ABSTRACT A MICRO-ECONOMIC ANALYSIS OF THE DETERMINANTS OF HUMAN FERTILITY IN RURAL NORTHEASTERN TANZANIA By James Edward Kocher The purpose of this study is to analyze the relationships between socioeconomic characteristics of women in rural Africa and their fertility, and to quantify the absolute and relative importance of the major determinants of fertility, particularly those affected by socioeconomic change and development. The data were collected in late l973 by interviewing the adult members of about 1500 households in four rural areas of northeastern Tanzania. These four areas were part of the sample of the l973 National Demographic Survey of Tanzania. Two of the areas are in Moshi district in Kilimanjaro region and the other two are in Lushoto district in Tanga region. The received theory of the determinants of fertility, and especially the theory as refined by the so-called Chicago school of economists, assigns a dominant role to demand variables (incomes, prices, preferences, and the opportunity cost of parents' time) in determining fertility. In contrast to received theory, the reformula- tion of the theory of the determinants of fertility in Chapter II assigns a dominant role to supply variables in the early stages of development while demand characteristics dominate only in later James Edward Kocher stages. This reformulation postulates that socioeconomic transition affects fertility through a set of changes which initially cause the average number of surviving children (supply) to rise, somewhat later bring about a decline in the average number of children desired (demand), and eventually these changes cause parents in their later years of childbearing to find themselves in excess supply rather than excess demand situations. Continued development causes relative prices and preferences to shift further against children vis-a-vis other con- sumer goods and the absolute amount of excess supply of children (per parent) typically continues to increase until parents are motivated to try to prevent further births. Survey data indicate that considerable change and development occurred in the study areas between about 1900 and l973, but that there are now substantial socioeconomic differentials both among and within the areas. The theoretical framework indicates that some of these differentials should cause differentials in number of children born, number surviving, and number desired. Separate supply and demand models of number of children born are constructed. Multiple regression analysis is the principal tech- nique used to test the efficacy of both supply and demand models. Independent variables in the basic supply model are woman's current age, her age at marriage, the average number of months she breastfed, her number of deceased children, and whether or not she is married to a polygynous husband. Independent variables in the demand models are various proxies for household income (both current and permanent-- i.e., wealth), relative prices of children, and relative preferences for children. Multiple regression is also used to test models of the James Edward Kocher determinants of the survival rate among a woman's children, her age at first marriage, and her reported duration of breastfeeding. The models are tested at the aggregate, district, and area levels. Results of these analyses show that as expected, supply vari- ables dominate while demand variables have as yet had very little to do with levels, differentials and changes in fertility in these four rural areas. The supply models at the aggregate level produce relatively large R25 and high levels of statistical significance for the indepen- dent variables. The R2 for age group 30-39 is .45 and coefficients for all independent variables have the expected signs and are statistically significant at the .01 level. R2 for age group 20-29 is .56 and coefficients for all independent variables except the dummy for poly- gyny have the expected signs and are significant at the .0l level. Several formulations of the demand models are tested using various combinations of independent variables, but the R25 for all formulations are very low and few of the coefficients are statistically significant although most signs are in the expected directions. For age group 20-29 all R25 at the aggregate level are below .04 and they are all below .02 for age group 30-39. Further analysis of this data--primarily by the use of discri- minant function analysis--shows, however, that demand detenninants are also changing as a consequence of socioeconomic development but that the expected positive relationship between income and desired number of children is weak or non-existent while the positive effect of development on the relative price of children seems to be rather strong. Thus, the effects of rural develOpment in the study areas include a rise in the average number of surviving children and a James Edward Kocher relatively weaker decline in the average number desired with the net effect on fertility to date depending on the mix of the changes in the supply variables. A major policy implication of the analysis is the importance of specifying the stage of socioeconomic and demographic change or transition which the target population is in before deciding what types of policy interventions may be most suitable for inducing fertility decline. Policies which affect appropriate changes in supply variables can cause desirable but modest changes in both fer- tility and number of surviving children in the short run, but large and sustained declines in fertility require large reductions in demand for children which will result only from substantial devel0pment and major improvements in living conditions over the long run. A MICRO-ECONOMIC ANALYSIS OF THE DETERMINANTS OF HUMAN FERTILITY IN RURAL NORTHEASTERN TANZANIA By James Edward Kocher A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Agricultural Economics T976 ©Copyright by JAMES EDWARD KOCHER 1976 ACKNOWLEDGMENTS The study on which this thesis is based was directed by me from its inception and hence any shortcomings are my responsibility alone. However, many people and institutions made significant contri- butions, some of which were critical to the success of the under- taking. My deep gratitude cannot be adequately conveyed by the brief acknowledgments below. The study benefitted from four sources of funding. The Popu- lation Council provided a grant to support data collection and initial data processing and analysis in Tanzania. This was supple- ' mented by funds from the Bureau of Resource Assessment and Land Use Planning (BRALUP) of the University of Dar es Salaam in support of the work in Tanzania. The Interdisciplinary Communications Program of the Smithsonian Institution provided funds during the period October l975 through May 1976 in support of further data analysis and writing. I benefitted from an appointment as a graduate research assistant in the Department of Agricultural Economics at Michigan State University during the Summer of 1976 and during my earlier graduate studies. During the period August 1972 through July 1975 I was a staff member of the Population Council and a Research Fellow in the Bureau of Resource Assessment and Land Use Planning (BRALUP) at the Univer- sity of Dar es Salaam. I am particularly grateful to my colleagues at the Population Council for their encouragement and support during ii the two years I was a member of the New York staff and the three years I was in Tanzania. I also greatly appreciate the support and coopera- tion given by my colleagues at BRALUP and the University of Dar es Salaam. I feel deeply indebted to both the Population Council and the University of Dar es Salaam for the opportunity to live and work in Tanzania. The field work and part of the data processing were carried out in conjunction with the 1973 National Demographic Survey (NDS) of Tanzania on which I also worked; the N03 was a joint undertaking of BRALUP and the Bureau of Statistics of the Government of Tanzania. My appreciation goes to the staff of the Bureau of Statistics for their cooperation and assistance. Numerous colleagues and friends in Tanzania and elsewhere gave valuable assistance at various stages of the work. Those in Tanzania include Professor A. C. Mascarenhas, Director of BRALUP; Mr. J. J. Mpogolo, Commissioner of Statistics; and Professor R. A. Henin, Dr. Douglas Ewbank, Mr. Tom Zalla, Mr. Ian Thomas, Mr. E. K. Sekatawa, and Mr. S. Jiwani, all of the University of Dar es Salaam. Messrs. Doamsi, Ngallaba, Kalembo, Lyakura and Mbura, all staff members of the Bureau of Statistics (Government of Tanzania), were helpful colleagues and warm friends. I am grateful to the members of the interviewer teams for their diligence and hard work and for the friendships we formed. My deepest gratitude is reserved for Dr. Carl Eicher who went far beyond his responsibilities as thesis advisor and major professor in helping shape my professional development. I greatly appreciate the support and assistance provided by the other members of my thesis committee--Drs. Derek Byerlee, Carl Liedholm and Robert Stevens--and also the assistance and cooperation of Pam Marvel and Paul Winder of the Programming Unit of the Department of Agricultural Economics. Others who made helpful suggestions at various stages in the work include Drs. Lee Bean, Donald Heisel, Paul Schultz, Frederic Shorter, and George Simmons. Finally, my wife Sally showed great patience throughout the time I spent as a student and also gave indispensable assistance and spent long days in helping to organize and supervise the field work in Tanzania. My son, Keith, who was born in Moshi in September 1973 during the field work, made Tanzania an even happier place to live and work. My son, Craig, who was born in November 1975 when I was beginning to write this thesis, kept me company during many long nights and bolstered my occasionally lagging spirits. iv TABLE OF CONTENTS LIST OF TABLES . LIST OF FIGURES. LIST OF MAPS. Chapter I. INTRODUCTION Background to the Study . Objectives of the Study . Chapter Organization . II. A REFORMULATION OF THE THEORY OF THE DETERMINANTS OF FERTILITY. . . . . . . . . Introduction. . Demand for and Supply of Children. Determinants of Demand . . Determinants of Supply . . Other Formulations to be Tested III. DESCRIPTION OF THE STUDY AREAS IN NORTHEASTERN TANZANIA . . . . . . . . . Location, Area, and Population Size . Sample Selection Some Agricultural and Settlement Characteristics: Tribe, Religion, and Literacy in the Study Areas. IV. RURAL DEVELOPMENT IN THE STUDY AREAS The Meaning of Rural DevelOpment. Some Comments on the Measurement of DevelOpment : Descriptions of Agricultural Changes. . Wealth, Incomes, Consumption, and Investment . Honsing. Crops and Livestock. Page viii xii xiii VOW-H 44 44 53 55 60 61 64 67 68 7O Chapter VI. VII. VIII. Occupations. . Modern Consumer Goods . Agricultural Equipment . Religion, Education, and Health. The Spread of non-Indigenous Religions. Educational Attainment . Education and Religion . Health Summary. ANALYSIS OF THE DETERMINANTS OF SUPPLY OF CHILDREN BORN. . . . . . . . . . Specification of the Basic Supply Functions. Information on Some of the Variables . . Regression Results at the Aggregate Level Regression Results at the District Level. Regression Results at the Area Level . ANALYSIS OF DETERMINANTS OF SURVIVAL RATE, MARRIAGE AGE, AND LENGTH OF BREASTFEEDING . Specification of the Models . Descriptions of Some of the Variables. Regression Results for Survival Rate of Children : Regression Results for Age at Marriage Regression Results for Duration of Breastfeeding : Comments on the Effects of Woman' 5 Education ANALYSIS OF DEMAND FOR CHILDREN Specification of the Basic Models . Descriptions of the Variables Regression Results . . . Results of the Discriminant Analysis . EFFECTS OF RURAL DEVELOPMENT 0N FERTILITY . The Impact of Rural Development on Supply Variables and Fertility. . Age at Marriage Number of Deceased Children Duration of Breastfeeding. Polygyny. vi Page 78 80 82 82 84 85 88 9O 93 97 97 99 107 115 118 122 122 125 130 133 137 139 143 143 149 151 159 173 174 175 1765 1753 181 Chapter IX. SUMMARY AND IMPLICATIONS . Summary . . . . . Policy Implications . APPENDICES Appendix 1. DESCRIPTION OF DATA COLLECTION AND PROCESSING . 2. THE QUESTIONNAIRES . 3. CODING INSTRUCTIONS FOR THE SOCIOECONOMIC QUESTIONNAIRE . . . . . . 4. INTERVIEWER'S AND SUPERVISOR'S MANUALS FOR THE SOCIOECONOMIC QUESTIONNAIRE . . . REFERENCES . Rural DevelOpment and Changes in Demand Characteristics. . . . . Summary Observations. vii Page 181 186 190 190 193 199 221 235 250 270 Table 10. 11. 12. LIST OF TABLES Tribal and Religious Composition of the Study Areas: Percentage Distributions for all Women and Husbands of Currently Married Women, 1973. .' Percentages with Some Formal Education, by Age and Sex, in 1973 Types of Housing in l973 . Index of Housing Quality in 1973 Cr0ps and Livestock: Percentage of all Households which Reported Harvesting the Crop Listed or Owning One or More of the Types of Livestock Listed, 1973 . Estimated Average Shilling Values (per Household) of Thirteen Cr0ps Produced and Sold in the Twelve Months Preceding the 1973 Survey. Indicators of Estimated Distributions of Shilling Values of All Thirteen Crops Produced, All Thirteen Crops Sold, and Coffee Produced and Sold in the Twelve Months Preceding the 1973 Survey Major Occupation Categories: Percentages of Husbands and Ever-Married Women in Each Category in 1973. The Possession of Selected Consumer Goods in 1973. The Possession of Agricultural Equipment in l973 . Highest Educational Attainment: Percentages in Each Years-of Schooling Category as of 1973. Differentials in Educational Attainment by Religion as of 1973 . . . . . . . . . . viii Page 56 58 69 71 72 74 76 79 81 83 86 89 Table 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. Medical Care During Pregnancy: (A) Percentage of Women Who Visited a Clinic at Least Once During the Final Three Months of Their Most Recent Pregnancy, and (B) Percentage of Women Who Delivered their Most Recent Baby in a Hospital. Averages and Standard Deviations for Numbers of Children Born per Woman, Controlled by Women's Current Age Group Averages and Standard Deviations for Reported Age at Marriage and Reported Duration of Breastfeeding for Currently Married Women, Married Once Only, Controlled by Current Age Group. Total Numbers of Women in Each Age Group and Percen- tages Who are Currently Married and have been Married Only Once Reported Incidence of Polygyny: (A) Percentage of Husbands Currently Polygynous, (B) Percentage of the Fathers of All Husbands Who Were Polygynous (By Current Age of Husband), and (C) Percentages of Women Who Are Currently Married to a Polygynous Husband. . . . . . Aggregate Level Regression Results for Model of Determinants of Number of Children Born, Women (Currently Married, Once Only), Ages 20-49, 40-49, 20- 29, and 30- 39. . . . . . . . Aggregate Level Regression Results: Inclusion of Geographic/Ethnic Dummy Variables (L-Dummies for Lushoto), Compared to Basic Models (Table 18), for Women (Currently Married, Once Only) Ages 20-29 and 30-39 . . . . . . . . . . . . . District Level Regression Results for Model of Deter- minants of Number of Children Born, Women (Currently Married, Once Only), Ages 20- 29 and 30- 39 Area Level Regression Results for Model of Determi- nants of Number of Children Born, Women (Currently Married, Married Once Only), Ages 20-29 and 30-39. Average Numbers of Surviving Children (and Standard Deviations) and Implied Survival Rates, Controlled by Woman's Current Age Group. ix Page 92 100 102 105 106 109 114 117 119 127 Table 23. 24. 25. 26. 27.. 28. 29. 30. 31. 32. Regression Results for Determinants of Percent Surviving. Regression Results for Determinants of Woman's Age at Marriage. Regression Results for Determinants of Woman's (Reported) Duratflw1 of Breastfeeding Percentages of Women and Husbands who want more Children, Controlled by Number of Children Currently Surviving . . . . Opinions of Women and Husbands: Percentages who think there are (A) Advantages to Large Families, (B) Advantages to Small Families, and (C) That Infant and Child Survival Prospects have been Improving in Recent Years . Comparison of R25 for Various Supply Models and Demand Models for Age Groups 20-29, 30-39, and 40-49. Aggregate Level Regression Results: Comparison of Results for Selected Demand Models of Determi- nants of Number of Children Born, Women (Currently Married, Married Once Only), Ages 20- 29 . . . District Level Regression Results: Comparison of Results for Selected Demand Models of Determi- nants of Number of Children Born, Women (Currently Married, Married Once Only), Ages 20-29 . . . . . . Area Level Regression Results: Comparison of Results for Selected Demand Models of Determi- nants of Number of Children Born, Women (Currently Married, Married Once Only), Ages 20-29 . . . . . . Averages and Standard Deviations for the Independent Variables Used in the Discriminant Analysis for Women (Currently Married, Married Once Only), Ages 25- 44, with Five or More Surviving Children (Standard Deviations in parentheses). . . Page 131 134 138 150 152 154 155 157 158 161 Table Page 33. Averages and Standard Deviations for the Independent Variables Used in the Discriminant Analysis for Men, Ages 25- 29, with Five or More Surviving Children (and Wife Under Age 45 if Monogamous) (Standard Deviation in Parentheses) . . . . . . 162 34. Discriminant Analyses Results for Women: Standard- ized Discriminant Function Coefficients and Various Indicators of the Discriminating Power of the Resulting Functions . . . . . . . . . . . 164 35. Discriminant Analysis Results for Men: Standardized Discriminant Function Coefficients and Various Indicators of the Discriminating Power of the Resulting Functions. . . . . . . . . . . . . 165 A1. Numbers of Households and Persons Included in the Study . . . . . . . . . . . . . . . . . 204 xi Figure 1. LIST OF FIGURES Hypothesized relationShips between desired and actual numbers of surviving children for two representative women in a traditional rural African society. .‘ Hypothetical trends in fertility variables associated with economic and social modernization. Hypothetical relative levels and trends in fertility variables for a typical woman who has recently completed childbearing in a transitional rural African society . . . . . . . . Comparison of two alternative patterns of changes in actual and desired numbers of surviving children during traditional and transitional stages Hypothetical illustration of the interaction between income, prices and two different sets of indifference curves in determining demand for children and other goods. . . . . . . . . . . . Hypothetical budget lines showing price effects with welfare curves relatively in favor of children . Model of hypothesized socioeconomic and demographic transition in rural Africa and the linkages to fertility change and eventual decline . xii Page 14 16 18 21 26 28 39 Map 01-500 01 LIST OF MAPS Location of Lushoto and Moshi Districts. Study areas within Moshi and Lushoto Districts ”Sketch of M1 (Kibosho study area) and Vicinity Sketch of M2 (Vunjo study area) and Vicinity . Sketch of L1 (Bumbuli study area) and Vicinity Sketch of L2 (Soni study area) and Vicinity xiii Page 45 46 49 50 51 52 CHAPTER I INTRODUCTION Background to the Study The overriding goal of the Government of Tanzania is to bring rapid improvements in the living conditions of all Tanzanians. The Government essentially monopolizes the provision of educational Opportunities, medical care and public health services such as clean water supplies. Most other Government investments and expenditures are made with the expressed intention of enhancing the economic potential of the people and the country. Reliable infonnation about demographic characteristics of the population is required if planning in many sectors (including education and health) is to be adequate. InfOrmation is needed not only on current demographic characteristics (birth rates, death rates, age distributions, migration) but also on likely future demographic characteristics. Future demographic characteristics are determined in part by current and future social and economic conditions. Reduction of birth rates is not an explicit objective of the Government of Tanzania. However, it is an objective of the Government that women should have access to methods for achieving the interval or spacing between births which will most enhance the health and well-being of themselves and their children. Substantial Government resources have been committed toward making available maternal and child health care (MCH) services throughout the country; included as part of MCH services are family planning counseling and free provision of contraceptives. Widespread utilization of MCH services will affect both mortality and fertility rates in Tanzania. It is a policy of the Government of Tanzania to encourage research and the acquisition of knowledge about current demographic characteristics, about the relationships between socioeconomic and demographic characteristics, and about possible future demographic characteristics. One of the objectives of this study is to make a contribution to the acquisition of knowledge which is useful to the Government of Tanzania. In particular, it is the objective of this study to analyze the determinants of fertility levels and fertility changes in selected rural areas of Tanzania. In addition to its potential utility to planners and policy makers in Tanzania, this study is intended to make a contribution to the ongoing discussions and controversies in academic circles con- cerning the fertility decline process and the relationships between socioeconomic and demographic characteristics of low income countries. During the past few decades the so-called population explosion in low income countries has generated a great deal of interest about the causes and consequences of rapid population growth and about ways of reducing high rates of population growth as expeditiously as possible. It is generally recognized that the rapid fall in death rates-~rapid in comparison to the comparable fall in death rates in the Western world in the nineteenth and twentieth centuries--is the major cause of high population growth rates. It is also recognized that the only satisfactory method of reducing population growth rates is to reduce birth rates; large-scale international migration is impractical while a rise in death rates would cause more suffering and be anti- developmental. During the past century or so, fertility in high income countries declined in response to social and economic development. With few exceptions, fertility declined without benefit of government programs and/or public policies to encourage lower fertility; on the contrary, in many countries fertility declined despite efforts of government and/or Church to slow or reverse the decline. Some scholars and policy makers expect a comparable fertility decline in low income countries in response to their social and economic development. At the 1974 United Nations Population Confer- ence this view was epitomized by the slogan, "Take care of development and population growth will take care of itself." Indeed, birth rates have declined in many low income countries in recent years, in some countries by 25 to 50 percent (e.g., Chile, China, Costa Rica, Malaysia, South Korea, Sri Lanka, and Taiwan). Birth rates have declined in the more socially and economically develOped regions of most Latin American countries [Oechsli and Kirk, 1975], most of which have not had strong government anti-natal policies. On the other hand, a large number of scholars contend that birth rates are unlikely to decline much in most of the low income *world, or that if declines do occur they will come about much too slowly and too late to save these countries [and perhaps the world?