Zimbabwe is not so straightforward. 27 5. CONCLUSIONS For many countries, the Contraceptive Prevalence and Demographic Health Surveys are among the first nation-wide demographic and socio-economic surveys made available for research at the primary level. There can be little doubt that the CPS and DHS, in general, have made important contributions to the understanding of demographic processes in developing countries. In Botswana and Zimbabwe, it has only been with the collection and release of these high quality surveys that analysts have been able to examine the evidence regarding fertility decline in both countries. It is clear that there have been fertility declines in both Botswana and Zimbabwe during the mid-1980s. What is less clear, however, is the magnitude of those declines. Comparisons of aggregate fertility estimates based on the 1984 CPS and 1988 DHS data appear to have resulted in over-estimation of the rate of decline since it can, in part, be attributed to differences in sample composition and survey methodology. Apparently, the evidence for dramatic reductions in fertility during the mid-eighties in Botswana and Zimbabwe is not as strong as has been previously claimed and so population projections based on these estimates will probably turn out to he too low when results from the latest Censuses are released. In particular. the evidence presented in this paper indicates that, relative to the 1984 CPS, women of the same cohort in the 1988 DHS tend to be better educated. For example, Batswana women age 25 to 44 in the 1984 survey report themselves as having completed half a year less schooling than the same cohort of women in 1988. The difference is significant in both Botswana and Zimbabwe and the evidence suggests that ic reflects sampling differences between the CPS and DHS. This fact has important implications for the conventional wisdom regarding the rate of demographic change in both countries. Among women aged 25 to 34 in Zimbabwe in 1984, between 20 and 50 percent of the observed fertility decline can be attributed to differences in education across the surveys, between 20 and 30 percent of the decline among women aged 35 to 44 in Botswana can similarly be explained. For the other cohorts, the discrepancies between the surveys are smaller. Sinceeducation is positively correlated with contraceptive prevalence, simple comparisons of aggregate data from the pairs of surveys is likely to over-estimate increases in prevalence. Thus, the argument, which some have made, that fertility must have declined dramatically because contraceptive prevalence rose dramatically is far from convincing. Similarly, part of the observed increases in child survival might also be attributed to differences in the surveys although this comparison is complicated by differences in both the manner questions were posed in the surveys and also the education levels of women in the two samples. In general, it is not obvious that inferences on the magnitude of the time series trend in demographic outcomes in Botswana and Zimbabwe should be based solely on the CPS and DHS data. As the final reports of both the 1988 BFHS-II and ZDHS recommend. prudence is needed in interpreting these data. It is likely that it will only be with the release of more socio-demographic data at the primary level that researchers and policy makers can hope to understand the complex process of social change taking place in Southern Africa. 29 REFERENCES Ainsworth, Martha, Kathleen Beegle and Andrew Nyamete. 