-- e.g., the Club of Rome Report, see Meadows et a1., 1972] from economic and perhaps ecological disaster. This conclusion is sometimes based on comparisons of current levels of per capita incomes and other conventional indicators of economic growth and development in contemporary low income countries with the comparable economic levels which prevailed in today's high income countries during periods immediately prior to and during their fertility declines. Most contemporary low income countries cannot hope to achieve within the near future the so-called "threshold" economic levels achieved by contemporary high income countries prior to their fertility declines [Teitelbaum, 1975]. The clear implication of such comparisons is that if fertility is to decline in these countries in the near future, it requires government policies to successfully promote the value of small families and the use of contraceptives, abortion, etc.. to achieve small families. A few studies have shown, however, that it is not so much the average national levels of the various economic indicators or the presence of government anti-natal policies and programs which affect fertility levels and rates of decline as it is the extent to which ordinary families have participated in and benefitted from develop- ment. Average values of per capita incomes, for example, obscure income distribution differences among countries. Those contemporary low income countries which have experienced substantial fertility declines have relatively less inequality in their income distribu- tions than do other countries with similar (and in many cases, higher) levels of per capita income which have experienced little or no fertility decline [Kocher, 1973].1 It is clearly not the aggregate relationship between economic indicators and overall fertility that is relevant; the relevant relationships between levels of living and fertility is at the micro- or family-level. It is also obvious that if birth rates are to decline sub- stantially within the next few decades in most low income countries, and particularly in Africa, they must decline in rural as well as in urban areas. The United Nations has estimated that about 70 percent of the people in the low income world now live in rural areas; about 75 percent in Africa do so. Despite a projected increase in the urban population in the low income world from about 720 million in 1970 to about two billion in the year 2000, the rural population is also expected to increase by almost a billion and still comprise almost 60 percent of the total in 2000 [Kocher, 1973: 14, 16]. If fertility is to decline in rural areas, it will be the result of a large pr0portion of individual parents deciding that they want to have fewer births than the average number of births parents in those societies traditionally had. Although this is the obvious requirement for fertility decline, relatively few studies have been made of the determinants of fertility in rural areas, and particularly in rural areas in Africa.2 1Here income distribution is defined broadly to include not only incomes per §g_but also access to education, health care, acquisition of modern consumer goods, etc. 2One study was carried out in 1961/62 in the Sudan by Henin. Results of his analysis are described in Henin, 1968, 1969. Cald- well's numerous studies in West Africa also provide some insights into the relationships; see for example, Caldwell, 1967, l976a, l976b. Objectives of the Study This study analyzes determinants of fertility in rural areas which have until fairly recently (i.e., the beginning of this century) been traditional African communities, largely undisturbed by outside-- particularly Western--influences, but which are now in relatively early stages of social and economic--and demographic--transition and development. This study is not specifically intended to address the question as to whether fertility in rural Africa will decline spontan- eously and if so, how soon and how fast. It has the more limited objective of analyzing the determinants of fertility levels, fertility differentials and fertility changes in these rural areas as they are transformed from basically traditional societies to increasingly more developed and modern societies. Nevertheless, a better understanding of the interrelationships between social, cultural, educational, health, and economic changes and changes in mortality and fertility in rural areas may be a prerequisite to addressing questions about how best to reduce fertility and population growth rates in rural areas of low income countries. The Specific objectives of this study are as follows: (1) To develop a theoretical framework for analyzing the determinants of parental demand for and supply of children in low income rural areas. (2) To describe the salient features of the four rural areas in Northeastern Tanzania, including a description of some important aspects of develOpment during this century. (3) To test supply models of the determinants of fertility. (4) To test models of the determinants of survival rates, age at marriage and duration of breastfeeding. (5) To test demand models of the determinants of fertility, to compare their results with those obtained for the supply models, and to analyze relationships between the desire of mothers and fathers for more children ("demand") and some of their socioeconomic characteristics, particularly income, relative prices and preferences. Finally, (6) to analyze the effects of rural development in the four study areas on the fertility of women in the areas, based on results obtained in objectives 3 through 5. Chapter Organization The basic theoretical framework will be developed in Chapter II. Chapter III will provide a brief general description of the study areas focusing particularly on their locations, areas, population sizes, agricultural features and tribal, religious, and literacy characteristics. Chapter IV will describe many of the important aspects of development in these areas, both at the time of the survey (1973) and over the course of the twentieth century. A discussion of the meaning and measurement of development is given in the first part of Chapter IV. This is followed by presentation of data on and analysis of specific indicators of development, including agricultural changes, housing, crop production, livestock holdings, occupations, acquisition of consumer goods and agricultural equip- ment, religious changes, educational attainment, and medical care during pregnancy. Chapters V through VIII are the analytical chapters. In Chapter V various formulations of a supply model of the determinants of fertility will be analyzed using multiple regression. In Chapter VI, multiple regression will be used to test a model of the determinants of survival rates among children within families. Multiple regression will also be used in Chapter VI to test models of determininants of two of the independent variables in the basic supply model (analyzed in Chapter V)--age at marriage and duration of breast- feeding. Various formulations of a demand model of the determinants of fertility will be tested in Chapter VII using multiple regression. Results will be compared to those obtained with the basic supply models. Discriminant function analysis will also be used in Chapter VII to analyze the demand for children as indicated by the responses of both women and husbands to the question, "do you want more children?" In Chapter VIII the effects of rural development on fertility will be analyzed drawing on the results obtained from the various models tested in Chapters V through VII and on the development experi- ences in the study areas as described in Chapter IV. Finally, Chapter IX will present brief summary conments and some policy implications. CHAPTER II A REFORMULATION OF THE THEORY OF THE DETERMINANTS OF FERTILITY This chapter will present the theoretical framework which will form the basis for the analysis in Chapters V through VIII. The first section of this chapter will briefly summarize the received theory of the micro-economic determinants of fertility, identify some important issues and present a theoretical framework which will more adequately explain levels of and changes in fertility in low income countries. Introduction1 The received economic theory of the determinants of fertility is set within a utility-maximizing framework. It is assumed that parents are rational, that they have limited material resources, that they prefer to allocate their material resources so as to maximize 1A number of economists have made significant contributions to the theory, notably Liebenstein [1957, 1974], Easterlin [1969, 1975], Becker [1960, 1965], T. P. Schultz [1969, 1973], Nerlove [1974], and T. W. Schultz [1973, 1974]. The two volumes edited by T. W. Schultz [1973, 1974] bring together several papers and commentaries presented at a 1972 conference on the economic determinants of fertility, although most of the papers in the two collections are directed to high income, small-family countries, primarily in the United States. 10 their satisfaction (or utility) subject to their tastes or prefer- ences and to the prices of available goods (and in some formulations, subject also to available technology), and that children represent one of a large number of potential acquisitions among which they can choose in attempting to maximize satisfactions. Hence, parents' demand for children is assumed to be the outcome of the interaction of their income, relative prices of children vis-a-vis other goods and their preferences for children vis-a-vis other goods. Liebenstein [1974] identified three types of satisfaction or utility which children (potentially) provide to parents: income utility (income or wealth, net of their costs), security utility (financial and/or emotional support, particularly in parents' later years), and con- sumption utility (personal enjoyment or satisfaction from children). An important contribution of Becker and the Chicago school of economists has been the explicit recognition given to the oppor- tunity cost of time spent by parents (especially mothers) in bearing and rearing children. “Thus, an increase in permanent income (house- hold wealth) due to a rise in a parent's permanent (average lifetime) wage rate implies both a positive wealth effect on fertility together with an offsetting negative fertility effect due to an increase in the cost of parents' time required in bearing and rearing children [T. P. Schultz, 1973: 72]. Analyses of demand for children in high income countries where small family norms prevail have had considerable success in ll explaining differentials in timing and numbers of births.2 However, applications of demand models of fertility determination to societies where large family norms still prevail, and where use of contracep- tives is not widespread, have been less successful. The theoretical framework developed in this chapter suggests that this may be because in many of these societies demand for children has typically exceeded supply; hence. fertility levels and differentials are attributable to differentials in characteristics which determine the supply of children (actual number of births or number surviving) and not demand (number of children desired). In the formulation presented in this chapter, demand variables are not absent, they simply have little or no impact on fertility--at least until the number of children born reaches or exceeds the number desired. Within the theoretical framework developed in this chapter, both supply and demand characteristics of parents in rural areas are expected to change as the consequence of the socioeconomic transition from a traditional towards a modern rural society (i.e., development). The independent effect of changes which take place in supply deter- minants is hypothesized to be an increase in average number of children surviving. The independent effect of changes in demand determinants is hypothesized to be a decline in the number of children desired. However, the timing of the supply and demand changes is expected to differ. The net result is expected to be that eventually during the process of socioeconomic transition parents will typically 2See, for example, papers in T. W. Schultz, 1973 and 1974, reprinted as T. W. Schultz, 1975. 12 find themselves in excess supply situations while in traditional societies parents typically experience excess demand. The achievement of excess supply will cause parents to want to hold down future fertility; eventually this may lead to the use of modern contracep- tives, if available. Demand for and Supply of Children3 The theoretical framework for this study views the numbers of children born and surviving as the outcome of the attempts by parents to match their actual numbers of surviving children with their desired number; that is, parents are viewed as attempting to achieve an equilibrium of their supply of children with their demand for children.4 Mothers (and fathers) in rural Africa are assumed to have goals about family size and composition. These goals may be something very general such as "as many as God gives" or very specific such as "four boys and three girls." Presumably if parents do have a family size objective, at any particular point in the process of family 3Throughout much of both the theoretical and analytical portions of this thesis, fertility and its determinants will be addressed from the perspective of mothers; this is primarily a con- cession to brevity, semantics and convention, and reference to women or mothers should usually be read as a short-hand for reference to parents or mothers and/or fathers. It is not intended to imply that mothers have an exclusive or necessarily dominant role in determining fertility within families. 4The approach in this section borrows from the work of Easterlin, particularly from his 1975 paper in which he outlines a framework for conceptualizing determinants of fertility from both economic and sociological perspectives. (Also see Figure 2 in this section.) 13 formation they would be able to indicate whether they want more children than they currently have. If the number of children a woman now has equals or exceeds the number she wants; that is, if her supply exceeds her demand, presumably in response to this question she would say she wants no more children. Conversely, if the number she wants exceeds the number she has, she would say she wants more children. Figure l is an illustration of the hypothesized relationship between the desired number of surviving children and the actual number of surviving children for two different women in a rural African society, commencing from the date at which each woman first gives birth. It is assumed that both women would like to eventually have an identical number of surviving children (0 = desired number of surviving children for both) but they differ in the actual number of surviving children they have at various points in time--either due to different fertility experiences or different survival experiences among their children or both. It is further assumed that both women began their childbearing experiences at the same time (to). Figure 1 shows that both completed childbearing at the same time (t]) although this is not essential to the argument. Finally, it is assumed that the general relationships between desired number of surviving children and actual number of surviving children for these two women are broadly representative of women in rural Africa. D is made identical for both women in Figure l for simplicity only; it is not necessary to assume that all women in a traditional society would like to have an identical number of surviving children. The important feature of Figure l is the relationship between S Number of Surviving Children Figure 1. 14 ------- ('1' d- O (+- CD (+- H Time 0 = Desired Number of Surviving Children for both Woman 1 and Woman 2 ("Demand") 51 = Actual Number of Surviving Children for Woman 1 ("Supply") 52 = Actual Number of Surviving Children for Woman 2 ("Supply") to = Time of the first births of each Woman t1 = Time of the last birth of each Woman t = Time when Supply of Surviving Children begins to exceed Demand, for Woman 1 Hypothesized relationships between desired and actual numbers of surviving children for two representative Women in a traditional rural African society. 15 (actual number of surviving children) and D (desired number of surviving children). After point te, Woman 1 has more surviving children than she orginally wanted, but Woman 2 never achieves as many surviving children as she wants. Figure 1 is also drawn to show that for both women the total number of surviving children peaks at the time of the birth of what is to be the last child for each, and-- as indicated by the slight downward dip in both S1 and S2 before they become horizontal--the typical woman may experience the death of one child or more following the birth of her last. Again, however, this assumption is not necessary to the argument. Other assumptions underlying the relationships illustrated in Figure l are that the typical woman has a goal about the number of surviving children she would eventually like to have (or at least as she gets older and her number of surviving children increases, at each stage she knows whether or not she would prefer having one or more additional children) and that presumably she would prefer to have no additional children after the number of surviving children she actually has equals or exceeds the number she initially wanted. Of course, a woman's uncertainty or anxiety about possible future deaths among her children may cause her to prefer to have somewhat more births than she would otherwise want, as a hedge against the unknown future. Figure 2 is reproduced from Easterlin [1975]. Figure 2a depicts hypothesized relative relationships between demand for children (Cd), supply of children if fertility is unregulated (Cu) and actual supply of children (C) in a low income society experi- encing (over time) socioeconomic development. Cd and Cn are 16 ' individual control social control : 1 =2 ’ I 1 r’ 1 I d I I I I I 1 i l A! m h I 1 excess excess E demand 5'5 SUDD'Y .. c" (b) Legend The following definitions all refer to the total number over the reproductive career of the “representative" household: -- ---~ ~ C" is the number of surviving children parents would have in an unregulated fertility regime. Cd is the desired number of surviving children in a perfect contraceptive society. —--—- C is the actual number of surviving children. sX is unwanted children, the excess of the actual number of children over the desired number. E sR is the degree of voluntary fertility regulation, measured in terms of children averted. Source: Easterlin, 1975, p. 60, Figures 2c and 2f. Figure 2. Hypothetical trends in fertility variables associated with economic and social modernization. 17 reproduced in Figure 26 together with a curve which indicates the extent to which unregulated fertility exceeds desired fertility (Cn - Cd). To the left of point m in Figures 2a and 2b, demand exceeds supply; to the right of point m supply exceeds demand. The diagrams illustrate changes in the relative levels of actual and desired numbers of surviving children for representative women who have recently completed childbearing (although over time the social, cultural, economic, health, and educational characteris- tics of "representative women who have recently completed childbearing" are presumably continually changing). The diagrams can be viewed as representing the passage of time within a rural society (perhaps several decades); over time, women in the childbearing ages would generally move from left to right although the transition would not necessarily be made by all women simultaneously. Alternatively, they can be viewed as representing one point in time in a rural society with individual women (who have completed childbearing) scattered or grouped at various points in the transition illustrated by the diagrams. Figure 3 extends and expands upon the Easterlin model. One traditional and three transitional stages of socioeconomic and demographic transition and development are identified in Figure 3. The traditional stage (Stage 1) would characterize rural societies which have experienced little or no modernization. The average number of surviving children (S) is low relative to (later) transi- tional stages, and the desired number of surviving children (0) is relatively high. The Easterlin model is also modified in Figure 3 by the absence of a solid line for D to the left of t3. This is 18 ' ,,.,.y.'o.';i,c~.i S — - — v ' O ‘ 2-7-3 :3 " Number of 7' N"’“’HOONVVOH1AW§31§~ SDIV1V1n5 ' I‘ANVONVVMOVT Children : | 01.053330” Sc 1 . . 9.0" ' l I 1 D I . . 1 Note: I F—excese demand—a'h—excess supply—» I i 1 STAGE 1 L STAGE 2 1 Lsmom L STAGE 4 t _t t2 t 9 t4 1 mm" # 3TRANSI TION \ TIONAL ‘ ’ D = Demand: Desired Number of Surviving Children (for com- pleted family size) S = Supply: Number of Surviving Children with Unregulated Fertility (completed family size) Sc= Decline in Actual Number of Surviving Children Due to Deliberate Fertility Control VII/l], Deficit Fertility (excess of desired number of sur- viving children over actual number) Surplus Fertility (excess of actual number of sur- viving children over desired number) Amount of Deliberate Birth Prevention, measured by number of children averted r v v v v v v {0202024020301 Transition Stages t2 = beginning of Stage 2: Average Number of Surviving Children begins to Rise t3 beginning of Stage 3: Desired Number of Surviving Children begins to Decline 4 = beginning of Stage 4: Beginning of Individual Practices (contraception, abortion, etc.) to Intentionally Reduce Fertility The absence of a solid line for D to the left of t is t intended to portray the possibly indefinite nature3of desired surviving children in traditional and early transitional societies. The relevant characteristic of D in this area is that it exceeds S. Source: Derived in part from Figure 2a. Figure 3. Hypothetical relative levels and trends in fertility variables for a typical woman who has recently completed childbearing in a transitional rural African society. 19 intended to portray demand as being perhaps only vaguely specified (by parents) in Stages 1 and 2 but nevertheless exceeding supply. As depicted in Figure 3, the transition is initiated (begin- ning of Stage 2) by a rise in the average number of surviving children per woman (5). This would presumably be at least in part the conse- quence of improved health conditions and declining infant and early childhood mortality, but it could also be in part the result of social, cultural, health, economic or educational changes which cause either directly or indirectly (see Figure 7) an increase in fertility. Stage 3 is initiated by a decline in the level of demand (0) for surviving children--i.e., the desired number of surviving children or desired completed family size of the woman. D would fall because changing cultural, social, economic, educational, and health conditions alter the mother's perceptions of the advantages of having the large number of surviving children which was traditionally desired.5 D could start to decline either while it still exceeds S 5It is possible that D (desired number of children) could decline prior to a rise in S (actual number of surviving children) but it seems unlikely that this would be due to a socioeconomic transition which brought improving living conditions. This is more likely to be the consequence of a Malthusian situation in which economic circumstances drastically worsen causing parents to desire fewer children than in the past because of their inability to provide minimum sustenance for themselves and their children. In such cir- cumstances child mortality rates might even rise and both average number of children born and average number surviving decline. Nine- teenth century Ireland may provide an historical example. A fall in 0 might have occurred in Ireland following the potato famine of the 18405. The Irish population had increased from about 4 million in 1780 to over 8 million in 1840 (plus 1.75 million abroad which were either emigrants or their children) [Connell, 1965]. In the second half of the nineteenth century the population of Ireland declined to about 4 million (it is now 3 million) as a result of very high mortality rates (in the years immediately following the famine), emigration, late marriages and high proportions of adults who never 20 (as in Figures 3 and 4b) or after S has already surpassed D (as in Figure 4a).6 The diagrams illustrate the assumption that both in a traditional society and into the early part of Stage 2 (and perhaps even into Stage 3), the typical woman who has recently completed childbearing does not have as many surviving children as she would like to have; 0 exceeds S. It is possible that some women in tradi- tional rural African societies end up with more children than they originally wanted, but it is assumed that most do not. Stage 4 is initiated by a decline in the actual number of surviving children brought about by the deliberate use of contra- ceptives, abortion or other practices by the woman or her spouse to {hold down her fertility. Line Sc in Figures 3, 4a, and 4b indicates the extent to which the average number of surviving children declines due to individual fertility control. As described above, Stage 2 (the first stage in the transi- tion) begins when the average number of surviving children per woman starts to rise; 0 is assumed to still exceed S at this time. However, so long as S continues to rise and D remains either constant or declines, eventually 5 must equal 0. Obviously, until the point is reached where 5 equals or exceeds 0, a woman will still want more children. That is, regardless of the levels and direction of changes married. However, although women married very late (average age at marriage for women approached 30 years), marital fertility as meas- ured by birth intervals remained high suggesting little or no contra- ceptive practice within marriage. ‘6Figures 4a and 4b compare different sequencing of changes in D and 5. Figure 4b is identical to Figure 3; 0 starts to decline while it still exceeds S; S = 0 (point e) is located in Stage 3. In Figure 4a, 0 does not start to decline until after S exceeds 0 and e is located in Stage 2. - -....« as;z S . v 9 o r v r I I ‘ s IIVNVVIIIIOVVQVIgESsfiiés;' '- - , """ ”IfIII}:’I."‘.IIIII'OII V1971§Z Number of V / 1 u..¢,:,:,1.o,o.o.i'.m KW“. Survi ““5 | 1 ' "111." {61111.0 0‘. I (a) - ‘ .II ‘ Children 1 . ‘JVVMVVI SC S exceeds D I I l 1 "AQ‘ 'bekane I) 1 ' 1 D declines 1 ' I I c—excess demand e—excess suinply-—————p i l 1 l l 1 I J t I 1 1 i l ' l I I I .... : g. n. s I . "’.“.‘.““r_‘, - ‘- -" V 17"0'....‘ ‘ KL;'Z;ZTT Number of '/ "" " "‘«"31.'1':':'I‘:" ”1'31KT Surviving 1 ‘ 0.1,1‘.3.0,1.1.“1,§s' _- Chi ldren I I “11.13.. I”? ( b) . ; | : “tfi$ D declines i ' I . ‘ D m S 1 : ' l exceeds D I ~<—— excesfs demand—a e—excess supply—s I I STAGE 1 1 STAGE 2 1 4 STAGE 3 1 STAGE L1 t . t1 I t2 123 e t1). IRAD ’ Q D = Demand: Desired Number of Surviving Children (for completed family size) SI= Supply: Number of Surviving Children with Unregulated Fertility (completed family size) Sc = Decline in Actual Number of Surviving Children Due to Deliberate Fertility Control fiZZZZZ Deficit Fertility (excess of desired number of surviving children over actual number) WSurplus Fertility (excess of actual nmber of surviving children over desired number) - Amount of Deliberate Birth Prevention, measured by number of children averted Transition Stages 2 beginning of Stage 2: Average Number of Surviving Children begins to Rise 3 - beginning of Stage 3: Desired Number of Surviving Children begins to Decline f? I 4 = beginning of Stage 4: Beginning of Individual Practices (contraception, abortion, etc.) to Intentionally Reduce Fertility. The absence of a solid line for D to the left of t3 is intended to portray the possibly indefinite nature of desired number of surviving children in The relevant characteristic Notes: - (b) is identical to Figure 3. traditional and early transitional societies. of D in this area is that it exceeds S. Figure 4. Comparison of two alternative patterns of changes in actual and desired numbers of surviving children during traditional and transitional stages. 22 in S and 0 over time, as long as D is greater than S the woman will want to increase S which can be accomplished by either increasing the proportion of her children who survive, increasing the number born, or both. However, at the point where S starts to exceed 0 (point e in Figures 3, 4a, and 4b), the woman will start to want to hold down S. Presumably the only way she would find acceptable would be to reduce the number of children born; she would not want to increase the number who die. Disequilibria exist on either side of point e (the point at which S = D), but the two disequilibria are very different. To the left of e (where D exceeds S) a woman would definitely ngt_want to hold down fertility and would most likely want to raise it. To the right of e (where S exceeds D) she would like to hold down her fertility provided it could be accomplished in a manner compatible with her social, cultural, and economic circumstances.7 Point e can be located anywhere within Stages 2 or 3; its location is simply the coincidence of actual and desired numbers of surviving children being equal. The only conditions are that this must come after the average number of surviving children starts to rise (beginning of Stage 2) and before the beginning of the use of 7Women (in the childbearing ages) for whom D exceeds S would presumably answer "yes" to the question, "do you want more children?" Women for whom 5 equals or exceeds 0 would presumably answer "no," although their attitudes about the desirability of more children may be affected by their subjective assessments of the probability of early mortality among their currently surviving children. Responses of both women and their husbands to this question are analyzed in Chapter VII. 23 contraceptives or other individual action on the part of the woman to try to prevent more births.8 To the left of t4 (beginning of Stage 4), the number of surviving children is generally not limited by deliberate actions of the woman (or her spouse). The limits are broader social, cultural, economic, and health conditions in the community. However, to the right of point t4 the numbers of children born and surviving are limited (in part at least) by deliberate individual actions of the woman and/or her spouse to hold down fertility. (See Figure 2 where Easterlin has labeled these two areas "social control" and "individual control" respectively.) To the left of t4, actual fertility (as contrasted to the number of surviving children) may be either falling, unchanged, or rising. Some health, educational, economic, social, and cultural changes cause fertility to rise; other changes cause fertility to decline. For example, the average age at first marriage could be rising and infant and young child mortality rates could be falling (this latter would have the effect of increasing average birth intervals); these two changes would have the independent effects of lowering fertility. On the other hand, women could be weaning their children earlier (perhaps because they are using purchased breast-milk substitutes) and women could be discontinuing their adherence to , 8Women who want more children (D exceeds S) may practice birth control to help achieve a more desirable interval between births. The criteria for entering Stage 4 of the transition, however, is that S either equals or exceeds 0 and that the reason a woman is practicing birth control is that she wants no more children. 24 prolonged post-delivery sexual abstinence; both of these changes would have the independent effects of reducing average birth intervals and raising fertility. The net result could be rising, falling, or about constant fertility during the period before time t4 (the beginning of deliberate individual fertility control). Determinants of Demand In the economic theory of household or parental choice, the three basic components of the determinants of demand are income, relative prices and tastes (or preferences). If children are a normal good, a rise in parental income would produce a rise in the number of children wanted; a rise in the price of children vis-a-vis other goods would cause a decline in the number of children wanted; and a shift of tastes in favor of children relative to other goods would cause a rise in the desired number of children. However, a rise in income (due to a rise in wage rates) is hypothesized to have both a positive effect on demand for children but also an offsetting negative effect because the value of time parents spend with their children is also increased thus increasing the price of children.9 9The treatment of children as a "good" is complicated by the quantity and quality components of children. Some analysts treat quality and quantity of children as two distinct goods, substitutable in part for each other and with separate income and price elastici- ties. In the analysis in this paper I will assume that the price of children reflects prevailing community (or reference group) standards and that the price may change over time as community standards and norms and the roles of children and families change. In practice it is virtually impossible to estimate the objective net price (cost) of children, particularly in low income countries; the approach selected to handle the price component of children is therefore 25 Indifference (or welfare) curve diagrams are used below to illustrate hypothesized relationships between income, relative prices, and tastes as they determine the demand for children. An indifference curve represents a set of combinations of "x" and "y" values among which a person is "indifferent"; that is, each combination would pro- vide exactly the same amount of satisfaction as every other combina- tion on the curve. A straight line drawn from the origin would intersect ever-higher curves representing ever-greater amounts of satisfaction. It is assumed that a person wants to maximize total satisfaction but is indifferent as to the particular combination of x and y, so long as it is one of the combinations along the highest attainable curve. In trying to maximize satisfaction, one ordinarily faces two economic constraints. The first is total income; this is indicated by the budget or income line, GO in Figure 5, and can be measured as the total amount of either x or y which can be purchased with the given income (0G or DC respectively in Figure 5). The second constraint is the prices of x and y. This is indicated by the slope of the budget line, and it can be interpreted as the amount of x that one important primarily for its contribution to theoretical clarity and consistency. Analysis of both demand and supply for children is further complicated by pregnancy and sexual pleasure being joint products of sexual intercourse and possibly having very different demand func- tions. Demand for children and demand for sexual pleasure may also be partial determinants of the demand for each other. Obviously, the desire for a child may affect the demand for sexual intercourse. Demand for sexual intercourse may affect demand for children in that pregnancy may be viewed by one or both parents as interrupting the future flow of sexual pleasures. Goods Index Figure 5. 26 —————— y—— _—_ n I- — _ I | l 1 1 I 1 g 3 A 5 6 7 8 9 10 Number of cthren Hypothetical illustration of the interaction between income, prices and two different sets of indifference curves in determining demand for children and other goods. Set of indifference curves which are relatively in g favor of goods as compared to children 2 II - Set of indifference curves which are relatively in favor of children as compared to goods 2 I ch 27 could purchase by foregoing one unit of y. Given income and relative prices as represented by the budget line, a person will prefer that combination of x and y at which point the budget line is just tangent to an indifference curve because this would be the highest attainable curve given these prices and income.10 It is quite conceivable that two people in a particular society with similar incomes and facing identical prices might have somewhat different sets of indifference curves representing different tastes and preferences. As illustrated in Figure 5, the person with the W9 set of indifference curves (relatively in favor of goods) would prefer the combination of 3 children and G9 goods, while the person with the Wch set of curves would prefer the combination of 6 children and GCh goods. If the price of one of the items increases, this changes relative prices and reduces the person's income as measured by that item. In Figure 6, the relative price of children as represented by budget line GD is double that represented by line GC (i.e., total income would purchase 10 children with line GC but only 5 with line GD, although income as measured by goods has not changed). As drawn in Figure 6, as the price of children doubles (from CC to GO), Wchl becomes the highest attainable indifference curve, and it can be reached only with the combination of 4 children and 0.5 Gch goods. That is, in adjusting to the new price and income situation, the 10In this context, the "price" of children refers to "net" price; that is, the child's contribution to parental income less the costs which parents incur in bearing and rearing the child (with appropriate discounting of the child's future costs and contribu- tions . W' 3 wth‘ A 131.1---__-______- ————| I I I Goods 1 I Index : I I G :3 B I I I 1 I 1 I GchL —————— — ——————— I I '5 Bch‘" ________ ' I : I J I, l I D l I 1 C 0 1 2 3 4 5 6 7 8 9 10 Number of cthren Figure 6. Hypothetical budget lines showing price effects with welfare curves relative in favor of children. Budget lines: GA = Price per child is negative (i.e., net income/ wealth gain from an additional child) GB = Price per child is zero GC = Price per child is positive but "low" (i.e., net income/wealth loss from an additional child) GD = Price per child is positive and double that of GC 29 preferred number of children has been reduced by 33 percent (from 6 to 4) and the preferred amount of goods reduced from Gch to 0.5 G he ch’ or she is now at indifference level Wchl, down from Wch2. The analysis to this point has assumed that the prices of both goods and children are "positive." In Figure 6, however, line GA shows the price of children being negative; that is, one obtains more income with an additional child. As drawn in Figure 6, a person with budget line GC and Wch set of indifference curves will choose 6 children and GCh goods. He or she would not want a seventh child because the amount of goods given up to attain that child would cause a shift to a lower welfare curve. However, with the same initial income (as measured by 0G on the y-axis) but with a budget line of GA, one would want as many children as possible because each additional child increases income (thereby enabling one to obtain more goods) and also moves the person to successively higher indifference curves. If the net price of an additional child is negative, the shape of one's indifference curve is irrelevant in determining the number of children and amount of goods preferred, because regardless of the shape of the curves (i.e., relatively in favor of either goods or children), each additional child increases income and thereby the amount of goods which can be acquired, and the increase in both goods and children together provide the person with ever higher levels of satisfaction. There is little theoretical or empirical attention in the literature to the possibility or implications of the price of children being negative. This may be in part because it is difficult and costly to produce objective estimates of the costs of children; few 30 data are available, especially for low income countries. Mueller [1975] concluded that the economic costs of children to parents almost always exceed their economic contributions, even in peasant societies. Caldwell [1976a, 1976b] is highly critical of her data, her inferences and her conclusions, especially with regard to rural Africa. Based primarily on his and others' studies in West Africa, he concludes that the economic costs of children to parents are generally modest and are exceeded by the actual (or at least expected) economic returns to parents from children. Perhaps more significantly, the perceptions of parents are that the economic return to them from their children will be (or is at least expected to be) greater than their cost. Cleave [1974] summarized the results of about 50 farm surveys in Anglophone countries of sub-saharan Africa. Data on labor inputs of children were collected in only a few of the studies, and even then values were not imputed to the work of children. Nevertheless, in most studies in which data were obtained on the labor of children, the time spent by them in household production was not insignificant relative to that of adults. A survey in Gambia found that about 75 percent of the boys and half of the girls ages 11-16 worked on the farm more or less regularly [Cleave, 1975: 38]. A farm survey in Toro, Uganda, found that children ages lO-15 spent only slightly less time in agricultural work than their parents even though over the entire year it averaged just under 3 hours per day [C1eave, 1974: 48-50]. There are no data from this study which directly address the ' question. There is no doubt, however, that children in the study areas and in rural East Africa in general make substantial 31 contributions--at least in relative terms--to the income of their parents, not only during a parent's old age but also while still children. Boys frequently look after grazing livestock as early as age 5 or 6. Girls age 5 or 6 may already be helping their mothers carry water or firewood and may be assisting with other household chores, and at an even earlier age a girl may help to look after a younger sibling, thereby freeing older female members of the household or community for more productive work. Usually in traditional societies children are available for such tasks during the entire day except in periods given to traditional forms of education or formal socialization such as initiation rites or study of the Koran (in Moslem societies), and in some cases even these can be combined with productive activities as when children learn domestic skills or animal or crop husbandry while working together with mother or father or other adults. School attendance may, of course, severely curtail the time available for productive household activities.]] Some information on the price of children implied by the models in Figures 3 and 4 is given below. If one is located to the left of point e, it implies that one of the following three price conditions prevails: (l) The price of children is negative, at least until D number of surviving children are obtained. (2) The price of children is positive but the woman is located on an indifference nCa1dwe11 [1975: 22] describes the situation in West Africa as follows: "Children, in traditional farming areas, enter the labor force at 4-6 years of age. They are partly economic and partly social assets in that they perform such labor as looking after animals, carrying messages, transporting water and kindling, and, in the case of girls, acting as nursemaids for the toddlers, tasks which adults find inconvenient or beneath their dignity." 32 curve which intersects the budget line (and hence moving along the budget line by increasing the number of children and reducing other goods would approach tangency with a higher indifference curve.) (3) The price of children is positive but the woman is somewhere below the budget line in which case increasing the number of surviving children will move her to successfully higher indifference curves. Specifically, if a woman is located between point t2 and point e (Figures 3 or 4)--i.e., the range where S is increasing but is still below D, one of the above conditions applies and movement along S in the direction from t2 toward e would cause attainment of successively higher indifference curves. If a woman is located to the right of point e, it implies that for her the price of children must be positive, at least for all children above D number (or else children are not a normal good) and she is located on an indifference curve which intersects her budget line at a point with a combination of too many children and too few goods. Had she, for example, given birth to one less child, she would have acquired more goods and would thus be located at or closer to a tangency with her highest attainable indifference curve.]2 12Before giving birth to the additional child which caused the number of surviving children to exceed the desired number she would have preferred not to have that child, but she would presumably not now want one of her children to die even though that would enable her to acquire additional (non-child) goods. Presumably the grief which would accompany the death of a child would entail a net loss of welfare even when the satisfaction attained from the acquisition of other goods is accounted for. 33 If a woman is located at point e, it implies that she is also located on her budget line at a tangency with the highest attainable indifference curve. The following can be inferred from Figures 3 and 4 about tastes for children relative to goods: (1) Should the price of children increase (relative to other goods) with income unchanged, if the demand for children declines little or not at all, it implies a price-inelastic demand for children meaning that the person's indifference curves are relatively in favor of children vis-a-vis other goods. (2) Conversely, if a modest increase in the price of children with income unchanged is followed by a substantial reduction in demand for children, then children are price-elastic and the person's tastes are relatively in favor of goods. This assumes that tastes have also not changed during this period. (3) Similarly, should income increase substantially with constant relative prices, if accompanied by a sizable increase in demand for children it implies that demand for children is elastic with respect to income (and tastes are relatively in favor of children). (4) If a sizable increase in income is accompanied by little or no increase in the demand for children, then the income elasticity of demand for children is low (and tastes are relatively in favor of goods, assuming again unchaged tastes during this period. Two techniques will be used in Chapter VII to analyze the effects of demand variables on number of children born and number desired. The first will be multiple regression in which the indepen- dent variables will be proxies for income, relative prices and preferences. Additional variables will be included in some 34 formulations of the model. The second technique will be discriminant analysis to study the characteristics of both mothers and fathers who responded "no" and "yes" respectively to the question, "do you want more children?" Because this is a dichotomous (0,1) variable, it is not suited to multiple regression analysis. Multiple regression has been used in many analyses of rela- tionships between fertility and socioeconomic characteristics of populations. Some have used households or individual women as cases [e.g., Harmon, 1970; Schultz, 1972; Snyder, 1974]; others--probably most--have used cross-sectional data in which average values for states or other geographic/administrative divisions were used as the cases [e.g., Schultz, 1969; Hicks, 1974; Seiver, 1975]. These models have used various proxies for the demand variables (income, prices and tastes); some have also included one or more supply variables (to be discussed in the next section); and most have used a mortality measure, often assumed to be a proxy for the extent to which parents respond to mortality risks by adjusting upward their desired number of births. The various regression models to be tested with demand proxies as independent variables will be specified in Chapter VII. Determinants of Supply The supply of children surviving to a woman is equal to the number she gives birth to less those who subsequently die. The number of children a woman gives birth to during her lifetime can be viewed as the result of (1) her age at the time of her first birth, (2) her age at the time of her final birth, and (3) the average interval (or spacing) between births. These three parameters are determined 35 by a large number of both biological and behavioral characteristics of the woman, her spouse and her society. These include the age at which she first has coitus, its frequency thereafter, and the amount of time the woman subsequently abstains from coitus during fecund periods. The latter may be the result of disrupted marriages (widow- hood, divorce or separation--either voluntary or involuntary), post- birth abstinence from coitus, etc. Age at first (or at least frequent) coitus may depend on age at first marriage. The frequency of coitus may depend in part on whether the woman is married to a monogamous or polygynous husband; the duration of infecundity (following childbirth) depends in part on the length of time the woman breastfeeds which in turn depends, among other things, on cultural practices, individual preferences of the woman, and whether or not the child survives at least to weaning. As an example of the way a cultural practice may affect fertility, women in many African societies traditionally abstained from sexual intercourse for up to two years or more after the birth of a child. One effect is to prolong the average interval between births, which in turn causes fertility to be lower than it would otherwise be. This postpartum or lactation taboo can be considered a cultural practice which also functions as a fertility determinant via its effect on the average interval between a woman's births. A set of social, cultural, and economic changes (which may be difficult to specify individually) may alter the adherence to this particular cultural practice. Some proportion of the childbearing women might discontinue the practice altogether or might follow it for shorter Periods of time than in the past as the consequence of changes in 36 social, economic, or educational circumstances. Even if these women continue to breastfeed while resuming coitus, there will be some chance of becoming pregnant. By changing the average birth interval, this would affect the level of fertility for both individual women and the society as a whole. It would presumably have no independent effect on the desired total number of surviving children.13 Changes in the average length of lactation would probably have a similar effect, provided that women resume coitus either prior to or soon after weaning. Breastfeeding itself prolongs the time interval between childbirth and the resumption of ovulation, and hence prolonged breastfeeding increases the average interval between the birth of one child and the next pregnancy, even if women engage in coitus regularly before weaning the child. Studies have shown that this ovulation-inhibiting effect exists for breastfeeding which lasts up to 1 1/2 to 2 years. However, the variation in resumption of ovulation is considerable [van Ginneken, 1974]. A study in Taiwan showed that "the relationships of lactation and amenorrhea ceased to exist when length of lactation exceeded 21 months and amenorrhea remained at a level of 12 to 13 months" [van Ginneken, 1974: 203]. This means that among this group of Taiwanese women, of those who breastfed for 21 months, ovulation resumed on the average at 12-13 months although some women resumed ovulation at 4 or 5 months after giving birth and some not until 21 months even though all breastfed for 21 months. In addition, full breastfeeding delays ovulation — 13Almost all women in the sample indicated that the post- partum sexual taboo is breaking down. Morgan [1975: 206] reports a similar finding in Nigeria. 37 more than partial breastfeeding (i.e., supplementing breast-milk with other foods in the baby's diet). Hence, socioeconomic and cultural changes affect the number of children born primarily by altering the average interval or spacing between births and by changing the average age of women at first coitus (and hence average age at the birth of the first child). Changes in the mortality rates of infants and unweaned children would also be expected to affect fertility by altering the average interval between births although there is clearly a causal relationship between a woman's number of deceased children and her number born going in both directions. On the one hand, the number of children a woman gives birth to partly determines the number who die because the more children that a woman bears, the greater the risk that one (or more) will die. Expressed another way, one would expect that due to simple mortality probabilities alone the average number of children deceased would be a direct function of the average number born. However, it is also hypothesized here that the number of deceased children a woman has is also a major determinant of the number born due to the effects of the mortality of infants and young children on the mother's fecundity and average birth intervals. Most child mortality occurs within the early months of life. Cessation of lactation due to a child's death would result in earlier resumption of ovulation. Since post-birth abstinence from coitus is traditionally usually conditional upon and corresponds to the period during which the woman breastfeeds, death of a child who had not yet been weaned would presumably result in earlier resumption of coitus by the woman 38 than would have been the case had the child survived (at least in traditional African societies). This combination of earlier resumption of both ovulation and coitus consequent upon the death of an unweaned child (usually accounting for the preponderance of childhood deaths) can be expected to result in a smaller average interval between the births of children where the first of a pair dies as compared to the average interval between births of children of whom none die.14 Figure 7 presents a model illustrating hypothetical relation- ships between (a) basic social, cultural, economic, health, and educational characteristics (and changes therein) of women and societies and (b) supply and demand determinants of the numbers of children born and surviving. The model in Figure 7 is matched with the transitional framework illustrating hypothesized changes in supply and demand over time in Figure 3. The upper portion of Figure 7 illustrates the role of selected basic supply determinants; the lower portion illustrates the role of basic demand determinants. The four transitional stages (from Figure 3) are identified along the bottom of Figure 7. Throughout Stages 1 through 3, both fertility and the actual number of surviving children are determined by supply and not demand factors, although it is assumed that initially demand exceeds 14For a summary of some studies which generally support the above hypotheses, especially for "pOpulations at the lowest level of development," see Preston, 1975. 39 .mcw—uwv Fazucm>m ecu umcagu auw__usum ou mommx:_— one can ~u_Lw< pass; :_ cowuvmcagu u_;nasmoEmu new u_Eo:ouwo.uOm uunvmwguoa»: mo —ouoz .u weaned A ue*h av nv N» we _ . . . : Hu? mooE o>3maom 1 550: .3 o man—mm 1mg eoesnom 1.5m .«o A - noeoofi - 11:8320 was? s. m Aoefi “26 on moofioaum n35: 1.5m mo Hope—:2 :ohflaflo wcawfifio mugBHHwQ Ugamwn— muqmcafimvwn o§o>< Wcfimfim $2.3? OHM gonev .8 wedsfiwom own»? game 5 no 538 mo :ofinamooom Law .8 _ H.352 h _ amend: . spoaoom Ho: " _ wcmvoowh ted 5503 sfiflfiom . _ 0535 .8 Rheum? concise one .8 83 313308 u . 2MB mo .m xo wane» 1mflovo§o _ . owmuo><1 We. .H. mom dad dacoafiozum vino p03 ongowwwwonmlwi mpfimcH and 5.53: 52289 no 2328 .wfinE "widowed—I mo 5.3m s. .93» o so.“ .3888 _ ed owe- mo oocoeMomH- Beam £355.90 _. _ _ “ fin... 3...... eye“... 3 H _ on one- enna ea em? a a _ _ $30.13 . _ mo monogocH1 —1 IIIqu wecdcfishopmo Mqhgm c." 1.: 40 the actual number of surviving children while later (commencing in either Stage 2 or 3) the actual number of surviving children exceeds demand.15 Figure 7 is intended to depict the basic societal and indi- vidual characteristics as continually changing throughout Stages 2 through 4 of the transition. They continue to impact, in presumably changing ways, on both supply and demand determinants throughout all the transition stages. From the standpoint of supply relationships, it is much less complicated to analyze determinants of number of children born (fertility) than the number surviving. First, a potentially larger number of variables affect mortality (and hence, survival) than affect fertility. Second, before one can survive one must first be born, but not vice versa. Hence, for each woman the potential number of surviving children is limited to the number she gives birth to. It is less than this number by the proportion of those who subse- quently die, so theoretically, number surviving should be the combined product of the determinants of fertility together with the determinants of mortality or the mortality rate (or conversely, survival rate). In Chapter V, multiple regression analysis will be used to test hypotheses about relationships between supply variables and 15As noted earlier, demand is assumed to be expressed only in terms of the total number of surviving children and not in the desired intervals between children, although of course differences in average birth intervals--whatever their causes--will affect a woman's numbers of children born and surviving at any given point in time. 41 number of children born. Models for determinants of the survival rate among a woman's children will be tested in Chapter VI. Other Formulations to be Tested In Chapter VI separate analyses will be carried out on two of the independent variables in the basic supply mode1--age at marriage and reported duration of breastfeeding. Multiple regression will again be used for the analysis. ' As outlined earlier, supply of and demand for children will be analyzed separately (in Chapters V and VII respectively) using multiple regression analysis. It is hypothesized that in Stages 1 through 3, supply variables primarily determine the numbers of children born and surviving and demand variables become important determinants of fertility only in Stage 4. However, some income, price or preference characteristics of the woman and her society which directly detennine demand may indirectly determine supply by altering the supply variables. For example, education may affect both demand and supply in several ways. Among possible effects on demand, one might hypothesize that women who have completed primary school have developed tastes which are relatively more in favor of goods vis-a-vis children than are the tastes of women with no education. Education may affect the price of children by increasing a woman's opportunity cost of childbearing and childrearing; educated women may hold higher "quality" standards for their children (implying a higher price) than women without education. Or. the income of husband or wife may be higher due to more education with a corresponding positive income effect on demand for children. 