1992. "The Impact of women's human capital on fertility and contraceptive use in sub-Saharan Africa". Policy Research Department, World Bank, Washington, D.C. Arnold, Fred. 1992. "An assessment of data quality in the Demographic and Health Surveys". Proceedings of the World Conference on the Demographic and Health Surveys, IRD/Macro, Maryland. Becker, Stan and Simeen Mahmud. 1984. "A validation study of backward and forward pregnancy histories in Matlab, Bangladesh". WFS Scientific Reports, No. 52. International Statistical Institute, Voorburg, Netherlands. Blacker, John. 1994. "Some thoughts on the evidence of fertility decline in Eastern and Southern Africa", Population and Development Review, 20(1):200-205. Brass, William, et al. 1968. The Demography of Tropical Africa, Princeton University Press, Princeton, N.J., USA. Boohene, Esther and Thomas Dow. 1987. "Contraceptive prevalence and family planning program effort in Zimbabwe". International Family Planning Perspectives 13(1): 1-7. Central Statistical Office and Institute for Resource Development/Macro Systems Inc. 1989. Zimbabwe Demographic and Health Survey, 1988. Final Report. Central Statistical Office, Harare, Zimbabwe and IRD/Macro, Columbia, Maryland, USA. Central Statistical Office. 1991. Intercensal Demographic Survey, 1987, Round 1. Preliminary Report. Central Statistical Office, Harare, Zimbabwe. Central Statistical Office and Institute for Resource Development/Macro Systems Inc. 1989. Botswana Family Health Survey II, 1988. Final Report. Central Statistics Office, Gaborone, Botswana and IRD/Macro, Columbia, Maryland, USA. Deaton, Angus and Christina Paxson. 1993. "Intertemporal choice and inequality". Research Program in Development Studies Discussion Paper, No. 168, Princeton University, N.J., USA. Freedman, Ronald and Anne Blanc. 1992. "Fertility transition: An update". Proceedings of the World Conference on the Demographic and Health Surveys, IRD/Macro, Maryland, USA. Government of Kenya and Institute for Resource Development/Macro Systems Inc. 1989. Kenya Demographic and Health Survey, 1989. National Council for 30 Population and Development, Nairobi, Kenya and IRD. Columbia. Maryland, USA. Johansson, Lief. 1989. "ICDS Round I and 2: Preliminary Analysis of Demographic Measures of Change Derived from Round I data". Report of a Mission to Central Statistical Office, Harare. Statistics Sweden, International Consulting Office, ZIMSTAT 1989-17, Sweden. Manyeneng, W. G., P. Khulumani, M. K. Larson and Ann A. Way. 1985. Botswana Family Health Survey I, 1984. Central Statistics Office, Gaborone, Botswana and IRD/Macro, Columbia, Maryland, USA. Mhloyi, Marvellous. 1988. "Fertility decline in Zimbabwe". Mimeo, University of Zimbabwe. Republic of China. 1989. Report on the Survey of Personal Income Distribution in Taiwan Area of The Republic of China. Taipei, Taiwan. Rutenberg, Naomi and Ian Diamond. 1993. "Fertility in Botswana: Recent decline and future prospects". Demography, 30(2): 143-159. Rutstein, Shea 0. and George T. Bicego. 1990. "Assessment of the quality of data used to ascertain eligibility and age in the Demographic and Health Surveys". An Assessment of DHS-1 data quality, DHS Methodological Reports, No. 1. IRD/Macro, Columbia, Maryland, USA. UNESCO. 1986. Education Statistics. United Nations, Geneva, Switzerland. UNICEF. 1989. Children, Women and Development in Botswana: A Situation Analysis. UNICEF, Gaborone, Botswana. van de Walle, Etienne and Andrew Foster. 1990. "Fertility decline in Africa: Assessment and prospects". Technical Paper, No. 125. World Bank, Washington, D.C., USA. World Bank. 1989. "Zimbabwe Population Sector Report", Volumes 1-3. World Bank, Washington. D.C., USA. Zimbabwe National Family Planning Council and Westinghouse Public and Applied Systems. 1985. 