42 However, a woman's educational level might also indirectly affect supply. School attendance could cause later age at marriage. Educated women may be more likely to purchase breast—milk substitutes and wean earlier thereby causing an earlier average resumption of ovulation. Both of these changes would affect number of children born. It can also be hypothesized that the survival rate among children of educated women will be higher than among children of women without schooling. A rigorous analysis would require that supply and demand functions be estimated independently--preferably simultaneously-~and that the independent variables be mutually exclusive in order to avoid the so-called identification problem. That is, none of the variables used in the supply models should be used in any form in the demand models and vice versa. In this case, education is used (in Chapter VII) as a proxy for tastes and relative prices in the demand model. Although it is not used directly in the basic supply model of number of children born (Chapter V), it is used as one of the indepen- dent variables in each of the models tested in Chapter VI--percent of children surviving, age at marriage, and duration of breastfeeding. The latter two are used as independent variables in the basic supply model of Chapter V, and hence education indirectly enters as a deter- minant of supply, so that requirements for avoiding the identification problem are not rigorously met. However, this violation seems warranted by the value in identifying the potentially significant effects of education as an independent variable in the various models. Finally, it is hypothesized that ethnic and geographical differences may significantly affect the relationships between 43 independent and dependent variables. The differences between the Moshi and Lushoto areas discussed in Chapters III and IV may be inadequately captured by the independent variables in the models. In Chapter V dummy variables will be incorporated into the basic supply model to test for significant differences in the independent variables due to geographic/ethnic differences. In addition, for nearly all phases of the analysis, results will be produced and analyzed for the district and area levels as well as for the aggregate level. CHAPTER III DESCRIPTION OF THE STUDY AREAS IN NORTHEASTERN TANZANIA Location, Area, and Population Size This is a study of four rural areas located in two districts in Northeastern Tanzania. Two of the four study areas are in Moshi district in Kilimanjaro region; the other two are located in Lushoto district in Tanga region (see Maps 1 and 2). Throughout this paper the study areas will be identified as M1, M2, L1, and L2 respectively. Both Moshi and Lushoto districts are located in mountainous areas of Northeastern Tanzania. The two are among the most densely populated of Tanzania's approximately 65 rural districts [Moore, 1973: 255]. At the time of the 1967 census, the rural population of Loshoto district totaled 210,484; that of Moshi district was 361,912 [Tanzania, Government of, 1969: 289, 290, 332]. In mid-1973 the rural populations are estimated to have been about 245,000 in Lushoto district and 425,000 in Moshi district. The average rural density for Lushoto district in mid-1973 would then have been about 70 people per square kilometer while that of Moshi district would have been about 98 per square kilometer. This compares to an estimated density of 16 per square kilometer for Tanzania as a whole. The eastern border of Lushoto district is about 60 kilometers inland from and parallel to the northeastern coast of Tanzania. The 44 45 ' I KENYA "\~. , gr\‘\\. .433 _ “em-'0' ‘HI 8710 ‘ l-I‘R‘s'i "S Mei-"1'1 Nosoi \. 1... miss: Nondem 'Pon.f.hl.fF£-;=E‘ 5"‘71-3- ondoo h- — 1:51."; .‘ '1. .. ’0, E? 11.33 41.30 r 0., ._-_1. -.- -~-.-e-. 0’ ' ~ '. .- o.-—-—..—. '— _ . -— o Noregotop/ “1.1.234" 3.- moron . _..... Intern. Mundane: I" District Boundaries C 0 Kilometres 240 :- L_ I g in it _- --’ze 03% Map 1. Location of Lushoto and Moshi Districts 46 \ lo '\, Nairobi Arusho {/24 District v." Some A4 a s a i s t e )3 pi e INSERT Dodomo O \TANZANIA Dar es Salaam. KENYA \ .:.: ‘t‘o 6::0::. . l o 0.0 O 0 e o 0.. .e.00.0 .000... to. o o °~ 0.9 9.0.0.303 ' . . C ’ ‘ O 0 o o v 0 o 4 \ o I o 1 o o e o 0 . \‘o '0'... a... o ’8. 0.9.0.0‘ '0. ' ‘ oo'ooooo‘ov’ee -‘ 0.0...o.o.o.o.o.o.ol e - 0.0.0 on o, /\ w More densly po- pulated areas Tonga 3 Korogwe 11L 211 411 511 911 120 4' K 1 L o M E 1 R E s 104., Dar es Sold-m 1» G) Clusters I to 4 —.—.-+— International boundary @Populated .area .. ....... District boundary 1n each District N Main roads 0 lawns and Urban centres Map 2. Study areas within Moshi and Lushoto Districts Lushoto ifl District 2 I-tfi -CY 47 southern border of Lushoto district is located in the Pangani Valley which runs southeast from Mt. Kilimanjaro to the coast below Tanga. The total area of Lushoto district is 3,497 square kilometers [Moore, 1973: 255], but the northern third of the district is part of the Mkomazi game reserve and is not inhabited by pe0ple. Of the southern two-thirds of the district, about half is sparsely populated lowlands and plains. The extreme southeastern one-third of the district is highlands with elevations up to about 2,300 meters (7,550 feet), but with the inhabited areas primarily within the range of 1,050 to 1,700 meters (3,450 to 5,575 feet) (see Map 2). This is the western range of the Usambara mountains. The average density for this most densely populated portion of the district was about 250 to 300 people per square kilometer in mid-1973. Moshi district is located on the southern and western slopes of Mt. Kilimanjaro which, with a peak of 5,895 meters (19,400 feet), is Africa's highest mountain and one of Africa's two permanently snow-capped mountains. Lushoto and Moshi districts are separated by Pare district. The eastern border of Moshi district is about 100 kilometers northwest of the western border of Lushoto district. Moshi district was created in 1970 by dividing the former Kilimanjaro district into two new districts--Moshi and Rambo.1 At the time of the 1967 census, the area that became Moshi district had 76 percent of the 1Moshi district was itself subsequently divided (in 1975) into two new districts. The eastern part is now known as Vunjo district while the western part retains the name Moshi district. Throughout this thesis, however, the name "Moshi district" will refer to Vunjo district and the present Moshi district combined. 48 rural population of what was then Kilimanjaro district. The total area of Moshi district is 4,336 square kilometers, but as with Lushoto district, much of it is uninhabited or only sparsely inhabited by people. The densely settled portion of Moshi district is located on the southern slope of the mountain above Moshi town at elevations ranging from about 1,060 to 1,850 meters (3,480 to 6,070 feet). This accounts for less than one-fifth of the entire area of Moshi district. Elevations above about 1,850 meters are part of Kilimanjaro National Park and are not inhabited by people. Both the western slopes and the surrounding southern and western plains suffer from considerably lower rainfall and are relatively sparsely populated. The average density for the more densely settled portion of Moshi district was about 400 persons per square kilometer in mid-1973. Each of the four study areas had a total population of about 5,000 at the time of the survey. The study areas themselves are geographically very small; all four are between about seven and ten square kilometers in size implying average densities on the order of 500 to 750 per square kilometer. M1 is located around an area known as Kibosho (or Kiwoso) about 15 kilometers north and slightly west of Moshi town. M2 is also called Vunjo (it is in the division of East Vunjo in the new Vunjo district) and is located about 50 kilometers east-northeast of Moshi town and about eight kilometers northwest of the village of Marangu. L1 is also known as Bumbuli cluster; the small settlement of Bumbuli is located on the northeastern edge of the study area. L2 is known as Soni area; the settlement of Soni is located in about the middle of the study are. (See Maps 3-6.) 49 m 0 kilometers I appronmate Cluster boundary_.’_-.__._._ I I I {jekaug'731 Map 3. Sketch of M1 (Kibosho study area) and Vicinity 50 A Na kilometers --3 3 8 approximate cluster boundary,-..-. u.- (jek. aug '73) Map 4. Sketch of M2 (Vunjo study area) and Vicinity 51 “‘EIEMUI Fifes' CSETVC N | 11mm ‘1 \ 0 kilometers HLDL 1:111111 ’ HLDLA W fiver P00 5 KIUmUmega River Map 5. Sketch of L1 (Bumbuli study area) and Vicinity 52 kilometers I IIEK' 5991 '73) Map 6. Sketch of L2 (Soni study area) and Vicinity 53 Sample Selection These four study areas were among the 72 rural areas included in the sample for the 1973 National Demographic Survey (NOS) of Tanzania. Random sampling was used in the N05 to select four rural areas in each of Tanzania's (then) 18 rural regions. Two of the four rural sample areas in Kilimanjaro region and two in Tanga region were then intentionally chosen because of some of their known agricultural, economic, demographic, and historical characteristics to be included in an expanded study which provided the data for this thesis. Sampling of households within each of these four areas was random. Every second household in each of the two Moshi areas and every third household in each of the two Lushoto areas were included in the sample; these households accounted for about one percent of the estimated total number of households in each district at the time of the survey in late 1973. (Further information on methodology is given in Appendix 1.) Some Agricultural and Settlement Characteristics Residents in both districts are predominantly small-holder farmers growing staple foods; most households keep livestock and most produce some crops for market. The most important food crops in Moshi district are bananas, maize, and beans. In the Usambaras they are bananas, maize, cassava, beans, and sweet potatoes. A variety of other vegetables are also raised and consumed in both areas . 2 2In Chapter IV, Tables 5, 6, 7, and 10 present detailed data on some agricultural characteristics of the study areas. 54 Settlement patterns differ somewhat between the two districts. Most of the population in Moshi district reside in the highlands (elevations above 1,000 meters) but people generally cultivate land both in the area immediately surrouding their homes as well as farther down the mountain toward the plains. The highlands and the plains are two different ecological zones based on soils, rainfall, temperature, and altitude differences. Maize is grown at the lower elevations and grass is cut and carried up the mountain to be fed to the cattle. For the most part, people do not live in villages but. reside in separate homesteads located in the midst of or immediately adjacent to their small coffee and banana farms. Beans are grown interSpersed among the coffee and bananas. Even before European contact, terracing and irrigation were practiced, animal manure was applied to the land, and shade trees were maintained for fuel and building materials. Historically, livestock were grazed in both lowlands and highlands, but in this century most of the former grazing lands in the highlands have been converted to permanent cultivation due to increasing land scarcity. Many people in Lushoto district also cultivate land both near their homes on the mountains as well as down on the plains. However, in Lushoto district most people live in small villages of about 10 to 50 households rather than in dispersed homesteads as in Moshi district. A typical village is situated along the crest of a ridge with cultivated areas running down the hill or located in nearby valleys, sometimes at a walking distance of more than an hour. A large proportion of the land cultivated in the highlands is located some distance from the village in contrast to Moshi district where 55 most of the cultivation in the highlands is located immediately adjacent to the individual homes. Tribe, Religion, and Literacy in the Study Areas The slopes of Mt. Kilimanjaro are the homeland of the Chagga tribe, and the Western Usambara mountains are the homeland of the Sambaa (or Shambaa or Shambala) tribe, although substantial numbers of the Pare tribe--whose homeland is the Pare mountains located betWeen Mt. Kilimanjaro and the Western Usambaras--also now live in the Western Usambaras. The upper part of Table 1 gives the tribal composition of the four study areas. In both M1 and M2, over 99 percent of both women and men are Chagga. In L1, 93 percent of the women and 91 percent of the men are Sambaa; 4 percent of the women and 6 percent of the men are Pare. Three percent of both women and men are of other tribes. In L2, 80 percent of the women and 79 per- cent of the men are Sambaa; 16 percent of both women and men are Pare. Five percent of both women and men in L2 are of other tribes. The lower half of Table 1 shows the religious composition of the study areas. In M1, 87 percent of the women and 77 percent of the men are Christians (virtually all Catholics), 2 percent are Moslems and the rest consider themselves followers of traditional beliefs. In M2, 99 percent are Christians (about 7 percent of whom are Lutherans) and about 1 percent are Moslems. In the two Moshi areas Protestants account for a smaller proportion and Catholics account for a larger proportion of the total than is the case for Moshi district as a whole. In L1, 82 percent are Moslems; the rest are Christians (almost all Lutherans). About 75 percent in L2 are 56 .moosa Assam one» as» ea zoom no ouogo>u uouzm.o3:= .m .mumL~ xenon acu>o.og oz» oz» :0 sumo Loo momuucwueod m>.uuonmag you we wooeu>u uuuzm.ox:= .N £93: 32:53:33”. we 3.53:; 3 239. .133... .3 was .33 :26: :u 3.3.2: :55? .. $30: $3. .5: .88 :3. .83 .38 .33 .3: .5 o.oo. o.oo. o.oo. o.oo. o.oo. o.oo. o oo. o.oo. o.oo. o.oo. o.oo. o.oo. o.oo. o.oo. .auoh ..m ..o o.o n.o ..o. m.m ~.. m.o o.o ~.o o.o o.o e..~ o... .a:o.u.v-L» o.oe o.on o.o. o..~ N.. e.. o.o. o.o. ~.~o ~.~o o.o o.. o.~ 5.. E~.mo: o.o. o... o.o. o.o. o.o m.n o... o.oN o.o. o.o. m.. o.o o.o o.o acoumuaogn o.ne n.oe ..m o.o ~.qo o.oo o.m o.o ... ..o o.~o ¢.~o o.o. nnoo o..o;uuu 8.2.2. .38 8e: .83 .3: .33 .88 .32 :m: .5 odo. odo. odo. odo. odo. odo. odo. odo. odo. odo. odo. odo. Qoo. Qoo. 3.8 ..N m. md md o.o o.o o.o mé od ed md o.o o.o o.o mcwfio m.~e adv odd mom o.o o.o To. mdn o.om Ndo o.o o.o o.o o.o $9.3m 9m o.o o... o.o Nd o.o dd. m.m. .d N... md o.o o.o o.o to... o.ov o.oe ~.o o.o o.oo o.oo o.o o.o o.o ~.o o.oo o.oo. o.oo ..oo doomed eat. 3.: 3: 3: .N: :: 8: .3 .8 E .3 .3 E .2 E .5 .95: :25: .33: :25: .33: :98: .93: :25: AS: :25: .33: :25: .53: .cqu: neonate? No.22: ~23: S 3 N: .: E838 Lo» moo.u=n.sum.o omoucuULuo "muab< xvaum ogu on copu.moasoo m:o.m..o¢ on. .oa.s» .. o.o.» n3. €26: 3..—.3: 323.59 we 3:33: .8“. EB :95: :u 57 Moslem. Almost all the rest are Christians (about 5 percent being Catholic).3 Until the l9505, and particularly the l9605, most formal schooling was provided by Christian missions. Due to the relatively early missionary work in both Moshi and Lushoto districts, a few mission schools were established in both districts around the turn of this century. Table 2 shows the percentages of both sexes and of various age groups which have received some formal education in each of the study areas. Figures under the heading "Children" are for surviving children for all ever-married women included in the study (regardless of the child's current residence). Data in the lower panel are for all women (regardless of marital status) and for husbands of currently married women. Nearly all males ages l5-l9 have had some schooling in Ml, M2, and L2 and about three-fourths have in Ll. Females ages l5-l9 fared somewhat less well although 90 percent in M2 have received some schooling; only 58 percent in Ll have. About three-quarters in both Ml and L2 have received some schooling. Roughly comparable differ- ences among the areas and by sex exist for children ages lO-l4. Ages 7-9 are too young to draw inferences about differences although 3According to the 1967 census, of those heads of households (males) in all of Kilimanjaro district who gave their religion (97 percent of the total), 71 percent were Christians, 8 percent Moslems, 20 percent followed traditional beliefs, and 1 percent were of some religion other than these three. In Lushoto district, 98 percent of the household heads stated their religion and of these, 13 percent were Christians, 8l percent Moslems, and 5 percent followed traditional beliefs [Tanzania,Government of, 197l: 249-50]. 58 mUPMEOm H m mm.mz u z e N. . m. e mN e N. . m . .N are.5 +eo N mm e. on o. om e on N. .m m NN mm-o.o. omueo N. No .m .o N. on N do on eN NN Nm mo-omo. omuem eN oN Nm em mm No oN oN mm me me NN mm-ooo. oN-eN mundane: new :msoz no em mm mm oN do do mN om No oN om mm-omo. o.-m. do me NN do No NN mm mm oN No do we mo-ome. o.-o. m. N. NN oN NN e. N N NN «N .N mN eouooo. o -N edge..;o 3: .m: .N: .E .o: .3 .3 E .8 .3 .3 .2 .N. .5 u z N z N z m z N z N 2 seem meow» noose mo< ouocmod .gmoz .4 N: .2 .xo.eo< uemggoo NNo. :. .xmm new mo< an .eo.umuoem .msgom meow :p.: mmomucmugme .N m.amN 59 the exceptionally low percentages in Ll imply a later average age at first enrollment which is consistent with what is to be expected in areas with an acute shortage of classroom places. For men there are only relatively small differences among the areas, even among older men. For example, among those born between 19l4 and l933 (ages 40-59), the percentages with some educa- tion are 28, 3l, 30, and 36 in Ml, M2, Ll, and L2 respectively. In fact, a slightly higher proportion of men age 40 or older in the Lushoto areas than in the Moshi areas received some formal education (see columns ll and l3). Probably all (or nearly all) of these attended mission schools. A higher proportion of women of all ages in the Moshi areas received schooling in the Lushoto areas, although proportions are low in both districts for women age 40 or older. The difference is particularly marked for women ages 20-29. Considerably more detailed information will be given in Chapter IV on important socioeconomic characteristics of the study areas in l973 together with a description of important aspects of development in these areas during this century. CHAPTER IV RURAL DEVELOPMENT IN THE STUDY AREAS The purpose of this chapter is to document the socioeconomic environment in these areas in the early l900$ and also in l973 and to trace the development of these areas during this time period. Around the early 19005 these were still traditional rural societies. Almost all residents still adhered to traditional religions. Very little commercialization of agriculture had taken place. Birth rates were high and death rates were probably not yet much below their historical average levels. Crude birth rates were probably close to 50 per thousand and it is unlikely that crude death rates were below about 40 per thousand (with average expectation of life at birth of about 25 years) giving a natural rate of increase of about 1 percent per .year.‘ Almost no one had received any formal education. Non- ‘traditional forms of housing had only recently been introduced by European settlers but had not yet been adopted by Africans. By l973, crude death rates had declined to about 15 per thousand (or less) in the study areas, average expectation of life at birth was 50-55 years, relatively few residents still followed lCaldwell [1975: 4] estimates that in west Africa prior to this century, crude death rates were close to 50 per thousand with average expectation of life (at birth) about 20 years or less. 60 6l traditional religions, most children were receiving at least primary education, substantial numbers of households (in some areas) resided in non-traditional houses, and almost all households participated in the cash economy. While these were clearly not yet modern rural societies, they were also no longer traditional. They had experi- enced very considerable social, economic, and cultural transition although there were large differentials both within and among the four areas in the extent to which households had participated in and benefitted from these changes. Some of the more important of these changes will be described in detail in this chapter. The Meaning of Rural DevelOpment The l970s have witnessed a searching critical reappraisal of the meaning of development and how best to achieve it. Gunnar Myrdal [l968, l970] and Dudley Seers [l970] were among the first to ask penetrating questions and to argue for development strategies which directly attempt to alleviate poverty and deprevation rather than to maximize aggregate economic growth in the h0pe that the fruits of growth will eventually trickle down to ordinary pe0ple. Owens and Shaw [l972] emphasized the importance of direct participation in the development process by those intended to be the beneficiaries. Kocher [l973] identified the complementary relationship between development patterns in which participation in and benefits from development are widely shared (rather than concentrated in the hands of a privileged class) and the widespread desire among families for fewer births. Recently, Hunter and Nelson [l976] have analyzed the conflicts between the objectives of broad political participation, rapid 62 economic growth and socioeconomic equity. Their analysis indicates that in the early stages of modernization and develOpment the main conflict is between equity and political participation while in later stages the main conflict is between political participation and economic growth. Although political participation is certainly one important aspect of development, this chapter will focus only on growth and distribution in the study areas. For the purposes of this study, "rural development" will mean, (l) general improvements in living conditions and material well-being of rural peOple, (2) the capability for sustaining continuous improvements over time, and (3) absence of gross inequalities in the distribution of the gains during this process [Kocher, l973: 4]. All three of these development criteria are important and one would expect to find various mixes of the three among different development experiences. The second of the three criteria is generally more difficult to measure than the other two. Information on some of the important components of development in the study areas will be given in this chapter, but descriptions will be limited to the first and third criteria above. Some Comments on the Measurement of Development "Development" implies a process; it also implies changes for the better taking place over time. This in turn requires something akin to time series data for proper analysis of the development pro- cess and its many facets and effects. Time series data REE §g_were not generated by this study. In fact, the nature of the relationships being analyzed would require time series data which extend over a 63 period of at least two to four decades. Such opportunities just do not yet exist in rural Africa. However, the survey data do provide two different types of information about changes over time. One is the comparison of condi- tions which prevailed in each area at the time of the survey with conditions presumed to have prevailed at some time in the past. The most satisfactory reference period is about the turn of the century, which is reckoned to be--for all practical purposes--the approximate beginning of socioeconomic transformation in each of the four study areas. Since all four areas were initially exposed to non-traditional outside influences (e.g., missionaries, European settlers, schools, and new crops including coffee) at about the same time, comparisons of the four areas provide some insights about the rates at which various changes have taken place in each of the four areas and the extent to which various changes had permeated the areas by l973. The second comparison is in the characteristics of the different age groups currently residing in the areas at the time of the survey as indicators of the timing and rates of changes. These comparisons are particularly useful for education, health care, POTyQyny, and to some extent incomes. Age groups can also be ComDared among the four areas. Indicators of development which will be described in some deta‘i l in this chapter include, (l) quality of housing (used in the later- analyses as a proxy for wealth or permanent income), (2) types 0f CV‘Ops produced and livestock owned, (3) estimated values 0f CY‘OPS ham’es‘dmd and sold per household (value of crops harvested is used as a Proxy for current income), (4) wage employment, (5) OWNEY‘ShlP by 64 households of selected consumption goods, (6) ownership of agricul- tural equipment, (7) levels of formal education, and (8) prenatal medical care and delivery of babies in hospitals. Data will be presented in Chapter V on average numbers of children born, age at marriage, duration of breastfeeding, and incidence of polygyny. Data will be presented in Chapter VI on average numbers of children surviving and implied survival rates. Descriptions of Agricultural Changes In the late nineteenth and early twentieth centuries the Western Usambaras (including Lushoto district) were described by European observers as being a densely populated but prosperous area with healthy inhabitants and a flourishing agriculture. The Sambaa were known by the German administrators as progressive farmers. The possessed "rich banana cr0ps" which . . . took in "half of all the fields. " They practised furrow irrigation and . . had "a significant wealth of livestock" . . . of "superior quality" [Attems, l968: 140, quoting German travelers writing in the l880$ and l8905]. 