1984 Zimbabwe Reproductive Health Survey. ZNFPC, Harare, Zimbabwe and Westinghouse, Columbia, Maryland, USA. 31 APPENDIX TABLE 1: REPORTED NUMBER OF CHILDREN EVER BORN Age Cohort: 15-24 25-34 35-44 1. ZIMBABWE Survey: ZRHS ZDHS ZDHS ZRHS ZDHS ZDHS ZRHS ZDHS ZDHS Date: 1984 1984 1988 1984 1984 1988 1984 1984 1988 None 1.78 1.59 1.65 4.34 4.19 4.29 7.11 6.08 6.24 [0.15] [0.11] [0.15] [0.20] [0.15] [0.14] [0.27] [0.23] [0.21] Pre-school 1.49 1.23 1.81 4.51 3.77 3.84 7.10 6.87 6.41 [0.17] [0.14] [0.20] [0.291 [0.22] [0.23] [0.46] [0.35] [0.32] Some primary 1.04 1.18 1.10 4.06 3.89 3.74 6.60 6.57 6.19 [0.06] [0.06] [0.06] [0.11] [0.10] [0.09] [0.18] [0.18] [0.16] Completed primary 1.20 0.98 0.62 3.45 3.54 3.57 6.00 5.60 5.57 [0.07] [0.071 [0.05] [0.13] [0.10] [0.10] [0.28] [0.27] [0.19] Form 2 0.58 0.39 0.45 2.87 3.37 3.25 5.31 4.46 4.87 [0.07] [0.05] [0.04] [0.18] [0.19] [0.17] [0.61] [0.39] [0.32] > Form 2 0.40 0.17 0.38 1.80 1.85 1.96 3.06 4.45 3.71 [0.05] [0.02] [0.03] [0.18] [0.13] [0.12] [0.37] [0.46] [0.39] 2. BOTSWANA Survey: BFHSI BFHS2 BFHS2 BFHSI BFHS2 BFHS2 BFHSI BFHS2 BFHS2 Date: 1984 1984 1988 1984 1984 1988 1984 1984 1988 None 1.31 1.27 1.16 3.57 3.49 3.43 5.72 5.36 5.57 [0.09] [0.07] [0.09] [0.12] [0.11] [0.10] [0.25] [0.24] [0.19] Pre-school 1.09 1.34 0.49 4.07 4.17 4.00 6.51 5.28 5.23 [0.25] [0.21] [0.18] [0.31] [0.25] [0.27] [0.31] [0.39] [0.37] Some primary 0.91 0.99 0.88 3.81 3.78 3.54 5.85 5.40 5.56 [0.07] [0.06] [0.06] [0.12] [0.11] [0.10] [0.20] [0.17] [0.16] Complete primary 0.88 0.71 0.69 3.30 3.12 2.87 4.74 4.87 4.87 [0.05] [0.04] [0.03] [0.08] [0.11] [0.07] [0.32] [0.41] [0.24] Form 3 0.51 0.52 0.54 2.39 2.53 2.31 4.49 4.63 4.50 [0.04] [0.03] [0.03] [0.10] [0.11] [0.08] [0.34] [0.33] [0.25] > Form 3 0.52 0.32 0.49 2.13 1.89 1.79 4.34 3.06 3.33 [0.081 [0.04] [0.05] [0.15] [0.121 [0.08] [0.611 [0.18] [0.18] Note: [Standard errors in parentheses]. First and third columns based on 1984 CPS and 1988 DHS resp. Middle column based on 1988 DHS evaluated at 1984; fertility calculated using birth history information. Distributors of World Bank Publications ARCENTiNA The Middle East Observer ITALY PORTUGAL CarisGHmrsch, SRL 41 Shenf Seet Licosa Comniionara Sansou SPA Livraria Portugal Ga"ena Guemes Cairm Via Duca Di Calabria, 1/1 Ria Do Carno 70-74 Flonda 165,4th Floor-Ofc. 453/465 CA5ella Postale 552 1200 Lisbon 1333 Buenc Aires FINLAND 501ZS Firenze AUSTRALIA PAPUANEWGUINEA, Akateemnnen Krorakxlppa JAPAN SAUDI ARABIA, QATAR AUSTRALIA, PAPUA NEW GUINEA P.O. Box 128 APNJarir Book Store FIJI, SOLOMON ISLANDS, SF-00101 HelsinkL 10 Eastern Book Service P.O. Box 3196 VANUATU, AND WESTERN SAMOA Hon.go 3-Chome. Bunkyo-ku 113 Riyadh 11471 DA. Infomahon Service FRANCE TokYo 648 Whitehorse Road World Bank Publications SINGAPORE. TAIWAN, Mitcham3132 66.avenued'Ina KENYA MYANMARBRUNEI Victora 75116 Pari Arina Book Sermice (EA,) Ltd. Gower Asia Pacific Pte Ltd. Q~uaran House. M4/angano Street Golden Wheel Building AUSTRIA GERMANY ro. Box 45245 41, Kallang Pudding, #04-03 Gerold and Co. UJNO-Verag Nairoi Singapore 1334 Graben 31 Poppelsoorfer Allee 55 A-1011 Wien D-iSoBonn I KOREA. REPUBLIC OF SOUTH AFRICA, BOTSWANA Pan Korea Book Corporation For sngletitles: BANGLADESH HONG KONG, MACAO Par Bo. 101. Kwangwhamun Oxford Universitt Pren Micro Industnes Development Asia 2000 Ltd. Seoul Southern Africa Assistance Society (OIDAS) 46-4BWvndhamSteet P.O. Box 1141 HouseS, Road 16 Wunrung Centre Korean Stock Book Centre Cape Town 8(80 Dhanmondi R/Area 2nd Floor P.O. Box 34 Dhaka 1209 Central Hong Kong Yeoeldo For subsor btwn aid en: Seoul InternationAl Subsmrption Servce B13anch ofic: HUNGARY P.O. Box 41095 Pine View, Ist Floor Foundation for Market Economy MALAYSIA Craighall 130 Agrabad Commercial Area Dombovan Ut 17-19 University of Malava Cooperative Johannesourg 2024 Chittagong 4100 H-1117 Budapest Bookshop. Luruted r O Box 1127. Jalan Pantau Baru SPAIN BELGIUM INDIA 59700 Kuala Lumpur Mundi-Prensa Libros. SA. Jean De Lannoy Allied Pulishera Pruvate Ltd. Castello 37 1060 Brussels 751 Mount Road MEXICO 28001 Madrid 1060 Bruwels ~~~~~~Madras - 6003002 INFOT'EC Aparrado Postal 22-860 Lubreria Intemnacional AEDOS CANADA Bmnch offs 1 40r0 llalpan, Mexico DY. Conseil de Cent. 391i 151A Diffulsdeur MortageIS[N. HerediaMarg 08009 BaNcelona ISIA Boul. dMrnBalard Estate NE ERLANDS Boucherville, Qubec Bombav - 400 038 De Lndeboom/InOr-Publikaties SRI LANKA AND THE MALDiVES J4B5SE6 'rP.O. Box 202 Lake House Bookshop 13/14 Asaf Ali Road 7480 AE Haaksbergen P.O. Box 244 1294 Algoma Road New Delhi -110 002EW EALAND ner Mawatha OtB3Wa 17nChittaranla Avenue EBSCO NZ Ltd. Colombo2 Calcutta - 700 072 Private Mai Bag 99914 Newv Market SWEDEN CHiLE Javadeva Hostel Building Aucklard Far singe tits: Invertec ICT SA 60th Main Road, Candhinagar Fntzes Fackboksaoretaget EdAifiio INTEC Of. 201 Bangalore - 560 009 NIGERIA Regeningsgatxn 12. Box 16356 Santifio iNTEC Of. 201Unuversity Press Limited 5-103 27 Stockholm Sanhiago 3-5-1129 Kachiguda Three Crowns Building Jericho CHINA Cross Road Private Mail Bag 5095 For suibcrpwton rden: China Financial & Economic Hyderabad -500027 Ibadan Wennergren-Wiliams AB P hmublishingi Hc Eonomk 1P. O. Box 1305 Publishing House Prarthana Flats, 2nd Floor NORWAY S-171 25 Solna B Da Fo Si Dong lie Near Thakore Baug, Navrangpura Narvesen Information Center Beijing Ahmedabad - 380 009 Book Departnent SWlTZERLAND COLOMBIA P.O. Box 6125 Etterstad For single title: Infoenlace Ltda. Patala House N-0602 Oslo 6 Librairie Pavot Apartado Aereo 16-A Ashok Marg Ca-e poLsaln3212 Bogota DoA Lucknow -226001 PAKISTAN CH1002Lausnne begota D.E ~~~~~~~~~~~~~~Mlnza Book Ag;ency Central Bazaar Road 65, Shahrah-e-Quaid-e-Azam For subscrrtion oden: COTE D'iVOIRE 60 Batal Nagar P 0. Box No. 729 Librairie Payot Centre d'Edibon et de Diffusion Nagou r 440 010 Wore S4000 Service des Abonnements Africaunes (CEDA) 'Caste poa ule 331i2 04 B.P. 541 INDONESIA PERU CH 1002 Lausanne Abid jan 04 Plateau Pt Indira Lumuted Editonal Desarrollo SA Jalan Borobudu r 20 Apartado 3824 THAILAND CYPRUS RPO. Box 8 Linal Central Department Store Center of Applied Reslearch Jakiw 10320 306 Silom Road Cyprus Ciolege PHILiPPINES Bangkok 6 DiogenesSteet.BEngomu :RAN International Book Center PiBox2006 Kowkab Publishers Suite 1703, Citviand 10 TRINIDAD &TOBAGO, ANTiCUA Nicosa P.O. Box 19575-511 Condominium Tower I BARBUDA. BARBADOS, DENMARK Te,ran Ayala Avenue, H.V. dela DOMINICA, CRENADA. CUYANA, SamfundsUtteRatKur Costa Extension JAMAICA, MONTSERRAT, ST. SamfundsLetteratur IRELAND Makati, Metro Maula KITS & NEVIS, ST. LUCIA. Rosenoerne Ales 11 CovemmentSupplies Agenc ST. VINCENT & GRENADINES DK-1970 Prederiksberg C 4-S Harcourt Road POLAND Systematics Studies Unit Dublin 2 nternational Publishing Service #9 Watts Street DOMINICAN REPUBUC Ul Piekna31/37 Curepe Editora Taller, C. por A. ISRAEL 10677 Warzawa Tnnidad Wes Indiest Restauracidn e Isabel lA Catolica 3D9 ISoRm7t Literature Ltd ApartadodeCorreos 2190P -O. Box 56055 For siiscrition ordem: UNITED KINGDOM Santo Domingo Tel Aviv 61560 IPS Journuas Microminfo Ltd. UlI. Okrezra 3 P.O. Box 3 EGYPT, ARAB REPUBLIC OF 02-916 Warsawa Alton, Hampshire CU34 2PG Al Ahram Engand Al Glaa Steet Cairo Recent World Bank Discussion Papers (continted) No. 231 Agricultural Extension in Africa. Aruna Bagchec No. 232 Telecornmmnntications Sector Reforn in Asia: Tou'ard a Newv Pragmatism. Pcter L. Smith and Gregory