0n the slopes one sees thick clumps of banana palms and between these, fields of sugar cane, maize, beans, pumpkins, and tomatoes. All the fields are extremely well tended. In the pastures numerous well-fed cattle graze. . . . All the settlements and inhabitants of this delightful area display a certain prosperity [Bauman, l890, as quoted in Iliffe. l97l: 33]. However, at least according to some observers, during the first half of the twentieth century the general prosperity declined and the well-being of the Sambaa deteriorated. Iliffe [l97l: 36, 4l] notes a tendency towards agricultural involution (i.e., falling labor and/or land productivity) in the Western Usambaras, and Attems [l968: l39-40] describes a process of involution and pauperization 65 which has occurred despite the development efforts of colonial administration. There is hardly any other district in East Africa where development efforts began as early as in the Usambara Moun- tains, were so manifold and so often repeated. . . . One scheme after another failed and proved itself to be a fruitless investment of capital and personnel. . . . The absence of agricultural development in a region which had had a sharp increase in population for decades has resulted in the classical process of involution. Shifting cultivation and semi-pennanent farming have been widely replaced by permanent cropping with one or two annual har- _vests. The concomitant circumstances are erosion and a great number of unusually small holdings. The originally high-quality nutrition with maize and beans has largely been replaced by cassava, which is poor in protein and minerals. Whereas well-nourished herds of cattle were once present, we now find numerous emaciated animals and over grazed areas. It is to be assumed that the economic situation of the population is worse than it was seventy years ago, that is, before the development efforts were begun [Attems, l968: l39, l4l]. Other reasons often cited as contributing to the problem are the increase in population density, the lack of a suitable cash cr0p from which to bring income into the area and serve as a catalyst for increasing agricultural productivity, failure to maintain and develOp irrigation networks, and poor soil conservation practices. The Western Usambaras are sometimes contrasted with Kiliman- jaro which is commonly viewed as an area that has experienced con- siderable prosperity and economic improvement during the twentieth century. Substantial economic growth occurred in Kilimanjaro agricultural during the first half of the twentieth century, although these gains were apparently unevenly shared. In the former Kiliman- jaro district (now Moshi, Rombo, and Vunjo districts), there were 3,300 African coffee growers in l923/24, l6,800 by l933/34, 36,900 in 1953/54, and 87,000 in the late l960$ [Maro, l974: 237]. This 66 represented about 90 percent of the rural households in the late 19605. Average coffee income per grower increased from Shs 70 in 1933/34 to a range of about Shs 800 to 1000 during the 19505 and 19605. In the three years of 1953/54, 1955/56, and 1965/66, average coffee income per grower was in the range of Shs 1600 to 2000 [Maro, 1974: 237]. Land pressures have increased very substantially during the past century in both the Kilimanjaro and Usambara areas. The rural population of Moshi district is estimated to have increased between 400 and 500 percent between 1900 to 1967. Farm size in the highlands fell from an estimated average of about 2.2 hectares per household in 1921 to 0.5 hectares per household in 1967 [Maro, 1974: 85]. Similar changes have taken place in the Western Usambaras. One response to the rising population density in the Kiliman- jaro area has been more diversification and intensification of agriculture in both highlands and lowlands. Coffee was planted in the interstitial areas among the banana groves, and stall-feeding of cattle on banana leaves and stems and on grass brought up from the lowlands was introduced. More maize and beans were grown in the lowlands, former fallow and grazing areas in the highlands came under permanent cultivation, steep riverbanks were planted with coffee and bananas, and more marginal areas in both highlands and lowlands were cultivated [Maro, 1974: 86-88]. Accompanying the increasing land scarcity were fragmentation of land holdings to enable most sons to inherit land, increasing land litigation among kinsmen, and an estimated 700 percent increase in the value of land in the highlands [Maro, 1974: 88]. 67 One indicator of increasing pressure on the land is changes in average size of land holdings and average numbers of cattle owned per household from one generation to the next. Each adult male respondent in the study areas was asked how the size of his land holdings compared to those of his father when the respondent was a boy. He was also asked how the number of cattle he owned compared to the number owned by his father. In M1 and M2, 73 and 80 percent of the respondents said their land holdings were smaller than their father's had been; the percentages in L1 and L2 were 81 and 85. Approximately similar pr0portions of respondents said they owned fewer cattle than their fathers had owned; the percentages were 82 and 85 in M1 and M2, and 76 and 81 in L1 and L2. In L1, L2, and M1, nearly half of all respondents reported that they now owned no cattle while only 11 to 18 percent reported that their fathers had owned no cattle. This suggests that the proportions of residents in these three areas who do not own cattle have increased by 2 1/2 to 4 1/2 times during the past generation or two (of course, as a result of population growth, the absolute number who have no cattle has increased much more). Although there may have been an equivalent proportional increase in the numbers not owning cattle in M2, only 9 percent of the respondents and 4 percent of the respondents' fathers were reported as owning no cattle. Wealth, Incomes, Consumption, and Investment The following characteristics of households and individuals in the four areas in 1973 will be described in this section: types of housing including an index of building quality as an indicator 68 of household wealth; reported livestock holdings and estimated values of crops produced per household including indicators of the distribu- tion of crop values among households within each area; possession of selected consumer goods; and possession of agricultural equipment. Data will also be presented on occupations of both men and women. Housing One of the best indicators of economic improvements and social and cultural changes in these rural areas is the prevalence of modern and relatively expensive types of housing. Traditional houses in Lushoto district are either round or rectangular with grass or banana- leaf roofs and pole and mud ("mud and wattle") walls. Traditional Chagga (Moshi district) houses are conical with banana-leafed roofs which sweep down to the ground from about 12-foot peaks, although Moshi district houses may also be rectangular with dirt floors, mud and wattle walls and grass or banana-leaf roofs. Table 3 summarizes information on the extent to which modern types of houses existed in the study areas in late 1973. Typically a household's first major expenditure on improved housing is the installation of a metal (tin) roof on a rectangular house which has a dirt floor and mud and wattle walls. Line 3 shows the proportions of households which have at least one building with a good roof, regardless of the other characteristics of that particular building. Percentages are 65 in M1, 92 in M2, and 34 each in L1 and L2. Line 5 shows that 22 percent of the households in M1, 29 percent in M2, 2 percent in L1, and 4 percent in L2 have at least one building 69 .mm=.m> amen m>.»umemmg we» we mmomgw>m emugo.mzeo mew ... ecu o. mee:.ouv ouoemod ecu .emoz so» mm=.m> oer- .ee. oc...m=em .Nz Lowv m cso.ou e. mm=.o> xmoc. ..o co comma mew o ecu .N .m mes:.ou :. mm=.u> xmee.-- .ume emu; .euuozu .mmogo ”moo; Loom “Housmu "gonad. .meoum .xuwcn omcwo .xuo.p mumcueou um.—m3 «mom ”Loo.w oooe “mm..u .mumwem e.» $8.. .38 ”m..oz eooo .xu.gn out.» .xuo.a mamgueou "mcpmoo; we mode» moowgm> me“ :. umu:.ue. mew muwumwgouumgmgu o:.:o..om ogpuu .goo.m oooo Lo woo; uoom .m..a3 oooo Lmeu.m no.3 oc.u._on o.oe.m o no: we; o.o;mmoo; we“ memos N Emu. .u_um.gmuumgmzu em.o.umem an» no.) o:.o..:e wee ammo. no mo; o.o;mmeo: we» «we» cows e sooogeu . mew».-- "mauoz .oaN. ..omo .ooqo .quo .28 .auo. mm oe.. NN.N mm .m.. Nm .o.. oo. mN.N oo. NN.N m=.cumd m:.o:.uxu mo:.u..=m .oz .o>< .o .e me.. mo.m me .e.N om oN.. eo. we.m ee. No.m o.ocmmao: ewe mec.o..=m .oz .m>< .m mu_um.gmuuaguge nguo .m m.m m.mo N.eN o.o m.me N.o o.oe eo. N.N mNo N.om Loo.m\ooo¢\..mz oooe on mucmme< .N do e.¢ o.o. do o.o e. m.N ee. m.e. Nm m.N mzoue.3 mam.e .e .. N.N m.mN m. o.o o N.. oe. o.oN mN N..N Loo.u\moo¢\..o3 coco .m «N o.o m.om oN ..N. .. N.¢ ee. N..o mo o.oN Loo.“ uooe .e no o.om e.wN Nm m.mm Nm o.om ee. e.No .N N.mo woom oooe .m e. o.o o.om m. o.o o N.. ee. N.No mo N.eN m..m: eoou .N m. N.N o.oN m. o.o o N.. eo. ..eN mo o.NN m..mz “mom .. .mcovuwemghueoz .< .N: .E .e: .8 .8 .8 .8 .8 .8 .8 .8 .: xmue. n xmoe. m xmoc. a xmoc. a m.: oo a u a mu. m.. or: N. .4 N: .z mu.um.gmuumgogo em.m.umem gu.z moc.o..=m oe.m=oz ego: go moo m>mx no.2: mv.o;mmeo: mo «cmugme we mexw NNo. :. mammoo: yo mmexw .m m.an~ 70 which has good (modern) walls, a good roof (metal or tile), and a cement floor. An index of housing quality is given in Table 4. Index values range from zero (lowest) to 6 (highest).2 Average values are 1.71 in M1, 2.46 in M2, 0.47 in L1, and 0.69 in L2. (The index value for each household will be used as a proxy for that household's wealth in Chapters VI through VIII.) Thus, modern forms of housing which are both costly and qualitatively superior to traditional housing have become much more common in Moshi district than in Lushoto district. Among the four study areas, M2 is the area with the highest incidence of non-traditional housing and L1 is the area with the lowest incidence. Crops and Livestock Table 5 gives the percentage of households in each of the four areas which reported harvesting the crops listed and which owned one-or more of the types of livestock listed.3 The most important 2The index was derived as follows: A total score was produced for each household based on the presence or absence of one or more household building(s) with each of the six characteristics given below. Each household received 1 point if it had one or more build- ing with the characteristic. Maximum possible total score for each household is 6 points; minimum possible total score is 0 points. Characteristics: (1) Good Wall(s) (5) Good Wall(s), Good Roof, and (2) Best Wall(s) Good Floor (in same (3) Good Roof(s) building) (4) Good Floor(s) (6) Glass Window(s) (For a list of the specific types of walls, roofs, etc., in each of these categories, see notes at bottom of Table 3.) 3Bananas and cassava are excluded from this list because respondents said they were unable to estimate quantities harvested. Bananas are the most important item in the diets in the Kilimanjaro 71 Table 4. Index of Housing Quality in 1973 Value of M1 M2 L1 L2 Housing Quality % Cum. % Cum. % Cum. % Cum. Index % % % % (1) (2) (3) (4) (5) (6) (7) (8) (9) 0 34 34 7 7 66 66 64 64 l 35 69 40 47 28 94 21 85 2 . 7 76 20 67 3 97 8 93 3 2 78 4 71 l 98 2 95 4 l 79 4 75 0 98 l 96 5 14 93 12 87 2 100 2 98 6 2 100 7 100 0 100 2 100 Mean Value 1.71 2.46 0.47 0.69 (N) (442) (460) (300) (290) Derivation of Index Values: A total score was produced for each household based on the presence or absence of one or more household building(s) with each of the six characteristics given below. Each household received 1 point if it had one or more buildings with the characteristic. Maximum possible total score for each household is 6 points; minimum possible total score is 0 points. Characteristics: (1) Good Wall(s) (2) Best Wall(s) (3) Good Roof(s) 4) Good Floor(s) 5) Good Wall(s), Good Roof, and Good Floor (in same building) (6) Glass Window(s) (For a list of the specific types of walls, roofs, etc., in each of these categories, see notes at bottom of Table 3.) 72 .0.0000 0.00.0 000 .000. 0000 00 000.00. 0.0 0.0 0.0 0.0 ..0 0.0 00 NN .N... 00 ..N... 00 .00.. NN .00 . 0N 00000 van 000.0.50 -- -- -- -- -- ..0 . N .00 . . .00 . . .00.. N .00.. N 00.0 N.0 0.0 0.0 0.0 0.0 N.0 MN 00 .00 . 0N ..0 . MN .00.. 00 .0N . 00 00000 0.. 0.. 0.. 0.. 0.N 0.0 00 00 .00 . 00 .00 . 00 .00.. 0N ..0 . N0 00000 ..0 N.0 ..0 -- N.0 ..0 . .. .N. . N .N . . .00.. 0. .00 . 0 0.0000 0.00.: 0.. 0.. N.. 0.. 0.N 0.. 00 0N .N0 . N0 .00 . 00 .00.. .0 .N0 . 00 0.0000. 0 0 --- . --- 0 .00.. 0 --- 0 0.000 0N 0 .000.. 00 .N00. N. .00.. 0 .N0N. 0 . 0000000. N N .00... .. .000. 0 .00.. . .000. 0 0000000 0 . .00N.. N. .00.. . .00.. . .00.. . 0000000 0 0 .0N.. N. .0N . N .00.. 0. .0. . . 000.00 0N .N .0... N0 .00 . 0. .00.. N0 .0N . 0 0000000 . 0 --- 0 .0N . . .00.. 0 .0N . . 00...: 00 0N .N0 . NN .N0 . 00 .00.. 00 .NN .. 00 00000 N0 00 .N0 . 00 .0... 00 .00.. 00 .00 . 00 00.00 0 0 --- 0 --- 0 .00.. 0 --- 0 00000000 0 0 --- N --- 0 .00.. 0 --- 0 0.0003 0. 0 --- 0 --- 0N .00. 0 --- 0 00. NN 00 .0N . .N .NN . 0N .00. 00 .00 . 00 000.00 .N.. .0.. .0.. .0.. .0.. .N.. .... .0.. .0. .0. .N. .0. .0. .0. .0. .N. ... 0000000 .000: N0 .0 N0 .2 0 0 0000. 0 0000. 0 x000. 0 0000. N o.o;emeo: Lee 0005:: meece>< ouez0e. .z0o: N0 .0 N: .2 sea. 0N0. .0000.0 .00000>.0 we 00e0. 0;. we ego: No one oc.czo Le e000.. 0000 we. me.000>eez emugoeem £0.53 0e.o;00eox ..e we emeuceeeee ”x0e000>.0 ece meecu .m 0.00» 73 differences in livestock holdings among the four areas are for cattle (including hybrid or "Eur0pean" cattle) and goats. Only about half the households in M1, L1, and L2 reported owning cattle and only about one-third reported owning goats compared to 91 percent owning cattle and 79 percent owning goats in M2. Table 6 gives the estimated monetary values of the 13 different crops produced and sold per household in the l2-month period preceding the survey. Average total values produced are 837 Shillings in M1, 976 in M2, 725 in L1, and 1053 in L2. Household values will be used in Chapters VI through VIII as proxies for current household income. Coffee accounts for almost all the produce sold in the Moshi areas while it accounts for roughly one-half of the value of crops sold in the Lushoto areas. An average of 468 Shillings worth of food crops (per household) were reported sold in L2; reported sale of food cr0ps is almost insignificant in the Moshi areas and averaged only 93 Shillings per household in L1. One reason that the average value of food crops produced and sold is much higher in L2 than in the other three areas is that in recent years there has been a major govern- ment effort to encourage and assist farmers in the Western Usambaras to grow and commercially market vegetables, large quantities of which are transported some 400 kilometers to Dar es Salaam. As noted earlier, the Soni area is located along the main road up from the area and they are one of the major items in the diets of people in the Usambaras. Some bananas are also marketed, especially in Kili- manjaro. Very little cassava is marketed. Cassava is of more importance in the diets in the Western Usambaras than in the diets in Kilimanjaro. 74 Table 6. Estimated Average Shilling Values (per Household) of Thir- teen Crops Produced and Sold in the Twelve Months Preceding the 1973 Survey.1 Crop M1 M2 L1 L2 Produced Sold Produced Sold Produced Sold Produced Sold (1) (2) (3) (4) (5) (6) (7) (8) (9) Coffee 721 721 707 707 301 301 311 311 Tea --- --- --- --- 72 72 --- --- Wattle --— --- --- --- 10 10 13 13 Cardamon --- --- --- ~-- 5 5 25 25 Maize 64 2 113 6 213 32 147 25 Beans 38 2 69 3 32 ‘ 3 60 ll Millet --- --- 36 19 2 1 2 2 Cabbage 4 l 41 12 50 28 152 128 Onions --- --- 5 1 9 5 28 23 Carrots l --- l --- l 1 59 56 Peppers l --- --- --- --- --- 24 24 Tomatoes 8 3 4 --- 3O 23 230 197 Leeks --- --- --- --- --- --- 2 2 Totals 837 729 976 748 725 481 1053 817 Notes: The table excludes bananas and cassava for which it was impossible to estimate either values produced or sold. --— = less than 1 shilling Exchange Rate: Shillings 7.14 = $1.00. 75 lowlands (and hence, up from the rest of the country) and residents of L2 thus have much easier and cheaper access to outside markets than do residents of the Bumbuli area (L1) and most of the other parts of the Western Usambaras. Estimated average household consumption values of the nine food crops (i.e., the average values of the quantities reported produced and consumed within the households) were 108 Shillings in M1,.228 in M2, 244 in L1, and 237 in L2. The low figure for M1 suggests that bananas may account for a much higher proportion of total food consumption in M1 than in the other’three areas. As indicated in Table 5, only half of the households in M1 report harvesting maize compared to almost all households in the other three areas. Table 7 gives some measures of the distribution of the esti- mated values of the total of all 13 products produced and sold, and of coffee separately (produced and sold). Implications are that the distribution of production and earnings is relatively more equal in the Moshi areas than in the Lushoto areas. For example, the median value of all crops produced is 66 percent of the average in M1, 69 percent in M2, 48 in L1, and 59 in L2. For all cr0ps 591g, median values are over 70 percent of the average values in both Moshi areas but only 46 percent of the average value in L1 and 48 percent in L2. As another indicator of relative inequality, in L2 29 percent of all households reported selling crops valued at less than 10 percent of the average value sold for the area (i.e., less than 48 Shillings sold per household). Twenty-three percent of the households in L2 reported sale values of less than 10 percent (82 Shillings) of the Table 7. Indicators of Estimated Distribution of Shilling Values of All Thirteen Crops Pro- duced, All Thirteen Crops Sold, and Coffee Produced and Sold in the Twelve Months Preceding the 1973 Surveyl Item M1 M2 L1 L2 (1) (2) (3) (4) (5) ALL CROPS PRODUCED Mean 837 976 725 1053 Median 550 677 348 620 Standard Deviation 947 1068 1551 1701 Percent Less Than:1 1/10 of Mean 8 5 10 17 1/4 of Mean 15 12 29 29 1/2 of Mean 35 34 48 45 ALL CROPS SOLD Mean 729 748 481 817 Median 523 524 221 395 Standard Deviation 874 804 1318 1418 Percent Less Than;1 1/10 of Mean 11 9 29 23 1/4 of Mean 19 15 41 33 1/2 of Mean 37 37 50 50 COFFEE Mean 721 707 301 311 Median 522 522 96 144 Standard Deviation 866 700 810 430 Percent Less Than:] 1/10 of Mean 11 9 28 30 l/4 of Mean 18 15 4O 36 1/2 of Mean 37 37 58 51 1 That is, the percentage of all households in the study area which are estimated to have had one-tenth, one-quarter and one-half, respectively, of the mean shilling value for the study area. Exchange Rate: Shillings 7.14 = $1.00 77 average for the area. Only 11 percent in M1 and 9 percent in M2 reported sale values of less than 10 percent of the averages for the respective areas. Thus, the data suggest that although there is considerable differentiation with respect to the values of cr0ps produced and sold per household within each area, there is apparently much less vari- ability within the Moshi areas than within the Lushoto areas. In other words, while average production and earnings may be somewhat higher in L2 than in the Moshi areas, there is apparently also much greater inequality among households in production and earnings so that earnings for most of the households in L2 are probably lower than for most of the households in the Moshi areas. There may be even greater relative inequality in production and earnings within L1 than in L2, implying that production and earnings for the bulk of the households in L1 are very much lower than for the bulk of the households in the other three areas. Estimated coffee earnings are reported separately in Table 7 because ownership of coffee trees is widespread in all four areas, and because Table 6 indicates that in all four areas the value of coffee produced and sold per household is greater than the value of any other single commodity sold or produced. In M1 and M2, 98 and 99 percent respectively of all households reported owning coffee trees; corresponding percentages for L1 and L2 are 87 and 93 (although as indicated in Table 5, not all households who owned trees actually harvested coffee in 1973). Most respondents also reported that the number of trees they owned although it was not possible to indepen- dently verify the number given. The average number of trees per 78 household was about 360 in L1 and M2, 400 in L2 and 460 in M1. The reported number owned by a single household ranged up to 4500 in M1, 1600 in M2, 3000 in L1, and 3800 in L2. However, the number of trees owned by a household is not an adequate indicator of the value of coffee tree holdings and potential production because there can be considerable variations in the age and quality of the trees and in cultivation practices. Local records and experience indicate that average yield per tree is much higher in the Kilimanjaro area than in the Western Usambaras, and the data reported here are consistent with this. Although the reported average number of trees per house- hold ranges only from 360 to 460 in the four areas, the estimated value of coffee production per household is more than twice as high in the Moshi areas as in the Lushoto areas. Occupations Although all four study areas are rural communities in which agriculture is the predominate occupation for both women and men, and virtually all those with agricultural occupations are self- employed, a sizable proportion of adults do have wage-paying, mostly nonagricultural jobs. Table 8 provides information on the principal occupation reported by all ever-married women and by husbands of currently married women. In L1, 10 percent of the husbands reported having wage-paying jobs while corresponding percentages for the other areas were between 20 and 23 percent. Among ever-married women, the percentages who reported having wage-paying jobs ranged from 4 per- cent in L1 to 8 percent in L2. 79 Table 8. Major Occupation Categories: Percentages of Husbands and Ever-Married Women in Each Category in 1973 Major M1 M2 L1 L2 Occupation Husb. Women Husb. Women Husb. Women Husb. Women (1) (2) (3) (4) (5) (6) (7) (8) (9) Professional, Technical, Administrative, Executive 1.1 0.2 1.7 0.8 0.8 O 2.0 0.3 Clerical, Sales, 0 Service 14.8 5.9 10.4 5.6 2.7 4.2 4.7 7.4 Other Non- Agricultural 6.8 O 7.5 0.2 4.6 O 14.8 0.6 Agricultural Self-Employed 76.7 92.3 79.2 92.0 90.4 95.5 77.7 91.3 Agricultural Paid Laborer 0.3 0 0.2 O 1.5 O 0.4 0 Not Working 0.3 1.6 1.0 1.4 0 0.3 0.4 0.3 Total Percent 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 (Total N) (352) (505) (402) (501) (263) (311) (256) (312) Husb= Husbands of currently-married women Women = Ever-married women 80 The relative importance of nonagricultural jobs for men-—and to a lesser extent for women--in M1, M2, and L2 as compared to L1 can be explained by the many more opportunities for nonfarm employment for residents of the former three areas as compared to those for residents of L1. The settlement of Soni is located in about the middle of the L2 study area. Although small, Soni is located at an important junc- tion on the main road running through the Western Usambara mountains and_the only road up from the lowlands to Lushoto village (the dis- trict capital). Soni has several shops, bars, and small hotels and eating establishments; there are also some civil servants, road maintenance workers, and other government employees. Relatively fewer comparable Opportunities exist in L1. Although neither M1 nor M2 are located along major transportation lines, there are daily bus services from both areas into Moshi town, and many residents of both areas commute daily to work there--a distance of about 15 kilometers from M1 and 50 kilometers from M2. Modern Consumer Goods Table 9 provides information on the extent to which pe0ple in the study areas have acquired modern consumer goods. Columns 2-9 give the percentages of all households which own one or more of each item and also give an index value of that percentage for each study area with values for M2 equal to 100. Columns 13-16 give the average number owned per household for each study area. The 22 items are arranged in five groups of 3 to 5 items each, from those that are the most common in the households to those that are least common. A large majority of the households in all 8] .00..0u0..N 000....00 .0000 emcegexm .200..300 :00» 000000 000: 0 00 0:.0> 000000000 e» 000000 a. .0. 0000 .00..e:e 0:0 000 .00.000 0e 0020000 00.00» .mNo. 000. 0. 200. .00.000 0 we 0:.0> 0.00 00000>0 00003.000 000 0. .0. 003.00. 00:.0> 00..—.000 ....0 e» e: e0eeee0 0000 0>00 ..0 000 00.0 0003000 00:.0> ..0. 00.0 000» 000. me 0:.e> 0 00002... .0000: - 0000. --- --- --- --- --- --- - . . 0 . . 00-00.0N0000. 0000. --- --- --- --- --- --- - . . - . . 000 0000 --- --- --- --- --- --- - . - - . . 0.0000000: 000 --- --- ..0 --- --- --- 0 0 0 N N 0 0.000.0 000 --- --- --- --- --- --- 0 0 0 0 0 N 00.000: 00.300 00 --- --- --- --- --- --- N 0 0 N 0 0 0000000 00000000 00 ..0 ..0 ..0 ..0 ..0 ..0 00. 0 0 .00.. N ..N . 0 .00.. N .N0 W 0 0.000 0000 00. ..0 ..0 N.0 ..0 N.0 ..0 00. 0 0 .00.. 0 .00 . 0 00.. 0 .N0 0 0000 00. --- ..0 ..0 --- ..0 ..0 00 0 0 00 w 0 .0. w . 00. 0. .00 0 000.0 000 --- ..0 ..0 --- ..0 ..0 00 0 .. N0 0 .00 0 .00. .. .00. N. 0.0000 00000.0 00. N.0 N.0 N.0 ..0 N.0 N.0 N0 0. N. .00. 0. 00 . 0. .00. 0. .00 N. 00000 00.00000 00. N.0 0.0 0.0 ..0 0.0 0.0 00. 0N .0 .00 00 0N . 0 .00. N0 .00 00 0000: 00N 0.0 0.0 0.0 N.0 0.0 0.0 0N 0N 00 00 w N0 .00 w 0. 00. 00 M00 .0 0.000 00 N.0 0.0 0.0 ..0 0.0 0.0 00 0. 00 00 0. ..N 0 00. N0 0N 00 000 .000: 0. 0.0 0.0 0.0 0.0 0.0 0.0 00 00 00 .N... 00 ..N . 00 .00.. N0 .0... 00 .000.. -000... 0000. 0. 0.0 0.0 N.0 0.0 N.0 0.0 N0 N0 .0 .00 . 00 .00 . 00 .00.. 00 .0N . 00 0000000 0. 0.0 0.0 N.0 0.0 0.0 0.0 0N 00 00 .N0.. 00 .00 . 00 .00.. 00 .0... 00 0..00000 00 0.0 0.0 ... 0.0 0.. N.0 00 00 00 .00 . 00 .N0 . N0 .00.. 0N .0N . 00 0.000 00 0.0 N.. ... 0.0 0.. 0.. 0N .0 00 .N0 . N0 .N0 . 00 .00.. NN .0N . .0 0000000: 00 0.. 0.. 0.. N.. 0.. N.. .0 0N .0 .00 . N0 .0N . 00 .00. 00 .N0 . 0N 000 00000: 0N 0.N 0.N 0.N ..N N.N 0.N N0 0N NN .0... 00 .0 . N0 .00. - .0N .00.. .0 0.000 0N 0.. 0.. ..N N.. 0.N 0.. 00. 00 00 .00.. N0 00.. 00 .00.. 00 .00 . 00 000 00000.0 .0.. .0.. .N.. .0.. .0.. .0.. .0.. .N.. .... .0.. .0. .0. .N. .0. .0. .0. .0. .N. ... 0.: a a x000. a x000. u x000. a x00:. n 00 0 0m.0> 0-0 0.: N0 .0 N: . .0 .0-0 0-0 0.: N0 .0 N: .z 0.. -..:m e.e000=e: 000 000502 00000>< .00.uNz. 000.00. a 200. 0a.: 00.000000: 00 «000000 20». Mug :0 308 0232.00 vmuum—mw we :330300 of. .m 03o... 82 four areas own one or more of the three most common possessions-- clothes box, chair, wood bed. At the other extreme, only a small fraction of the households own one or more of each of the eight or so less common--and generally most expensive--items at the bottom of the table. However, among the l0 or ll items in between, a consider- able differentiation in the incidence of possession is apparent. Without exception, Ll has the lowest percentage owning each item, and in many cases the percentage is only about half as high as it is in the other three study areas. There is generally not much I difference among the other three areas. Agricultural Equipment The agricultural self-sufficient character of the study areas is also reflected in Table 10 which gives the percentages of house- holds which had at least one of each of the 13 different agricultural tools on equipment items. The items are grouped from most common at the top to least common at the bottom. The first three items--hoe, panga (similar to a machete) and axe--are found in almost every household. Between 97 and 100 percent of all households in each of the areas has a hoe; the average number per household ranges from 2.7 to 2.9. Between 95 and 99 percent have a panga; axes are owned by 89 to 95 percent of all households. The area with the highest proportion of households having each of the items is Ll. Religion, Education, and Health In this section the spread of non—indigenous religions during this century will be summarized and data will be given on the effect of religion on education. Detailed information will also be given 83 000 .00.000 00 0000000 00... u ....N 000....00 00.000 .mnm. 000. 