Staple No. 233 Land Reforn an d Farm Restructuring in Russia. Karen Brooks and Zvi Lerman No. 234 Population Growth, Sl fcing Cultivation, and Untstustainable Agricultuiral Development: A Case Study itn Madagascar. Andrew Keck, Narendra P. Sharma, and Gershon Feder No. 235 The Design and Administration of Intergovertvnental Transfers: Fiscal Decentralizatior in Latin America. Donald R. Winkler No. 236 Puiblic and Private Agricuiltutral Extension: Beyond Traditional Frontiers. Dina L. Umali and Lisa Schwartz No. 237 Indonesian Experience uitli Financial Sector Reforn. Donald P. Hanna No. 238 Pesticide Policies in Developing Coun tries: Do They Encourage Excessive Use?Jumanah Farah No. 239 Intergov'ernment Fiscal Relations in Indonesia: Issnes and Reforn Options. Anwar Shah and Zia Qureshi No. 240 Mllanaging Reduindancy it Overexploited Fisheries. Joshua John No. 241 Instituitional Change and the Public Sector in Transitiotnal Economies. Salvatore Schiavo-Campo No. 242 Africa Can Compete!: Export Opportumnities and Challengesfor Gannents and Home Products in the U.S. Market. Tyler Biggs, Gail R. Moody, Jan-Hcndrik van Leeuwen, and E. Diane White No. 243 Liberalizing Trade in Services. Bemard Hoekman and Pierre Sauvi No. 244 Womnen in Highler Eduication: Progress, Constraints, and Promising Initiatives. K. Subbarao, Laura Raney, Halil Dundar, andJcnnifer Haworth No. 245 V4hat We Knor' About Acquiisition of Adult Literacy: Is There Hope? Helen Abadzi No. 246 Formulating a National Strategy on Infonnatioti Technology: A Case Study oJ India. Nagy Hanna No. 247 Improving the TransJer and Use of Agricultural Itffonnatioti: A Guide to Infornnation Technology. Willemn Zijp No. 248 Outreacl and Sustainability of Six Ruiral Finance Institutions in Sub-Salaran Africa. Marc Gurgand, Glenn Pederson, and Jacob Yaron No. 249 Popuilation and Inicome Change: Recent Evidence. Allen C. Kelley and Robert M. Schmidt No. 250 Subhmission anid Evaluation of Proposals fr Private Pou.er Genieration Projects in Developing Coluntries. Edited by Peter A. Cordukes No. 251 Supply atid Demanidfor Finianice of Small Enterprises in Ghlana. Emest Arycetey, Amoah Baah-Nuakoh, Tamara Duggleby, Hemamnala Hettigc, and William F. Steel No. 252 Projectizing the Governance Appro ach to Civ'il Service Reforn: An Instituitiotnal Environment Assessmenttfor Preparing a Sectoral Adjustment Loan in the Ganibia. Rogerio F. Pinto with assistance from Angelous J. Mrope No. 253 S,nall Finns infornally Financed: Studies fromi Bangladesh. Edited by Reazul Islam, J. D. Von Pischke, and J. M. dc Waard No. 254 Indicators for Monitoring Poverty Reduction. Soniya Carvalho and Howard White No. 255 Violence Against W omen: The Hidden Health Buirdeni. Lori L. Heisc with Jacqueline Pitanguy and Adriennc Gemain No. 256 Women's 1-calthi and Nutritioni: Making a Difference. Anne Tinker, Patricia Daly, Cynthia Green, Hccn Saxenian, Rama Lakshminarayanan, and Kirrin Gill No. 257 Improving the Quiality of Primary Eduication in Latin Atmerica: Tou'ards the 21st Centuiry. Lawrence Wolff, Emesto Schiefelbein, andJorge Valcnzuela The World Bank Headquarters European Office Tokyo Office _1 1818 H Street, N.W. 66, avenue d'16na Kokusai Building Washington, D.C. 20433, U.S.A. 75116 Paris, France 1-1 Marunouchi 3-chome Chiyoda-ku, Tokyo lOOjapan Telephone: (202) 477-1234 Telephone: (1) 4(0.69.30.00 Facsimile: (202) 477-6391 Facsimile: (1) 40.69.30.66 Telephone: (3) 3214-5001 cs Telex: MCI 64145 WORLDBANK Ielex: 640651 Facsinmile: (3) 3214-3657 MCI 248423 WORLDBANK Telex: 26838 Cable Address: INTBAFRAD WASHINGTONDC 0 12993 PH 100 0- 8213 -2993 -6 How Fast i s Fert i lity 4000014 06 7 $6 .95 g ISBN 0-8213-2993-6