0. 500. .00.000 0 00 00.0> 0.00 00000>0 00005.000 000 0. .0. 050.00. 000.0> 00....00= .0000 000000xm .00..00a 000 . ....o 00 00 0000000 0000 0>00 ..o 000 00.0 000x000 000.0> ..0. 00.0 0000 000. 00 00.0> 0 00005--q .0000: 0000N 0 - 0 0 0 - 03000. 000.. --- --- --- --- --- --- - - 0 - - . 0000.00 .0000. 0~.0: 00. ..0 0 0. 0 m 0 0 0000 00 --- 0.0 --- 0.. 0.... N . N.. .N . . 0 .00.. 00 .00 . N0. .500 0000 00 0.0 ..0 0.0 N.0 ..0 ..0 000 ..N 0 .0N0. N0 .00.. 0. .00.. 0. .0N . N 00.00.000 00 ..0 0.0 N.0 ..0 0.0 0.0 00 .. .0 .00 . N. .0. . 0 .00... .0 .N.. . 00 000.000 00 0.0 0.0 0.0 N.0 0.0 0.0 N.. ..N .0 .00 . 0N .... . ..N .00.. 00 .0. . .N.. 000.00 000.00 00 0.0 0.0 0.0 0.0 0.0 N.0 .N 00 N. ..N . 00 .0. . 00 .00.. 00 .00 . 0.. 0.0.0 000.00 00. N.0 ..0 N.0 N.0 0.0 0.0 ..N 0. 00 ..0 . N. ..0 . 0. .00.. 00 .00.. .0 .60 00.0000. -000 000000 0N N.. 0.0 0.. N.. 0.. 0.0 ..N. 00 00 .00.. 00 .00 . 00 .00.. N0 .00 ._ 00 00> 00.000. -000 000000 0. N.. ... N.. N.. ... ... 00. ..0 00 .00.. N0 .N0.. 00 .00.. 00 .00.. 00 0.2 .0000005. 0. 0.N N.. 0.. ..N 0.. N.. 8. 00 00 .00.. 00 .00.. 00 .00.. 00 .00 . 00 00000 00 0.N 0.N 0.N 0.N 0.N N.N N0. 00 N0 ..0: 00 .00.. 00. .00.. N0 .00.. N0 00.. .0.. .0.. .N... .0.. .0... .0.. .0.. .N.. .... .0.. .0. .0. .N. .0. .0. .0. .0. .N. ... 0-0 0.: N0 3 N0 .0 m-.. 0 0 x000. 0 x00... 0 x00... 0 x00... 0. 0.000000: 000 000502 0a000>< mm-” 0-0 0.: N0 .0 N: .: Eng» .0070: 000.00. 0 000. 00.; 00.000000: ..0 0000000 000. m.m. 0. 00050.0cm .0000.00.0m< 00 00.0000000 000 .c. 0.000 84 on levels of educational attainment by age group for both women and men. Some information will also be given on health and health services and their utilization by pregnant women. Access to both education and health care are important components of levels of living and well-being, and differential participation indicates in part the extent to which social services are relatively evenly or unevenly distributed and utilized within and among rural areas. The Spread of Non-Indigenous Religions Christianity was introduced by missionaries in the late nine- l teenth century and the first Christian missions were established in both Lushoto and Moshi districts in the l8905. Records from the Lutheran and Catholic offices in Moshi indicate that in 1910 about 4 percent of the population in the entire Kilimanjaro area were Christians. By l935 about 40 percent were Christians, and by l970 about 80 percent were Christians [my calculations based on Maro, l974: 254]. Throughout this period approximately two-thirds of all Christians were Catholics and one-third Lutherans. In the post-World War I period Christianity expanded much more rapidly in the Kilimanjaro area than in the Western Usambaras although missionaries were active in both areas. This was due in part to the competition between Catholic and Lutheran missions for followers among the Chagga. A comparable situation did not develop in the Usambaras, but Christian missionaries there had considerably more competition from Islamic influences. The Moslem religion was originally introduced into the Usambaras from the coastal areas around Tanga and Mombasa, and apparently commencing during the latter 85 part of the nineteenth century, the number of followers of Islam among the Sambaa increased relatively rapidly [Feierman, l974]. As indicated in Table l (Chapter III), the pr0portions of the population who were Christians in each of the study areas increased from about zero at the turn of the century to 82 and 99 percent respectively in Ml and M2, and to 18 and 25 percent respectively in Ll and L2. Educational Attainment It is the policy of the Government of Tanzania to provide universal primary education to all children in Tanzania as soon as possible, which would be grades l-7 (until about l0 years ago there was also primary grade 8). Children are supposed to enroll in grade one at age 7. However, they frequently first enroll at older ages, sometimes as old as age l3 or l4. In fact, for the country as a whole in l973 only about half of all children of primary school age (approximately ages 7-l5) had ever attended school [Mwakalasi, 1975]. Information on proportions of residents with formal education was presented in Table 2 in Chapter III. Data on highest levels of educational attainment by age groups are presented in Table ll. They are given for surviving children ages 7 to 29 (regardless of where the children now reside), for all women over age l9, for husbands over age l9 of currently married women, and for mothers and fathers of all ever-married women.4 4Data for "children" ages 20-29 consistently show higher levels of educational attainment than for women and husbands ages 20-29, although the differences are much greater among men than among women. Differences are probably primarily because the data do not .cwso: voNLgut Na Managua. £36 o o o ooN o o ooN o o o ooN o o o ooN o o o ooN o o. o ooN ma.aca +oo o o N mm m N o oo o o « mm o N oN mm o o N mm o N N NM 2.9.5 om-o« o N N No N m MN NN o M m mm N M «N on o o w «a o m NN MN me.a MM-ON Noun m.umvcamwm Any mucmcoaa o o o ooN o N oN MM 0 o o ooN o M 0N NN o o o ooN o o NN mm «N.-ucq +oo o N M Ma N o MN No 0 o oN om M N MN «m o N M mm o e «N oN MM-«NmN om-o« N N m mm N mN o« M« N « NN Mm N «N «m NM N o N NM « oN eN «m M«-«MmN MM-oM 0 ON mN «N N MN mm NN o «N MN No M oM o« MN N m «N cm 0 oN on «N M - mvconmax cwsoz N NN NN No NN NN m« MN N MN mN mm m .«N on N o N MN mm o NM NM MN Mm-««mN mNuoN o NM MN «M M mM N« oN 0 MM oM MN « o« N« m o oM NN N« N OM M« MN Mm-«mmN aNlmN o NN NM om 0 MN om MM o NN om NM o MN om NN o NN MN mo N NN o« N« Mo-mmmN «N-oN o N «N mm o N NN mm o N ON MN o N mN «m o o N Mm o o N Mm oc-«omN m -N cucuNNnu ONozmao . N4 .4 o N o mm o o N mm o o o ooN o N o mm o N o mm o o N am mm.-oca +oc o o N mm o o N mm o o N mm o o N mm o o N mm o N N mm mN.-ocn mm-o« o N m «o o M N ca 0 N « mm o M o No C N m «a o « m mm mmrloca MM-0N Noun m.uuamw¢ may mucmgmac o o N am o N NN NM 0 o N am o o m mm o N o mm o m oN mN «N.-ugn +oo o M N oo o a NN oN o M o mm o m MN on o « « No o NN mN ,NN MM-«NmN mm-o« o M MN mo M 0N MM MM o m oN no M NN om OM N N MN MN N «N oN m« M«-«MMN MM-OM N MN MM M« N oM w« 0N N NN N« MM M MN Nm NN N «N «N Nm N NM NM MN Mm-« m MN-ON mucmnmzz case M MN «M MM NN w« mN N M MN N« MN 0N o« MN m « NN NN N« MN m« MN oN mm-««mN MN-MN M o« NM NN NN om «N « N Mm MM oN NN No «N M m «« NN «N NN Mm «N o mm-«moN MN-MN o MN mm mN o MN Mo «N o NN Nm NN N «N No MN 0 mN Nm «M o NN mo «N Mo-mmmN «N-oN o N oN mN o N «N «N o N ON MN 0 N MN mN o N oN aN o N MN NN oo-«waN m -N :mLuNNzu Nxmoz N: N: NoNN NMNV N«Nv NMNV NNNN NNNV NONV NMNV NMNV NNNV NoNV NmNV N«NN NMNV NNNN NNNV NoNv va va NNV Nov va N«V NMN NNV . NNV +m wim «nN o +m mnm «IN 0 +¢ mum «.N 0 +0 mum «.N o +m m-m «IN C +m mum «-N o mmNaewu mmNoz moneou mmNo: mmNoeoN mmNmz cgom usage meow» om< acNNoozum Ncsgou menu» .02 chNOOgum Nascou mumm> .oz chNoogum Naficom meow» .oz .xoggq< acosgsu MNNN No ma Ngomouuo chNoocum No-mcmo> comm cN amoucougua "NewscNouu< NacoNuouzum umonme .NN mNnuN 87 As noted in Chapter III, there are only small differences among the areas in the proportions of younger men who have received some formal education. However, differences are somewhat greater for the proportions who have completed more than four years of school, and the biggest differences are among male "children" ages 20-29. About two- thirds in both Ml and M2 received more than 4 years of education; 13 and 20 percent respectively received more than primary school education. In Ll and L2, 44 and 33 percent respectively received more than 4 years of schooling; 6 and 9 percent respectively received more than primary schooling. Very few husbands age; 20-29 have more than primary schooling which suggests--as would be expected--that almost all men ages 20-29 born and raised in the area who went beyond primary school are now residing outside the area. Among women ages 20-29, about one-fourth in both Ml and M2 received more than 4 years of schooling compared to only 6 percent in Ll and 14 percent in L2. (Figures for female "children" are only slightly higher.) Among older women, percentages who have received more than 4 years of schooling decline sharply; for women ages 30-39 the percentages are 8, 8, l, and 5 respectively in Ml, M2, Ll, and L2. Among women ages 40-59 the percentages are 4, 3, l, and 0 respectively while for husbands ages 40-59 the percentages are l2, 6, 6, and 10 respectively. Almost no women residing in the study areas have received any post-primary education. refer to quite the same populations. More highly educated "children" are more likely to be living outside the study areas, probably in an urban area. There may have also been some tendency for women to report somewhat higher levels of educational attainment for their children than were actually achieved. 88 Thus, educational data from the study areas show that educa- tional levels are above the national average in all four areas, and in M2 they are probably about as high as in any rural area in the country. They have improved gradually in all four areas during the last 50 years or so, but they have improved most rapidly and covered a higher proportion of the population--and particularly of the women--in M2 than in the other three areas. The least improvement has been in Ll. Education and Religion ' Table l2 presents data on religious differentials associated with differentials in educational attainment. The figures are striking. Data are presented only for Ll, L2, and Ml since 99 percent of the respondents in M2 are Christians. In Ll and L2, com- parisons are between Christian and Moslem respondents; in Ml compari- sons are between Christians and those who still follow traditional beliefs. Data are presented for husbands, all women, and male and female children of ever-married women identified separately by both father's and mother's religion. Among women and husbands ages 20-39, educational attainment among both Christian husbands and Christian women compares very favorably with the overall educational attainment of the same age groups in M2 (Table ll). Among children ages lO-l9 (and particularly among those ages lO-l4), religious differentials are much less pro- nounced for boys although they are still significant for girls. This no doubt reflects at least two things: One is that all primary schools were taken over by the government during the l0 years or so 89 MN MM MM .MN N MN MM MN .NM.N MM MM MM .NM N .M MM MN .NN N MM M. NM .MM N MM .M MN .M. M.-M. M NM MM .NM N M. MM MN .MMNN N MM MM .MM N MM MM .. .MM N .. M. MN .MM N M. .M MM .NN M.-M. M .N MN .M. N . M. MM .MM N M N. MM .MM N M MM MM NMN N M M NM ..M N M M MM. .M. M -N MM.M..MM M.LMMMM: MM MN MM .MM N MM MN N. .MN N NM MM NM NMM N MM NM M .M. N MM .N MM .MM N NM MN MN .N. N M.-M. M. MM MM NMM N M. MM MN NMM.N M MM MM NMM N MN MM M. WNM N .. MN MM NMM N M. MN NM ..N N M.-M. M N. MM .M. N . M. .M .MM N M N. MM .MM N M MM MM .M. N M M NM .MM N M M MM. WM. N M -N co.M..MM M.MMMMMN MMLM..MM o.osaN MM N. N. .MN NN MN M .MM N NM MM M .MM W NM MM M. .N MN MM NN .MM MM M. M. .N. N M.-M. MM MM .. .NN N. MM M. .MM W M. MM MN .MN M. MM M. MN .. MM MM NM NN MM MM M. M.-M. M. M. MN .M. M MN NN .MM . M. MM .MN N M M. .M .MN M M NM MN M M MM N. M -N MOMMMNMM M.LMM.M: MM MN M 2M N MN MN M ..o N MM .M M WMM N M MM MM .M. N MN MM NN NMM N NN M. M. NM. N M.-M. MM .M M .MM W N. MM M. WMM W N. MM MN MN W MN NM M. .MN W .. MM .M NM W MN MM MM MN M.-M. M MN MN .MN . NN NN WNM . M. MM MN M NN MN .NN M M NM MN M N MM M. M -N MM. ..MN M.LMMMMN 5.2.5 «.N..: N M MM NMM MN M. .M .MMMW M .. .M NNMNW N. MN NM .NM W M M MM .M.MN MN MN .M NMM MM. M M MM. .M. N M MM .MM M M MM. NMN M M MM. .M M M MM. WM. N M M MM. .M +MM M M MM. NMM M M MM .M..N M N MM .MM W M MM NM ..N N . M MM .MM N M .N MN M..N MM-MM M M MM. ..N W M. MN NM .MM N N .. NM NNM N. N. MM M. N M M MM .MM W M NN MM .. MM-MM M M MM. .M MN MN MM NNM.N N. N. MN .MN N NN MM MM MN N M M MM .MM N. MM MM .N. MN-MN 58.. ..M M . MM .MM N MN NM MM .MMNN .. MM MM NMM.N MM MM .N .MM N M NM MM .NMNN NN .M NN .NM N +MN M M MM. .M. N M MN NM .MN N M MN MM .MN N MN MN MM .M W M M. MM .NN W --- --- --- .M N +MM M M MM. .MM N M. MN MM ..N N N NN .N .MM N MM MN MM .M. M .N MN ..N M NM MM WN. N MM.MM M M MM. .M. N MM MM . NM .MM N M. MM MM NMM N MM MM M. ..N N M. MN MM .MM N MM MM M. .N N MM-MM M M. MM NN N .M .M M. .MM N MN MM MN .MM N MM NM M .M. N M. MM .M .MM N MM .M M .M. N MN.MN Muconmaz .MNN .MNN .MNN .NNN ..NN .MNN .M.N .M.N NN.N .M.N .M.N .M.N .M.N NN.N ...N .M.N .MN .MN .NN .MN .MN .MN NMN .NN ..N N N N M N M M M M M M M N N N N M M MM M-. MMMz NzN +M M-. oMoz .zN +M M-. acoz .zN +M M-. Mcoz NzN +M M-. ocoz .zN +M M-. MMMz NzN EoNMoz coNuMNgzu eoNMoz cMNuMNugu EmNmo: :MNMMNgsu Maogw N: NM NM oa< MNMN No MM conNNom Mn acoscNauu< NocoNuMoavN :N MNMNucucuuMNo .NN oNnMN 90 prior to the survey and religion became no longer a criteria for entrance into what were formerly mission schools. Hence, subsequently about as high a pr0portion of non-Christian as Christian boys were apparently enrolled in primary school. Second, the much higher pro- portions of non-Christian than Christian girls ages lO-l9 who never attended primary school no doubt reflects greater reluctance by non- Christian parents to allow their daughters to attend school. ,Thus, religious differences--and the historical pattern by which Christian missions and mission activities spread in the study areas--appear to account for many of the observed educational differentials among and within the study areas. 52213.1. In addition to their evangelistic and educational activities, the missionaries established hospitals and clinics. Largely as a consequence of mission medical activities, the Kilimanjaro and Western Usambara areas are relatively well-endowed with health facilities. In l972 Lushoto district ranked first and Kilimanjaro district (encompassing the current Moshi, Vunjo, and Rombo districts) third among the 56 non-island and non-urban-periphery districts in Tanzania in the proportion of the population within l0 kilometers of any health facility. Only l.5 percent of the population in Kilimanjaro district and 0.5 percent of the population in Lushoto district were not within 10 kilometers of a health facility [Thomas and Mascarenhas, l973: 23, 46, 48]. Female expectation of life at birth probably did not exceed 30 years at the turn of this century and may have been lower. Based 91 on reported age distributions, female expectation of life at birth today is estimated to be in the range of 50 to 55 years in all four areas. For Tanzania as a whole and for rural Africa generally, these are high levels. Although survival rates have certainly increased in all four study areas in recent decades, it is not known when they started to increase, how fast, or how much they have increased. Residents of all four areas lived relatively close to health care facilities; probably all lived within no more than a 2 or 3 hour walk from either a clinic or a hospital. Table l3 gives (a) the percentages of women who visited a prenatal clinic at least once during the final three months of their most recent pregnancies, and (b) the percentages of women who delivered their most recent babies in a hospital. Grouping all women age 20 or more together, 57 per- cent in Ml, 56 percent in M2, 58 percent in Ll, and 59 percent in L2 reported having visited a prenatal clinic. A comparison of ages 40-44 with ages 20-24 gives the following percentages: Ml, 59 and 86 percent for ages 40-44 and 20-24 respectively; M2, 60 and 97 percent; Ll, 57 and 74 percent; L2, 67 and 88 percent. Thus, the pr0portions of women attending a health facility for prenatal care have apparently been increasing steadily over time and are progressively higher among younger women. In all four areas, over half of the women ages 45-49 reported not having attended a clinic during their last pregnancies, while among women in the next younger age group (40-44) more than half did in all four areas. However, while all four areas are similar in the overall proportions having visited a clinic, among those under age 35 the highest percentages are in M2 and the lowest in Ll, although in both Ll and 92 Table l3. Medical Care During Pregnancy: (A) Percentage of Women Who Visited a Clinic at Least Once During the Final Three Months of Their Most Recent Pregnancy, and (B) Percentage of Women Who Delivered Their Most Recent Baby in a Hospital Current Age Group Ml M2 Ll L2 Moshi Lushoto (l) (2) (3) (4) (5) (6). (7) A. Percentage Who Visited a Prenatal Clinic 20-24 86 97 74 88 90 82 25-29 86 82 79 80 87 79 30-34 75 92 68 85 83 77 35-39 79 76 60 59 77 60 40-44 59 60 57 67 59 62 45-49 44 35 48 28 39 38 50-54 2l 12 48 25 l6 36 55-59 4 9 l8 7 7 l4 60+ 6 2 4 ll 4 8 20+ 57 56 58 59 57 59 B. Percentage Who Delivered in a Hospital1 20-24 74 7l 25 36 73 30 25-29 79 62 l9 42 70 29 30-34 55 52 13 39 54 27 35-39 El 57 ll l6 54 l3 40-44 46 42 ll 2l 44 l6 45-49 43 26 l9 7 34 l3 50-54 20 9 0 l0 l4 5 55-59 l2 9 0 0 ll 0 60+ 7 l 0 4 4 2 20+ 47 38 13 25 43 l8 WW 1Includes women who delivered at home attended by a medically-trained midwife (about two percent of the total in this category). 93 L2 (and particularly in L2) a large proportion of those who visited a clinic reported having done so more frequently than monthly. Overall, 47 percent of the women in Ml, 38 percent in M2, l3 percent in Ll, and 25 percent in L2 reported that their most recent baby was born in a hospital.5 Among women ages 20-24, the corresponding percentages are 74 in Ml, 7l in M2, 25 in Ll, and 36 in L2.6 In Chapter VI hospital deliveries will be used as a proxy variable for quality of health care received by the woman and her children. A few further comments about incidence of hospital deliveries will be made in Chapter VI. EEEEEIZ. This chapter has presented data on agricultural and other economic and social characteristics of the study areas in 1973 as well as some information on changes which have occurred since about 1900. Considerable changes have taken place in all four areas but the pattern and extent of the various changes have differed substan- tially among the areas. Land pressures have greatly increased in both districts resulting in ever-smaller land holdings. Residents have responded by reducing or eliminating grazing areas, reducing average cattle holdings and generally intensifying cultivation. Over 5This also includes women who delivered at home attended by a medically-trained midwife, but these deliveries accounted for only about two percent of all deliveries in this (i.e., hospital) category. 6Percentages were even higher in all areas but M2 among women ages l5-l9 who had given birth, but there were only 18 women in Ml, 7 in M2, 6 in Ll and 9 in L2 in this age group who had given birth (respectively, l4, 4, 9, and 15 percent of all women in this age group . 94 this same period, commercialization of agriculture has expanded throughout both districts although agriculture has been more prosperous in Moshi district than in Lushoto district. This is probably due largely to the greater success in producing coffee in Moshi district. Despite growth in cash incomes and acquisition of material possessions, average incomes remain rather low in all four areas. Some of the literature on agriculture in Lushoto district suggests that overall productivity may have deteriorated during this century; data from this study neither support nor contradict this contention. In both districts, however, the land is supporting about four to five times as many pe0ple as it did in l900. When the areas are ranked according to indicators of material or economic well-being, Ll invariably ranks lowest and in most cases M2 ranks highest. The most striking differences are in quality of housing where the ranking from best and most expensive to poorest is M2, Ml, L2, and Ll. M2 is also well above the others in ownership of livestock. In the possession of consumer goods and agricultural equipment, the residents of Ll are the worst off while the differ- ences between the other three areas are relatively modest. However, the area with the highestestimated average value of cr0ps produced and sold per household is L2. The ranking is L2, M2, Ml, and Ll. Sale of food crops--particularly tomatoes and cabbage--accounted for almost 60 percent of the total estimated value of crops sold in L2, but sale of vegetables on a substantial scale has come about only during the last few years. The value of nonfood crops sold per household in the Lushoto areas is only about half that of the Moshi area. Hence, the relatively high production and sales levels per 95 household in L2 are undoubtedly quite recent, while it is likely that average production and earnings from cash crops in Ml and M2 are not much higher (and may be lower) than they have generally been during the past 25 years or so. Moreover, analysis of evidence on distribution of crop production and earnings within the areas show relatively greater inequality in the Lushoto areas than in the Moshi areas and suggests that most residents in the Moshi areas produce more and earn more from their cr0ps than do most residents in L2. Most residents of Ll produce and earn far less than do most residents of the other three areas. The distribution of participation in education and health services is also somewhat more unequal in the Lushoto areas than in Moshi. In the Moshi areas there are relatively fewer women with no education and relatively fewer women who did not receive medical care, at least during pregnancy, although a large proportion (i.e., about 80 percent or more) of younger women in all four areas reported receiving some prenatal medical care. Estimates of crude death rates indicate that mortality rates are relatively low--for rural Africa-- in all four areas. Thus, social and cultural changes and health, education, and economic improvements have taken place in all four areas during the twentieth century, although the amounts and rates of change and the extent to which the changes have permeated and transformed the societies appear to differ considerably among and within the areas. Within the framework presented in Chapter II, these are societies in transition; at present they are neither "traditional" nor "modern." In the next four chapters interrelationships between some of the 96 social and economic characteristics of these societies and the supply of and demand for children will be analyzed. CHAPTER V ANALYSIS OF THE DETERMINANTS OF SUPPLY OF CHILDREN BORN The basic supply variables and the supply model of numbers of children born outlined in Chapter II will be analyzed in this chapter. First, a formal statement of the model will be presented. Some characteristics of the variables, such as average values, distribution by age groups, and trends will then be described. This will be followed by description and analysis of the results of the regres- sions. Specification of the Basic Supply Functions It is hypothesized that number of children born will be posi- tively related to woman's current age, negatively related to woman's age at first marriage, negatively related to polygyny, negatively related to duration of breastfeeding, and positively related to number of children who die (due to an expected shorter average birth interval between births of children who subsequently die and the following births, compared to the average interval between births of children who do not subsequently die). The basic supply model is given below: 97 98 (la) NBORNij = f (WOMAGEij, AGEMARij, DHPOLYij, BRSTFEDij, NDIEDij, e) where: NBORNi. the number of children born alive to woman i in J age group j WOMAGEij = the age of woman i in single years AGEMARi. = the age at which woman i married (as a proxy for 3 age at first coitus, or at least the onset of frequent coitus) DHPOLY . = a dummy variable whose presence indicates that 13 the husband of woman 1 is polygynous (as a proxy for frequency of coitus and perhaps for duration of post-delivery sexual abstinence) the reported duration of breastfeeding (in months) by woman i (as a proxy for duration of infecundity following the birth of a child) BRSTFED.. lJ NDIEDi. the number of woman i's children who have died 3 (as a proxy for duration of infecundity following the birth of a child) e = random sample i = each currently married woman, married only once, in age group j the age group; it may be a 5-year, lO-year or n-year (.1. II The direction of expected relationship between each indepen- dent variable and the dependent variable is shown in the parentheses below: WOMAGE (+), AGEMAR (-), DHPOLY (-), BRSTFED (-), and NDIED (+). Below is an alternative formulation of this basic model in which WOMAGE and AGEMAR are collapsed into the single varible, YRSMAR (for the number of years the woman has been married), where YRSMAR = (WOMAGE - AGEMAR). 99 (lb) NBORNij = f (YRSMARij, DHPOLYij, BRSTFEDij, NDIEDij. e) The expected relationship between YRSMAR and NBORN is of course posi- tive, and the directions of the expected relationships between the other independent variables and NBORN are unchanged from formulation (la). Regression results for both formulations will be presented and discussed later in this chapter. Information on Some of the Variables Table l4 gives average values and standard deviations for reported number of children born per woman, controlled by the current age group of the woman. Table 15 gives averages and standard devi- ations for reported age at first marriage and reported duration of breastfeeding. Table l6 shows the percentage of all women in each age group who are currently married and have been married only once. Table 17 gives the reported incidence of polygyny among husbands and their fathers and the proportions of women married to polygynous husbands. The upper portion of Table l4 shows only relatively small differences among the four areas in numbers of births for women who are currently married and married only once. Some differences are suggested and may be important; for example, the average number reported born is sharply lower for women age 50 or older compared to women 40-49 in the Lushoto areas while in the Moshi areas the average number reported born is up slightly. This probably results in part from less complete recall of deceased children among women in the Averages and Standard Deviations for Numbers of Children Born per Woman, Controlled by Women's Current Age Group Table 14. A A O D m 4" U) r- o V v M: U) 3 _J o A Q) N > I— < V A A D '— °t- U) v- .2 V v m 0 Z 0 A G) O > I'- < v A A D Ch (I) v N V .—l M .N > co < v A A D IN (I) v u— v _l q; A > £0 < V A A (3 LO (1) V N v Z d .N > <' < V A i A O m (I) V r— V Z M .N > N < V D. a): A mo '- Mg umugoaog cmsoz MN umcmmz Mama MNco NQMM MM.MM.>MM MNMMMMMM u MM. NM.N.N N.NM WM.M.N ..MM .M. M.N M.MM WM...N N.MM .M.N.N M.MM .M. N N M.MM +MM .M...N M.MM .M.M.N M..M NM. M N M.MM WN...N M.NM WM.N.N M..M WM. M N M..M MM-MM .N.M N ...M .M.M N M.MN .N. M N ..MM ...M N N.NM .M.M.N M.NN .M. M N M..M MM-MM .M.M N M..M WM.M N ..MN .M. M N N.NN .M.M N N.MM .M.M N M.NN .N. M N M.NN MM-MM WM.M M.MN .M... M.MN WN. M N M.MN .M.M W M.NM M..N.N N.NN .N. .. .W M.MN MN-MN .N.M. M.MN .N... M.NN .M. M N M.MN .M.M. M.NM ..N.N ..MN .M. M. M.MN MN-MN ngucoz cNN MMuszmz MmNnM m NMMNuouuuxmz NNoch ummummmgm :Meoz meNN No spmcmN um ugonmx .M.M N M..N .M. M N N.NN .M. M N N..N .M.M N M..N .M.M N ..MN .N. M N ..NN +MM WN.M N M.M. .M. M N M.M. ... M N M.M. .M.M N M.M. .M.M N M.M. WM. M N M.M. MM-MM W..M N N.M. WM. M N M.M. .M. M N M.M. .N.M N M.M. NN.M N N.M. .M. M N M.N. MM-MM .N.M N M.N. WM. M N ..M. .M. M N M.N. .M.M N N.N. .M.M N M.M. .N. M N M.N. MM-MM WM.M N M.N. .M. M N N.M. ... M N M.N. .M.N N ..M. ...M N M.M. .M. N W M.N. MN-MN .M.N N ..N. .M. N N M.N. WM. N N M.N. .M.N N M.N. .M.N N N.M. .M. N M.N. MN-MN NMgMMN :NN mmmNNNMz um mm< mmmgm>< vmugoamm .M.N NN.N ...N .M.N .MN .MN .NN .MN .MN WMN .MN WNN ..N WMMN .M>< .MMN .MMM .MMN .MMM .MMN .M>< WMMN .M>< ..MMN .M>< M :95 MMMMMM. .MMMz NM .4 N: .z MMM Maogw mm< pcmcgzu MM nmNNochoo .MNco mono MMNNNMz .cmsoz umNNgmz MNucmggau NoN chcmoNuMMmgm .M. M.MM. No :oNMMNMQ MMMNMMMN ucm MMMNNNMz MM mo< umugoamm Now McoNMMN>mn.oLMMMMMM ucm Mmmmgw>< 103 to those now ages 30-34.1 Data to be discussed later, however, suggest that once married, average fertility is probably higher (at least among younger women) in M2 than in the other areas. Table 15 also gives reported duration of breastfeeding; this is for the next-to-last baby weaned unless the woman weaned only one baby in which case it refers to that baby.2 Reported average lacta- tion declines almost uniformly from oldest to youngest age groups in a11.four areas. Among women 50 or older, the average is 39 months in L1 and 35 months in the other three areas. The decline in reported average length of lactation from women ages 50 and over to women 20-24 is 7 months in L1, 9 months in L2, 11 months in M1 and 15 months in M2--down to an average of 20 months in M2, 24 months in M1, 26 months in L2, and 32 months in Ll.3 1Even if reported accurately, one would expect average age at first marriage to rise as current age rises. Those not married when they were in a younger age.group raise the overall average when they subsequently marry. th is hoped that the length of lactation reported by each woman is representative of her entire breastfeeding experience. In pre-testing the questionnaire it was found that some older women reported an extraordinarily long lactation for their last child, apparently due to the woman's reluctance to wean the child which she assumed was her last. Since the nursing of these children is probably unrepresentative of these women's general breastfeeding experiences, the questionnaire was revised to obtain reported length of lactation for the last two babies breastfed to weaning (i.e., children who did not die before being weaned). 3The likelihood that average length of lactation has been declining in recent years is reinforced by the overwhelming pr0por- tions of women who believe that it is. When women were asked whether length of breastfeeding is longer, shorter, or about the same today as it was when their mothers were breastfeeding, the following per- centages said "shorter today": M1, 96; M2, 91; L1, 86; and L2, 90. 104 Since the analysis will be limited to women currently married and married only once, Table 16 gives the percentage of all women who are in this category. Again, among women under age 25, the propor- tions married are much lower in M2 than in the others, with relatively little difference among the areas for women age 25 and older except that among women age 50 and older, only half of all women are cur- rently married, once only in M1 compared to about two-thirds in the other three areas. It was hypothesized in Chapter II that polygyny would result in somewhat lower fertility. This may be due to numerous factors, including a lower average frequency of coitus for women with poly— gynous husbands, a longer average period of post-birth abstinence from sex, and a longer average duration of breastfeeding. Table 17 shows the reported incidence of polygyny among husbands and their fathers (controlled by age group of the husband) and the percentages of women currently married to a polygynous husband. The data suggest that a generation or two ago the incidence of polygyny was similar in all four areas; among husbands currently age 60 or older, the percent polygynous ranges from 36 in M1 to 47 in L2; the average for husbands' fathers ranges from 63 percent in L1 to 75 percent in L2.4 However, among men under age 60 in M2, the reported incidence ranges from 2 percent for those 20-29 to 12 percent for those 40-49. This compares 4Even if reported accurately, one reason for a higher fre- quency among fathers is that for fathers it means "ever polygynous" while for husbands it is "currently polygynous." In addition, one would expect that the proportion of sons of polygynous men would somewhat over-represent the proportion of those men who were actually polygynous, on the assumption that polygynous men have a larger average number of sons than monogamous men. 105 .MNco muco OMNNNMe .OMNNNMs MNucmggzu pcwugma u OZO pcwugma ON NMN NM OMN ON «M MM MM MM m«N OM m«N +Om MN MNN OM NwN NM MM NN OM mm NO «N mm m«-O« mm Mm mm MNN Nm M« «M m« Om MN NM MM mmumm MM Ow «w MMN Mm NM NM MM NM MO NO ON «M-OM OO NNN Mm MNN Nw «m MO NO mm «M NM Nm MN-MN MN MO MM N«N MN Om NN OM M« Mm NN «M «NTON NN OMN NN MOM MN NO MN MO m MNN NN OMN MN-MN .M: NN.N N..N WM: .3 .8 E .8 .MN 3 .MN .NN ..N OZO z OzO z OZO z on z on z OzO z acmugma NMMON Mcmugmm NMMoN pcmugwa NMNON ucmugma NMMoN pcmugma NmuoN ucmugwm NMMON M395 MMMMM=N . .MMMz NM .. N: .z MMM OMNNNMz MNNMMNNMO MNM 0:3 MmOMpcmugma new MMMNO MO< comm :N cmsoz Mo MNmnsaz NMNON mono MNMO OMNNNMz :mmn m>ms MOM .ON «NOON 106 Table 17. Reported Incidence of Polygyny: (A) Percentage of Husbands Currently Polygynous, (8) Percentage of the Fathers of All Husbands Who Were Polygynous (By Current Age of Husband), and (C) Percentages of Women Who Are Currently Married to a Polygynous Husband Current Approximate Percentage Polygamous A e Y Grgup 33:;5 L1 L2 M1 M2 Lushoto Moshi (1) (2) (3) (4) (5) (6) (7) (8) A. Husband 20-29 1944-53 12 14 11 2 13 6 30-39 1934-43 19 31 12 4 25 8 40-49 1924-33 35 51 27 12 43 19 50-59 1914-23 47 44 19 8 46 13 60+ pre-‘l4 38 47 36 42 42 39 8. Father (by Husband's Current Age 20-29 pre-‘20 7O 69 58 61 7O 60 30-39 pre-‘lO 44 71 71 69 57 70 40-49 pre-‘OO 61 68 54 76 7O 75 50-59 pre-‘9O 64 77 82 71 7O 75 60+ pre-‘80 78 92 85 82 85 84 Average for Fathers 63 75 7O 72 69 71 C. Women* 20-29 1944-53 26 29 11 O 28 6 30-39 1934-43 29 41 19 7 35 12 40-49 1924-33 39 4O 28 15 4O 21 50+ pre-‘24 50 54 33 16 52 23 *Women Currently Married, Married Once Only. These percentages are almost identical to percentages for all currently married women. Data on past incidence of polygynous marriages are not available for women currently widowed, divorced, or separated. 107 to 11 to 14 percent for those 20-29 in the other three areas, and among those 40-49, 35 percent in L1, 51 percent in L2 and 27 percent in M1 are polygynous.5 The proportions of women currently married to polygynous husbands show similar differentials. More than a quarter of the married women ages 20-29 in both Lushoto areas are married to polygynous husbands compared to 11 percent in M1 and none in M2. It appears that polygyny has all but disappeared in M2 and has become less common in M1 than in the past. No changes are detectable in the Lushoto areas. Regression Results at the Aggregate Level Tables 18 through 21 present regression results for both formulations of the basic supply model. Table 18 presents results at the aggregate level (all four areas combined) for the following five age groups: 20-49, 40-49, 20-39, 20-29 and 30-39. Three of the results given are for formulation (la) with WOMAGE and AGEGROUP; the other five are for formulation (lb) with YRSMAR. For all eight regressions, the signs of each of the dependent variables are in the expected directions, with the exception of the coefficients of DHPOLY for age group 20-29 (which are positive but low). That is, WOMAGE, YRSMAR, and NDIED are positively related to NBORN, and AGEMAR, DHPOLY, and BRSTFED are negatively related to NBORN. The t-ratios for WOMAGE, AGEMAR, YRSMAR, and NDIED are all significant at the .01 level. The constant is significant in six and BRSTFED in seven of the eight regressions at the .01 level. The largest R2 5With the exception of L1, polygyny is more frequent among men ages 40-49 than among those ages 50-59. 108 (.623) is for age group 20-39; the next largest R25 are for age group 20-49 (.574 and .572) although these are only slightly larger than those for age group 20-29 (.562 and .558). The relatively large R25 for age groups 20-39 and 20-49 are mainly because the larger sample sizes produce very large t-ratios for YRSMAR (or WOMAGE and AGEMAR) and NDIED. Examination of results for age groups 20-29, 30-39 and 40-49 indicates that the larger age groups (20-39 and 20-49) obscure some important differences in the effects of the independent vari- ables on fertility within the lO-year age groups. As expected, the substitution of YRSMAR for WOMAGE and AGEMAR has little effect on either R2 or the other three independent vari- ables. (Comparisons between these two formulations are given in Table 18 for age groups 20-49, 20-29, and 30-39). It does have the effect of increasing the size of the constant and greatly increasing its t-ratio. Although formulation 1b (with YRSMAR) is intuitively superior to formulation la in that it identifies the supreme impor- tance of total years married in explaining fertility (particularly among younger women), it has the disadvantage of obscuring the impor- tance of age of marriage as a determinant of fertility. The poorest performing regression is for women 40-49; R2 = .192. (For formulation la with WOMAGE and AGEMAR, R2 = .212). The signs are all in the expected directions, but only the constant and NDIED are significant at the .01 level; DHPOLY is significant at the .05 level. The coefficient for YRSMAR is nearly zero. Since most of these women have probably completed childbearing, it is to be expected that YRSMAR would be of relatively less importance in explaining fertility levels and differences than it is for younger 1(39 ..M>M. M.. MM M.MML-M MMMM.M.MM.MN .M>M. MM. MM M.MM.-. MMMM.M.MM.MN .MNM. .M. MM M.MM.-. MMMM.M.MM.M. .MMN.MN NMM..MN .NMM.MN NMM...N WMMN.NN WMN MM-MM MMM. MMM ..MM.M .NMM. - .MMN.- ..NN.M .NNM.N "MMMMM .MM..MN .MNN.MN .NMM.MN .NMM.MN .MMN.MN .NNM..N ..N MM-MM MMM. MMM .MMM.M «MMM. - .MNN.- .NMN. - .MMN.M NMMM.. ”MMMMM .MNM.NN .MMM.MN ..NM.MN NMM.M.N .MNM.MN ..N MN-MN MMM. MMM MMNM.M .MNM. - MMM.M .MNN.M .M.M.. "MMMMM .NMM.NN .NMM.MN .MMM..N WNM.N.N WMN.M.N .N.M.MN WMN MN-MN NMM. MMM .MNM.M MMNM. - MMM.M .MMN. - .MMM.M MMM.M "MMMMM .MN...N WMMN.MN .MM..MN NMM.MNN .M..N.N .MN MM-MN MNM. M.N MMMN.M .MNM. - M.NM. - .MMN.M «MMM.. ”MMMMM .M.N.MN .MNM..N .MNM.NN .NMM.MN .MM...N .MN MM-MM NM.. MMN .NMN.M .NM. - N.MM. - M.M M ..MMM.M ”MMMMM ..M.M.N .MMM.MN .NM..MN .MM.MNN NNM.M.N .MN MM-MN NNM. NMM MMMN.M .MNM. - .MMM. - .MMN.M .MMM.N ”MMMMM .MM.M.N WNMM.MN .MMM.MN ..N.N.N .MN.MNN .MM..MN .MN MM-MN MNM. NMM MMMN.M .MNM. - .MMM. - .MN.. - M..N.M .MMN.. ”MMMMM MMM zMMMz NM 2 MN.Mz MMNNMMNM .MMMMM MMzMN. MMzMMM MMMzMM game m .NM> MmNOMNNM> ucmucwqmch -NmucN ..MO MM-MM MMM .MN-MN .MM-MN .MM-MM .MM-MN MoMM .WMNMO muco .OMNNNMz MNNMMNNOON cwEoz .cgom :MLONNOO No cmnsaz No mucchELMNMO No Nmuoz LON MMNOMmm coNMMwNOMN Nm>wN MMMOMNOO< .mN oNOMN 110 women, although it was expected to be of some significance. The rela- tively low explanatory power of the mode1--and also the lack of importance of YRSMAR--are also probably due in large part to rela- tively greater inaccuracy in reporting age data--and perhaps to some extent, number of deceased children (NDIED) and hence NBORl--than among younger women. It is interesting, however, that coefficients for both DHPOLY and NDIED are rather large-- -.691 and O.752--with the former significant at the .05 level and the latter at the .01 level. The t-ratio for BRSTFED is almost significant at the .10 level.6 The lower half of Table 18 reports results for each of the two formulations for age groups 20-29 and 30-39. DHPOLY enters the equation quite differently for the two age groups. For ages 20-29, DHPOLY is of no importance; its coefficients are positive but quite low. However, for ages 30-39 it is negative and relatively large (just under -O.7) and significant at the .01 level. This implies that for women ages 30-39, marriage to a polygynous husband has the effect of reducing total number of children born by about 0.7 on average. Because of the apparently better data for younger women, subsequent analyses will be carried out only for women under age 40. The two estimation equations for age group 30-39 at the aggregate level are given below: 6It is possible that reported length of lactation of women in this age group is less representative of overall individual breastfeeding experiences as compared to the data for younger women. 111 (1a) NBORNij 1.864 + (.244)l~IOMAGEi‘j - (.207)AGEMAR1j - (.770)DHPOLYij - (.038)BRSTFEDij + (.686)NDIEDij (1b) NBORN]:j 2.872 + (.221)YRSMARij - (.753)DHP0LYij - (.037)BRSTFEDij + (.691)NDIEDij where: i - each currently married woman, married once only, in age group j j ages 30-39 Examples of the results produced by these models are given below: For women age 36, married at age 17, married to monogamous husbands, having breastfed for 24 months each, and with no deceased children, the estimated number of children born per women using model (1a) is 6.22, and for model (lb) it is 6.18. For women married to polygynous husbands with otherwise identical characteristics, estimated numbers of children born are 5.45 and 5.43 respectively. For women married to polygynous husbands and having two deceased children each but with other characteristics unchanged, estimated numbers of births per woman are 6.82 and 6.81 respectively. The two variables, DHPOLY and BRSTFED, are only proxies for what should presumably be (a) total number of married years during which the woman's husband was polygynous and (b) total number of months the woman breastfed all her children. No data are available on (a); (b) can be estimated in various ways, perhaps the simplest being to multiply the reported length of lactation (of the next-to- last baby weaned) by the woman's number of surviving children. Regressions were run with this variable (designated TBRSTFED). Regressions were also run with a variable designated YRSPOLY which 112 was DHPOLY multiplied by YRSMAR multiplied by a constant to represent the proportion of a woman's married years during which her husband was assumed to have been polygynous.7 These reformulations of the model was unsuccessful due to the very high levels of positive multicolinearity between estimated total number of months breastfed (TBRSTFED) and WOMAGE or YRSMAR; similarly between YRSPOLY and WOMAGE or YRSMAR. The simple correlation coeffi- cients for BRSTFED and the two age variables were around .5 and .6 and they were between .2 and .4 for YRSPOLY and the age variables; as a result, both TBRESTFED and YRSPOLY entered the equations with positive signs. Hence, DHPOLY and BRSTFED were retained in the basic models even though both are less than ideal proxies for the variables of interest. Because there is the possibility that systematic differences exist between Moshi and Lushoto areas which are not adequately incor- porated into the aggregate level equations, the basic models were expanded to include dummy (0,1) variables for each of the original variables, where each of the original independent variables is multiplied by either 0 or 1 depending on whether the case is in Moshi district or Lushoto district. The area dummy variables enter the equation for all cases from Lushoto (Lushoto values are all multiplied by 1 and re-entered in the equation a second time) but 7In the trial regressions, the constant used was .67. Of course, any value of l or less could be used. The drawback is that the same coefficient must be used for all wives of polygynous husbands because there are no data on length of time Spent by each woman in a polygynous union. 113 have a value of zero for all Moshi cases. Any systematic differences between the Lushoto and Moshi women will be indicated both by the area (L-) dummy variables and also by any differences in the original variables as compared to the variables in the original equation. Results for both formulations (la and lb) are given in Table 19 for age groups 20-29 and 30-39 and are compared to the results obtained with the basic (short) formulations from Table 18. The area (L-) dummies are given on the right side of the table. For all four comparisons, R2 is slightly higher for the expanded model (higher by a range of .006 to .021). For age group 20-29 the effect of the expanded model is to slightly reduce the size of the constant and to slightly increase the size of the coefficients for WOMAGE, AGEMAR, YRSMAR, and BRSTFED; none of the coefficients are changed by more than about 10 percent. For age group 20-29 the expanded model increases the coefficients for DHPOLY substantially (although t-ratios remain low) and reduces the coefficients for NDIED by about 25 percent. None of the area dummy variables are statistically significant and t-ratios are quite low except for LNDIED. For age group 30-39, coefficients for AGEMAR, YRSMAR, and NDIED are essentially unchanged. The coefficient for WOMAGE declines about 10 percent. Coefficients for BRSTFED are increased by a little over 10 percent and the constants are increased by almost 70 percent in the first formulation. The coefficients for DHPOLY are increased greatly--from under -O.8 to almost -l.6. This is also reflected in the coefficients for LDHPOLY (right side of the table) which are significant at the .01 level. LDHPOLY is the only area dummy 1'14 Nm>mN ON. NM oNumcuu ucMuNNNcOme Nm>mN mO. um ONumguu accuNchONMM Nm>mN NO. NM oNuoglu ucmuNNNcONmn .NMN.MN .MMM.MN .MMN.NN .NM..MN .M.M.MN .MMM.MN N.MN.MN .MMM.MN .MNM.MN .MNN.MN ..N MM-MM NMM. NMM.M . ..M.M .MNN.. MMM.M MMM.- .MMM.M .MMM.- .MMM..- .N.N.M .MNM.M "MMMMM .MMN.MN NMM..MN .NMM.MN NMM...N .MMN.NN .NN MM-MM MMM. ..MM.M MNM.M- .MMN.- ..NN.M .NNM.N ”MMMMM .N.N.MN .NNM.MN .NNM.NN .M.M.MN ..N...N .MMM..N .MMM.MN .M.N.MN .MMM.MN .MMM.NN .MMN.MN ..MM.NN .NN MM-MM .NM. MMM M ..M.M .NM... MMN.M MMN.M M.N M- ..MM.M «NMM.- .MNM..- .M.N.- .M.N.M MMN..M HMMMMM .MM..MN .MNN.MN .NMM.MN .NMM.MN WMMN.MN .NNM..N .MN MM-MM MMM. .MMM.M MMMM.- .MNN.- .NMN.- .MMN.M NMMM.. ”MMMMM .MMM..N .MMM.MN WMNN.MN ..MN..N .MM..MN .MMM.MN .MNM.MN N.NM.MN .MM.M.N N.MN.NN . .MN MN-MJ MMM. MMN.M MMM.M MMM.- MMM.- .MM.- .M.M.M .NNM.- NMM.M .MMN.M .NMM.. .MMMMM WMNM.NN .MMM.MN ..NM.MN .MM.M.N .MNM.MN . ..N MN-MN MMM. .MNM.M .MNM.- MMM.M .MNN.M .M.M.. .MMMMM .MNM..N .MMN.MN .MMN.MN .MMM.MN .MMM..N .MMM.MN ..MM.MN .MMM.MN .MMM.MN .M..M.N .MM...N ..M..MN . ..N MN-MN MMM. MMN.M MMM.M NMM.- MNM.M NMM.- MMM.M .M.M.M MMNM.- .MM.M ..NN.- .NMM.M M.. M .MMMMM .NMM.NN .NMM.MN .MMM..N .NM.N.N .MN.M.N .N.M.MN ..N MN-MN NMM. .MNM.M .MNM.- MMM.M .MMN.- .MMM.M MMM M ”MMMMM MMM ZMMMz MN.MZN MMNNMMMN .NMMMMN NMzMNNN MMzMMMN MMMMMMN MNMMMMN MN.Mz MMMNMMM NNMNMM MMzMMN MMzMMM MMMzM: M M ..N mu . 35> NNN MmNaMNNM> ucmvcwamncN -NmucN .OMO OM-OM ucM ON-ON Mmm< NMNMO mucO .umNNNMz MNNMMNNMON cmsoz Now .NON mNOMNN MNmuoz MNMMM cu OMNMMEOO .NouocmzN Low MMNEEOOlNN MmNOMNNM> MEEOO UNMONNNUNchNOooO No conaNucN "MONOMmm coNMMMNOmm Nm>oN muMOMNOO< .ON mNnMN ll5 variable that is statistically significant in any of the expanded models. It appears that there are some systematic differences between the two districts in the ways in which the independent variables relate to fertility (NBORN), but the results given in Table 19 suggest that for these two age groups the differences are relatively minor with the exception of DHPOLY for ages 30-39 and, to a lesser extent, NDIED for ages 20-29. The results indicate that for ages 20-29, NDIED has a greater effect on NBORN in Lushoto than it is in Moshi, although the negative relationship is relatively large and statistically significant in both districts. For ages 30-39, the only substantial difference between the two districts is in the importance of DHPOLY. The results indicate that polygyny is substan— tially more highly correlated with fertility in Moshi district than in Lushoto district, although again there is apparently a negative and statistically significant relationship between the two in both districts. There is essentially no difference between the two dis- tricts in the way in which NDIED enters the equations for age group 30-39. Regression Results at the District Level The basic supply models were also applied to data at the district and area levels. Results are given in Tables 20 and 2l respectively. The district level results provide another opportunity to look for systematic differences between the two districts in the relationships between the independent variables and fertility. ll6 Results in Table 20 are again given for both formulations la and lb and for age groups 20-29 and 30-39. The explanatory power of the model differs for the two dis- tricts. For age group 20-29 R2 is higher for Lushoto (.64 versus .52) while for ages 30-39 it is higher for Moshi (.49 versus .43). As expected based on the area-dummy results shown in Table l9, for ages 20-29 NDIED is of considerably greater importance in Lushoto than in Moshi (coefficients of about .80 compared to about .52). There are also larger differences between the two districts in the coefficients of the other independent variables than was evident from Table l9. In all cases the coefficients are larger for Moshi than for Lushoto. However, the directions of the signs are identical in both districts. (Although the coefficients for DHPOLY are positive, they are low and have low t-ratios for Lushoto). Comparisons between the districts for ages 30—39 show that the effect of NDIED differs little between the two; coefficients are .73 for Lushoto and .69 for Moshi. However, as expected, the role of polygyny is very different. The coefficients for Moshi are about -l.6 compared to about -0.4 for Lushoto. The coefficients for Lushoto are also smaller for BRSTFED (-.03l versus -.043--significant at the .01 level in Moshi but at only the .05 level in Lushoto) and for AGEMAR, but they are substantially larger for NOMAGE, producing a slightly larger coefficient for YRSMAR in Lushoto even though the coefficient for AGEMAR is smaller (-.l9) in Lushoto than in Moshi (-.22). l'l7 _m>m_ o_. so o_umc-o “and_a_co_mx .asap mo. um o_umt-u peauwc_=m_m‘ _m>m_ .o. an owuog-u acco_w_co_ma ona.qv Amm_.mv Asm~._v Aeo_.xv A~,~.¢v A“. om-oa owe. me. .mme.o spmo.- mmm.- .mmm.o .mko.m "caged obozmas Aomm.ev Ame_.~v AeOm.Fv Ame—.mv Amop.ov Aooo.ov Auv om-om «me. me, .omu.o e_mo.- NN¢.- .mm_.- .mmm.o mmc.- "ecaou oboxmas Aaom.ov Ammn.mv A_~N.¢V AoNq.mv Am_m.ov Auv om-om Fae. .NN .oao.o .meo.- .omm.P- .~_N.o .mmo.m ”tamed ~zmoz “Mkm.ov Amme.mv “mo“.qv Aaeo.av A~_N.mv Amoe.mv Auv om-om .me. .NN ._mo.o .Nqo.- .mam._- .m_m.- .m_~.o .o-.m ”ecoou _zmoz Ammo.~V Ammm.mv Ace—.ov A_m..Fv flamm.ev Auv ¢~-o~ new. mep ._om.o aw_o.- mmo.o .mmm.o .mme.p ”aemou owoxmss Aamm.NV Aeom.mv Am¢_.ov Aaqo.ov Aoma.mv Aoo_.Fv on oN-o~ eve. me. .mo~.o empo.- emo.o .we~.- .oNN.o _om.o "eeaou chozmas Aeoo.mv heow.mv “com.ov AoN.~_v A_mm.ov Auv mm-oN m_m. Now .m_m.o .Nmo.- Nmm.o .mmm.o .~m¢.P “eeaou ~xmo: Ao_o.mv Aom~.mv Amoa.ov Aoae.ov Aom.opv Aemp.ov Auv m~-o~ apm. NON .m_m.o .mmo.- _mm.o .PNN.- .Nmm.o m__.o "ecmou fixmoz 2¢Omz N« z ammo: owaemmm >soaxo aqzma> «(zqu m¢ manc_Lo> ucmvcwawvc~ -Lmucu .nwo mm-om we» mm-o~ mam< .AAFco coco .umwggcz x—ucmggzuv cw503 .cgom cva—_cu wo Lamazz vo mucoc_sgwumo we Paco: to» mu_=mmm cowmmmgomm Fm>w4 uuwgumwo .o~ v.2a» ll8 Regression Results at the Area Level Area level results are given,§n Table 2l for age groups 20-29 and 30-39 but only for formulation la (with WOMAGE and AGEMAR). The results show differences between the areas within each district which were not apparent in the district level results. For example, for ages 20-29, although NDIED has the expected positive sign for all four areas, it enters with quite different coefficients and t-ratios. In Ml, Ll and L2 it is significant at the .Ol level but it is not signi- ficant in M2. In Ml the coefficient is about 0.7 while it is only 0.2 in M2. In Ll it is about l.0 while it is 0.5 in L2. In the district level results, NDIED had a larger coefficient for Lushoto than for Moshi, but it is now clear that this was because for M2 the coefficient was very low and statistically insignificant. In fact, the coefficient for Ml is not all that much lower than the coefficient for Lushoto district (Table 20). There are also differences for BRSTFED; in L2 both the coefficients and the t-ratio are essentially zero while in both Ll and M2 the coefficients are -.032 and signifi- cant at the .05 level. In this example, the L2 cases cause BRSTFED to appear to be less important in Lushoto (for ages 20-29) than in Moshi (see Table 20) although the results in Table 2l show that the effect of BRSTFED on NBORN is greater in Ll than it is in Moshi district as a whole (Table 20). For ages 30-39 there are also differences for NDIED between the areas within each district, but they are much smaller than for ages 20-29. However, there are very large differences for DHPOLY. For Ml the coefficient is -l.l and for M2 it is -2.5. For Ll both coefficient and t-ratio are almost zero while for L2 the coefficient ll9 _m>m_ 0,. pm o_umc-u bemuwc_cmwmx _a>m_ we. be Osage-» acmuwcwcmwme Pm>m_ Po. be Osage-“ Headwa_cmwma Amma.mv Ammm._v “oem.~v Amep.mv AmN~.mV Ammm.ov Auv mm-om 0mm. MN .meo.o xeeo.- _m~.- .Nom.- .me.o mmm.o ”cameo N4 Aeoo.ev Aooo.ov Acme.ov Aomu.mv Ammo.mv “NmN.PV Auv mm-om moo. ON .me.o o.o Nep.o .wep.- .mom.o meo.m- ”Lemou .4 Aaom.ev “weo.mv Aeme.¢v “www.mv Aeoo.ev Amem._v Asv mm-om opm. ONP .omm.o .meo.- .Nom.m- .om~.- .me.o 04¢.N ”Lemou N: Aomm.¢v Ammm.wv Amwe.mv Aoma.mv Am_m.mv Am_s._v Auv mm-om mas. _o_ .eau.o ameo.- ammo.F- .Nmp.- .mmp.o \mwm.m ”Lemou .z Amee.mv “Noe.ov Am_o._v Ao_m.ov Ammm.kv Ammo.ov Apv m~-o~ New. mo .um.o $00.- NQN.- .m-.- .m_m.o wo..o ”cameo N4 Aomm.av A_mm.mv A_PN.FV Ammn.mv Am,¢.mv Ap_¢.Pv Auv mm-o~ New. «A .mpo.P .Nmo.- omN.o .NmN.- .Nmm.o _NN.F ”Lemou _4 Amfim.ov ANNo.mV Ammm.ov A_~m.NV Amo_.ov Auv om-om mmm. om cm~.o ammo.- pcmumcou .NmN.- «Nom.o me_.- ”tamed N2 Amm_.ev Aomo.Nv Ammw.ov Aom~.mv Ammo.NV Awmm.ov Auv mN-o~ mmm. _PF .m_e.o eomo.- omm.o «©e~.- .Now.o Ne¢.o "Laced P2 zaomz Na 2 omHoz omawmmm >soazo mezmo< wu mmFamem> pcmccmqmvcH -gmp:~ .uama mm-om wen m~-om mane .A»_=o muco ua_LLmz .umwgcmz appcmggauv :mEoz .cgom :anchu mo Luggaz we mucmcwEquwa mo Pmuoz Low wupzmmm cowmmmgmwm Pw>m4 mmg< .—N wpnmh l20 is .73 and almost significant at the .lO level. Despite these large differences, the overall impression produced by the district level results is not contradicted--i.e., in Moshi district polygyny is highly (negatively) correlated with fertility while in Lushoto poly- gyny is of only modest importance. For the two Moshi areas, the role of length of lactation for women 30-39 is very similar. Correlation of coefficients are about -.045 meaning, for example, that an increase of about ll months in length of lactation is associated with a decrease of about one-half a birth. The coefficient for L2 is nearly identical but in Ll the relationship between BRSTFED and NBORN is zero. Although for both age groups there are differences--in some cases rather sizable--in the coefficients for WOMAGE and AGEMAR, the relative magnitudes of the coefficients don't differ all that much. The constants differ greatly and are not significant (except for ages 30-39 in Ml at the .lO level). It might appear that these differences would produce substantially different estimates of NBORN. However, the example below shows that this is probably not the case. For a woman age 27, married at age 17, married to a monogamous husband, having breastfed 24 months, with one deceased child, the estimated NBORN using the equations given in Table 22 for ages 20-29 are as follows: Ml, 4.38; M2, 4.25; Ll, 4.59; and L2, 4.34. Retaining the same characteristics except assuming that the woman now has no deceased children produces the following estimates for NBORN; Ml, 3.66; M2, 4.03, Ll, 3.57; and L2, 3.83. All of these estimates seem highly plausible. l2l In summary, the model identified as a “supply" model for estimating number of children born to women of childbearing age in rural communities in Northeastern Tanzania apparently functions well. The hypothesized relationships between the various independent vari- ables and the dependent variable are in most cases supported, often with high degrees of statistical significance. In those cases at district and area level analyses and for certain age groups where the hypotheses about the relationships between polygyny and fertility, and the length of breastfeeding and fertility, are not supported, the basic model itself should probably not be rejected. Rather, the differences which result from the analyses using various age groups and geographical areas point to apparent diversities among communi- ties and societies and among age groups of women in the ways in which various of their characteristics are associated with their fertility. For some obvious groups, such as older illiterate women, misreporting of information on age and child mortality (and perhaps on length of lactation) also affect adversely the explanatory power of the model and the ways in which the independent variables enter the model. In the next chapter three additional supply models will be examined. The analysis will attempt to identify determinants of survival rates among women's children, age at marriage and length of lactation. Demand variables will be analyzed in Chapter VII. The utility of demand variables in explaining fertility will be compared with the utility of the supply model analyzed in this chapter. Finally, in Chapter VII discriminant analysis will also be used to analyze relationships between (primarily) demand characteristics and whether or not women and husbands want more children. CHAPTER VI ANALYSIS OF DETERMINANTS OF SURVIVAL RATE, MARRIAGE AGE, AND LENGTH OF BREASTFEEDING The determinants of three additional supply-related variables will be analyzed in this chapter. They are (l) survival rate (or percent surviving) among women's children, (2) age at marriage of women, and (3) length of lactation of women. Multiple regression models are first specified for each of the three independent vari- ables. Data are then given on reported number of surviving children and implied survival rates, and characteristics of many of the other relevant variables are summarized. The results of the regression models are then described and analyzed. Specification of the Models It is hypothesized that number of children surviving will be positively related to quality of health care, positively related to woman's years of formal education, and positively related to income. A model for the determinants of percent surviving among a woman's children is specified below: (2) PCNTSRVij = f (DLVRYij, NOMEDUCij, HUSB-NAGEij, CROPVALij, BLDQUALij, e) 122 where: PCNTSRVij DLVRYij WOMEDUij HUSB-WAGE.. lJ CROPVAL.. lJ BLDQUALij J l23 the percent of the children born alive to woman i (in age group j) who are still alive a dummy variable whose presence indicates that the last baby of woman i was born in a hospi- tal (as a proxy for quality of health care received by the woman and her children) the number of years of formal schooling of woman i (as a proxy for quality of health care and nutrition provided by the woman herself) a dummy variable whose presence indicates that the husband of woman i has a wage-paying job (as a proxy for income and other influences favoring relatively better health care) the estimated value of crops produced by the household of woman i during the l2 months preceding the survey (as a proxy for current income) an index of the quality and value of the housing of the household of woman i (as a proxy for permanent income or wealth) random error each currently married woman, married once only, in age group j woman i's age group It is hypothesized that the age at which a woman first marries will be positively related to the level of education of her- self and her husband, will tend to be higher if the woman is Christian (rather than Moslem or Traditionalist), higher if her husband is Christian, lower if her husband is a polygynist, higher if her husband has a wage-paying job, and negatively associated with the level of her household's (i.e., husband's) wealth. The model is specified below: = .., .., .., H E .. (3) AGEMARij f (NOMEDUCIJ HUSBEDUC1J NOMRELCIJ USBR LCU l24 DHPOLYij, HUSB-WAGEij, BLDQUALij, e) where: AGEMARij WOMEDUCij HUSBEDUCij NOMRELCij HUSBRELC.. lJ DHPOLYij HUSB-NAGEij BLDQUALij The direction the age of marriage of woman i in age group j the years of formal schooling of woman i years of formal schooling of woman i's husband a dummy variable whose presence indicates the religion of woman i is Christian a dummy variable whose presence indicates that the religion of woman i's husband is Christian a dummy variable whose presence indicates that the husband of woman i is currently polygynous a dummy variable whose presence indicates that the husband of woman i has a wage-paying job an index of the quality and value of the housing of the household of woman i random error each currently married woman, married once only, in age group j of expected relationship between each indepen- dent variable and the dependent variable (AGEMAR) is indicated in the parentheses below: NOMEDUC (+), HUSBEDUC (+), NOMRELC (+), HUSBRELC (+), DHPOLY (-). HUSB-WAGE (+), and BLDQUAL (-). It is hypothesized that the length of time a woman breastfeeds will be positively related to her age, negatively related to her level of education, longer if she is married to a polygynous husband, shorter if her religion is Christian, shorter if her husband has a wage-paying job, and negatively related to the income of her house- hold. The model is specified below: l25 = .. .., NOMREL .., (4) BRSTFEDij f (WOMAGEij, NOMEDUC1J, DHPOLY1J ClJ HUSB-WAGEij, BLDQUALij, e) the number of months woman i reported having breastfed the next-to-last baby she weaned (or the last baby weaned if she weaned only where: BRSTFEDi‘j one) NOMAGEi. = the reported age of woman i at the time of the 3 survey NOMEDUCiJ = number of years of formal schooling of woman i DHPOLYi. = a dummy variable whose presence indicates that J the husband of woman i is currently polygynous WOMRELCij = a dummy variable whose presence indicates that the religion of woman i is Christian HUSB-WAGEij a dummy variable whose presence indicates that the husband of woman i has a wage-paying job BLDQUALi. an index of the quality and value of the J housing of the household of woman i e = random error i = each currently married woman, married once only, in each group j The direction of the expected relationship between each inde- pendent variable and the dependent variable (BRSTFED) is indicated in the parentheses, as follows: WOMAGE (+), HOMEDUC (-), DHPOLY (+), NOMRELC (-), HUSB-WAGE (-), and BLDQUAL (-). Descriptions of Some of the Variables Variables of interest in this chapter are the dependent and independent variables included in models (2), (3), and (4). Table 22 gives the reported number of surviving children and the implied percentages surviving by age group of the women. The percentages of women whose last baby was born in a hospital are given in Table 13 l26 in Chapter IV. Data on age at marriage, length of lactation and incidence of polygyny are given in Tables l5 and l7 in Chapter V. Data on housing, values of crops produced, occupations and education are given in Tables 3, 4, 6, 8, l0, and ll in Chapter IV. Data on religion are given in Table l in Chapter III. There appear to be no striking differences among the areas in the average numbers of children surviving per woman (top of Table 22). The middle block in Table 22 shows that when averaged over all_women, the average number surviving per woman is much lower in M2 than in the other three areas for women l5-l9 and 20-24. This probably results mostly from later age at marriage--at least in recent years--in M2 (as noted in Chapter V). The lower portion of Table 22 gives implied percentages surviving by age group. There are some notable differ- ences among the areas which indicate probable under-reporting of number of deceased, especially among the older women in the Lushoto areas.1 From ages 40—49 to ages 50 and over, implied survival rates decline less in the Lushoto areas than in the Moshi areas. Survival 1The patterns in the data could result from other types of errors besides under-reporting of deceased, but the latter seems most likely and is to be expected based on experiences of other demographic surveys among illiterate populations, especially in Africa. One would expect that in the interview situation even illiterate women would report surviving children with a high degree of accuracy, but that older women may be more inclined to forget or neglect to mention long- deceased children, particularly those who died soon after birth. In some cases memory may simply be faulty; in others it may be due to reluctance to mention a grief-stricken event due to the belief that one might experience some additional misfortune as a consequence of discussing it. Regardless of the cause of the under-reporting of deceased, the result will be an undercount of number born (since number born is the sum of number surviving and number deceased) and an inflated survival rate, with obviously deleterious consequences for both models (1) and (2). ill! III I ‘ In. 'I I I'll! I l I II I I'll. I: III. l! I Ill-I l27 Lug econ cucu__zu Lassa: mmmgm>o .mpnmu mwzu yo ugmq swan: .Ae_ m_nmh Eong :meoz men mount Pm>w>gam o» uxmc mwmmgucmcma cp mmzpm> . mg» new mp mpamp seem cm>wgmv mcm mmumm Fu>w>gam - =o_ua_>ao venueaum u om. "mayo: AF.mV mum. A_.~V omn. A~.ev mmm. Ao.mv com. Am.~V as“. Aa.ov Fm~.. +om Ao.~v new. Am.ov omm. Am.kv aka. Am.mv mom. Ao.Nv omm. An.ov Nkm.. me-oe AN.mv 8mm. Ap.ov cmm. Am.ov mam. AN.QV mom. Ao.ov mmm. A~.ov «mm. mm-mm Am.¢v com. Am.mv mum. Am.¢ mmm. A~.¢V omm. Am.mv Nam. A..mv omm. em-om Ao.mv 8mm. Am.mv oma. Ak.m mom. A¢.mv com. Ae.mv new. Am.mw mom. m~-m~ Am..v mam. Aa.,v mmm. Am.Pv moa. Am.,v “mm. AQ.FV Nma. A_.~ emm. ¢~-om Axpco muco umwggmz .Ungmz hpucmggzu mm< :wEoz .._.o cmgupwzu mcoE_.>.5m umwpqEH Tease: Amo.Fv m..e Amm.Pv mm.e Aea._v No.4 Aem..v QN.¢ Amm.Pv em.e Amm._v ¢N.e em-om Ao¢.PV mo.m Amw.Fv mm.~ Am¢.Pv 5,.m Aum.Pv _o.m Amo.~v om.~ Lam._v mw.~ a~-m~ AMN.PV mm._ AFm._V m_.. Aom._ m¢.P Am.._v NN.~ Amo._w mk.o Ace._ m¢.P ¢~-o~ Amo.ov mF.o Amm.ov VF.o Amm.o mp.o Aom.ov mN.o Amm.o co.o AFN.o ¢~.o mP-mF :mEo: 44< .cmeoz gm; cmgcp_;u mcw>w>gsm we LmnE:z mmmgm>< A~.NV o.m Am.PV ¢.m A¢.Nv ~.m Ao.~v m.e Ao.~v o.m Am._v P.m +om Am.Nv F.o A¢.NV w.m Am.~v 8.8 AO.NV o.m Ae.Nv m.m Am.Nv m.m m¢-o¢ Am.PV o.m Am._v _.m Am.mv o.m Ao.pv o.m Ao.Nv _.m Am.pv F.m mm-mm Am._v m.e Am.Fv “.4 Ao._v m.¢ Am._v «.4 Am._v m.e Am._v ¢.¢ em-om A¢.Pv F.m Ao._v N.m Am._v m.m Am._v P.m Am.Pv N.m Am.Pv ~.m ¢~-m~ A_._v N._ Am.mv m._ A~.Fv m._ AF._V N.F A_..V m.F Ae.PV m._ ¢~-o~ A>_:o muco umwggoz .nmwggoz x—ucmcgzuv case: Lwa cmgupvzu mcw>w>gam mo Laneaz mmmgm>< Am_v ANFV Ap_v Aopv Amy Amy ARV Amy Amy Aev Amv ANV A_v Aomv .m>< Aomv .m>< Aomv .o>< “omv .o>< Aemv .m>< Aaomv .m>< a o a a L ouosmzs Pgmoz N4 _4 N: _z am< . asogo mm< acmggzu m.:msoz x5 uwppogucou .mmumm Fm>w>gam um_PaEH new Amcowumw>mo ugoccmum ucmv cmgvp_gu mcw>w>gzm mo mgw5532 ommgm>< .NN m4mop op. oo o_ooL-o pooUFcooowm‘ Po>o_ mo. om o_ooc-o oooowceoo_mo .osop _o. oo o.oot-o ooouwcwoo_m. Ammo.ov Ammo.oo flooo..v Hoop.ov Aopo.~v Am~.opo Ago om-o~ ooo. opp oop.o ooo.o- o_o.o- oo~.o coo.oo .o,.~o "ccooo No ooo.oo Apoo.oo Rumo.ov Amoo.Pv Amuo..v Aoo.~mv A“. om-o~ omo. “NF oho.o No~.m-. mm,.P -~.P- ‘oo,.o .oo.oo "ccooo _o - Awoo.oo Aooo.oo Aooo.ov “koo.ov Akmo._v Ao_.-o Aoo oo-o~ o.o. o~_ ooo. - o.o .No. - ooo. - Bop.o «om.oo "ccooo N: AMNN.PV “moo.ov Amko.ov Ammm.oo Ao-.mV Aoo.opv on oo-o~ moo. oop ooN. - NAP. - oao.o one. - .oo.m_ .No.o~ ”ccaoo _= Aoom.oo Aoom.,v Aoo~.Pv Amoo.ov “oom.~V Ao~.~mv on om-o~ moo. ooN ooo.o ooo.o- oo_.~- ooo. - .o-.o .oo.oo “ccaoo opoxmoo AomF.PV Aeso.oo A~o~.ov Aop_._o Aomo.oo Ao,.ooo on om-o~ Foo. oom o~_. - Now. - Pom.o ooo. - .ono.o .oo.oo ”Lcooo Homo: Aooo.ov Appo.oo A~o~.ov “moo._v Aoo~.mv Aom.oov Ago oo-o~ moo. ooo ooo. - ooo..- F~_. - moN.o- .ooo.~ ao~.oo "ccaoo ooo oompzoa No 2 Aooomzm "moomo oo<3-omoz oo>oo uomu n .su> mmpnowgm> acmucmamucn -LoucH .amo mcw>w>cam ucwucmo mo mucmc_eLopmc so» mu_:mu¢ :ovmmmsuum .mw a—nuh l32 three or more surviving children. As might be expected, the model performed better but only slightly so. R2 did not exceed about .120 (which was in M1). Three reasons for the poor performance of the model are noted below: (1) There are relatively high levels of multicolinearity among some of the variables, particularly the income variables. (2) The hypothesized relationships between some of the independent variables and.the dependent variable (PCNTSRV) are reversed; even the simple correlation coefficients are reversed. For example, the correlation coefficients between education and percent surviving are almost invariably negative, even among younger women. (3) Clearly, most of the determination of survivorship (or conversely, mortality) is excluded from the model. "Luck" is probably the main reason some children die and others survive and the major cause of differential child mortality among families in these communities. The model was reformulated in an attempt to eliminate some of the variables showing high levels of multicolinearity. It was reduced to the three independent variables of DLVRY, WOMEDUC, and one of the three income variables (HUSB-NAGE, CROPVAL, and BLDQUAL). The reduced model with BLDQUAL as the income proxy performed better than models with each of the other two income variables, but the performance of this reduced model was not superior to that of the original version. Of course the relatively high levels of multicolinearity between DLVRY and BLDQUAL, and DLVRY and HOMEDUC (both correlations positive in most cases), remained. Despite the poor performance of the various versions of this model, one interesting result is that in all formulations, DLVRY was l33 positively associated with PCNTSRV and in almost all runs DLVRY was statistically significant--in most cases at the .01 level. In the abbreviated versions of the model (with only one income variable), the coefficient for DLVRY was usually about 10 or above implying an increase of about 10 points in the percent surviving associated with children of women who had hospital deliveries. It is interesting to note that for women under age 40, the correlation coefficients between DLVRY and NOMEDUC, DLVRY and 'BLDQUAL, and DLVRY and HUSB-NAGE, respectively, are invariably positive. For the Moshi areas the correlation between DLVRY and .NOMEDUC, and DLVRY and BLDQUAL, are slightly above .2; there is a greater range-—from below .1 to above .2 in the Lushoto areas. The correlations between DLVRY and HUSB-HAGE are positive but below .1 in the Moshi areas and L1, but above .2 in L2. However, the correlation coefficients between DLVRY and CROPVAL are negative in the Lushoto areas (ranging up to -.2 in L2) and positive but low in the Moshi areas (ranging up to about .1). Regression Results for Age at Marriage Table 24 gives regression results for the model of determi- nants of age at marriage. Again, to minimize the effects of age misreporting among older women (generally overestimating both current age and age at marriage) and to minimize the downward bias among younger ages due to exclusion of those not yet married (e.g., 66 percent of the women ages 15-19 and 25 percent ages 20-24 in M2), regressions are reported for age group 25-34 (except at the aggregate level where results are also given for age group 25-39). The R25 are I34 Fm>wp op. no owuog-u ucouwwwcmwmx _o>o_ mo. we oooee-o ooeooc_eo_mo Fm>m~ Fo. no o_uoc-u ucouwwwcmwma AmoN.ov Aooo.oo Aom_._v Aooo.oo AoNo.ov Aooo.,v ANNP.ov Amo.NPv on om-oN o__. No ooo. - mom.o Noo.m oom.p- ooo.o mom. - Foo. - .oo.oo ”ccooo No AoN,.,V Aoo_._v “moo.Pv AooF.Fv Aooo.ov A_oo._o flooo.Nv Amo.N_v Aoo om-mN mo.. om _om. - oop._- oo_.o omo.o- Npo.o \Noo.o NNoN.o .FN.mF "ecooo _o AoN_.NV ANo_.oo Aoom.ov Aooo.oo Aooo.ov A_mo._v Aooo.o v Aoo om-mN ooo. oo Noon. - omp.o ooo.o- ooeomooo ooo.o P_P.o NooN.o .oo.NN Hoodoo N2 Amoo.ov ANNo._v Aooo.Fv homo.ov ANoo.ov Aooo.ov Aooo.oo AN_o.o v now om-mN ooo. NN om_. - moo.o oN_.N oNo._ Noo.o oop.o ooo. - eNo.m_ ”ccooo _z Loom.ov AmNm.ov Aomm.~v Aooo.ov ANmF.FV Aomo.oo Aooo._v Aom.N_v on oo-oN oNo. oo. mop. - oom. - omo.o oNo.,- ooo.o omp.o omN.o .om.o_ “ecooo oeoxmoo AoNo.Nv Aooo.Pv A_oN.ov Aooo.ov A_om.ov ANoo.ov A_NN.ov Aoom.o o Roy om-oN Nmo. No. NoNN. - Nmm.o mom. - ooo.F opo.o Noo.o oNo.o .No.NF Hoodoo ozmoz Amoo._o ANoo.Fv AmoN.ov AmMN.oV A__o.ov AmmN.ov Aooo._v Amo.va on om-mN omo. oNo \omp. - Nooo.o ooN.o NoN.o oop.o mNo. - oop.o ._N.NF Hoodoo ooo Noeoa “ooo._o ANNo.oV AN_N.oo Aooo.ov Amo_._v A.NN.oV AoNo._v ANm.NNV Ago om-oN -moz omo. NNN NNoN. - moN.o oNo.o omm.o ooo.o moo.o o_F.o .mm.o_ "ccooo oo< No< a z oqooooo oo<3-omox ooNoomo: oooozoz >oooxo ooooomo: oooozo: N unmu n .eo> mmpnmwco> ucovcmomucfi -LoucH .omo wmmwccoz um mm< m.:osoz mo mucocwsemumo Low mu_:mmm cowmmoemmz mpaoh 135 low, ranging from .037 in Moshi district as a whole to .185 in L1. For individual areas, the R25 are .064 in M1, .090 in M2, .185 in L1, and .110 in L2. In all cases the constant is significant at the .01 level. None of the independent variables are significant at the .0l level although there are some instances of significance at either the .05 level or the .10 level. The signs of the independent variables are of interest, however. At the aggregate and district levels, woman's level of education is positively related to age at marriage--as hypothesized, and NOMEDUC is almost significant at the aggregate level. The coefficients are not large; they are about .12 at the aggregate level (implying that women who complete primary school-- grade 7--marry about a year later than women who don't attend school, ceteris paribus). There are, however, notable differences among the areas. In Ll the coefficient is relatively large-~0.707--(implying that primary school leavers marry about 5 years later than those with no schooling) and significant at the .05 level. The coefficient for M2 is .29 (e.g., primary school leavers marry about 2 years later than those without schooling) and significant at the .10 level. The coefficients for M1 and L2 are negative but very small. If anything, the data probably understate the positive effect of woman's education on age at marriage. Presumably, women with education are more likely to report their ages accurately than women without education. As mentioned earlier, most analyses of age misreporting have indicated that respondents tend to over- rather than under-estimate their ages. If this is more prevalent for uneducated than for educated women--which is believed to be the case, 136 it would tend to bias upward not only current age but also age at marriage for uneducated relative to educated women thus reducing the overall strength of the positive relationship between level of educa- tional attainment and age at marriage. Husband's educational attainment is also usually positively related to woman's age at marriage in the results given in Table 24, and it is significant at the .10 level (with a coefficient of .542) in L1 although it has a negative coefficient (-.353) in L2. Polygyny is also positively associated with woman's age at marriage although both coefficients and t-ratios are not large. The effect of religion is mixed. In both Lushoto areas Christianity among women is associ- ated with earlier age at marriage while among their husbands it is associated with later age (of the women) at marriage. Both are positively related to age of marriage in M1. In M2, all women in this age group are Christians, and although the sign is negative for men, there are very few non-Christian men. In the Moshi areas and L2 the coefficients for HUSB-wAGE are positive; the coefficient is negative in L1. The index of building quality is interesting. In all cases the coefficient is negative--as expected--implying that woman's age of marriage is lower the greater is her husband's wealth, and it is significant at the .10 level for the aggregate-level and for Moshi and M2. Although the R25 are relatively small and very few of the t-ratios are significant, with few exceptions the signs of the coefficients in the equations for woman's age at marriage are in the expected direction; the only consistent exception is polygyny which 137 has a positive sign while a negative relationship was hypothesized. The consistency in results for the wealth proxy (BLDQUAL) and for woman's education (especially given the probable offsetting bias) lends some support to the hypothesized relationships despite the absence in some of the regressions of statistical significance. Regression Results for Duration of Breastfeeding Table 25 gives regression results for the model of determi- nants of length of lactation. Regressions are shown at the aggregate level for age groups 25-34 and 25-39; the age group is 25-34 for Lushoto district and the Lushoto areas, and the age group is 25-39 for Moshi district and the Moshi areas. Lushoto data are only for women up to age 35 because, as discussed earlier, data for women 35-39 are presumed to be less accurate for the Lushoto areas than for the Moshi areas. The R25 are not large but are above those obtained from the regressions on age at marriage. For the areas they range from .091 in L2 to .297 in L1. At the aggregate and district levels they range from about .11 to .15. With the exception of the aggregate level, the intercepts are not significant. The results for the dependent variables are interesting, although again-~with the exception of woman's age--they are in most cases not statistically significant. Woman's age is in all cases positively related to reported length of lactation and it is significant at the .01 level in M1 and M2 and at the .10 level in Ll. In these three areas the coefficients range from about .6 to .8 implying that an increase of one year in age is 1138 .Fo>o_ op. oo oeooe-o oooo_cwoowmx _o>o_ mo. oo oeooe-o ooeo_coooemo _o>o_ _o. oo oeoee-o oooo_e_eo_mo Aomm.ov Aomo.Fv Aon.oo Aoom.ov ANmN.ov AooN.oV A__o._v Ago om-oN _oo. No on.o \NNN.m- NNN.P- Noo.F oom.o ooo.o oo.NN "ecooo No Aooo.Nv Ao__.ov Ao_o.Pv Room.Fo ANoo.Nv Aooo._v Aomo.ov on om-oN NoN. om .moN.m NNN. - NoMN.o omo.m emmo.P- Nooo.o _oo.m ”ccooo _o Amoo.ov Aomo.ov AmNN._v Aomo.ov AoNo.Nv Ammo.ov on om-oN ooo. mm_ NNN.o Noo.P- ooeomeoo ooo.o- mNo. - .Nom.o ooo.m "ccooo N: ANoP._V Aooo.ov Aooo.,o ANoN.oV Ammo.ov Ammo.ov A_oN.ov Aoo om-mN oo_. No_ oPo.o Noo.P- omN.o- ooo.o NoN. - .ooo.o ooo.o Hoodoo _z Aooo._v AFoN.Nv AoNo.ov A_No.Fv Aom_.oo AooN.Fv AoNo.ov on oo-oN oop. ooo ooo._ NNmo.o- oPN.,- Nom.m _oo. - ‘moo.o Noo.o "ccooo oeozmoo AooP.Fv Aooo._v Aoo_._o A_o_.ov Amoo.ov Aooo.ov “Nom._v on om-oN o.o. ooN omm.o ooo._- oom.o- mom. - oNN. - .Noo.o oN.o_ "ccooo Homo: AmoN.oo ANmo.Nv Aooo.mv Ao_o.ov Aooo.ov Aooo.ov ANop.ov on om-oN oo_. on oNN.o NNoN.N- .o_o.o- omm._ o_P. - .Noo.o .oo.mp Hoodoo ooo owe -»mooaxo oooozoz Noozoz 38 n ...; $23.»; 23:8qu L35 d8 mcwvmmmummmLm um cowumgzo AuwuLonmv 9:02.03 we mucmcpELmumo Low mu pzmmm cowmmmmem .mN m—nmh 139 associated with an increase of a little under a month in reported length of lactation. In all runs except L2, education is negatively associated-~as expected--with reported length of lactation, although it is signifi- cant (at the .05 level) only in L1. As expected, polygyny is posi- tively associated with length of lactation (except in M2 where there is very little polygyny). Women who are Christians are likely to have breastfed for shorter periods than non-Christians in M1 and L2 but-- for unknown reasons--by about half a year longer in L1 (significant at the .10 level). (All women ages 25-39 in M2 are Christians.) As hypothesized, wives of husbands who had wage-paying jobs (at the time of the survey) breastfed for shorter periods than other women in all four areas (although both coefficients and t-ratio are very low in L1); the coefficient in L2 is -5.7 (months) and the t-ratio is signi- ficant at the .10 level. Contrary to expectation, BLDQUAL is uniformly positively related to length of lactation. In Ll the coefficient is 3.8 and it is significant at the .01 level. The original hypothesis was that level of permanent income--for which BLDQUAL was used as a proxy--would be negatively related to duration of breastfeeding based on the assumption that women with higher incomes would have been more likely to purchase breast-milk substi- tutes and wean earlier. Comments on the Effects of Woman's Education The amount of formal schooling of a woman was hypothesized to be associated with all three dependent variables analyzed in this chapter--survival rate, marriage age, and length of lactation. 140 Although in most cases the coefficients and t-ratios for HOMEDUC are low, some of the results are of interest. Woman's level of education was hypothesized to be positively related to the survival rate among her children and her age at marriage and negatively related to length of lactation. Although the coefficients are, with few exceptions, low and not statistically significant, the direction of the signs are also--with few exceptions-- consistent at all levels of analysis. Contrary to what was expected, the signs were negative for PCNTSRV implying that the children of educated women apparently do ggt_have a higher survival rate than do children of uneducated women. A study by Lindner [1975] of health and nutrition practices in the Kilimanjaro area similarly found that infant and child mortality rates did not decline as mother's educa- tional level increased. This was apparently because educated women did not provide much better health care or more hygenic home situ- ations while they were more likely to wean their babies earlier and to bottle-feed them thereby (frequently) adversely affecting their nutritional status. In M2 and L1, woman's level of education is positively related to her marriage age (significant at the .10 and .05 levels respec- tively). In the other two areas the relationships are essentially zero. Since the proportions of girls who attend school have been increasing fairly rapidly in recent years, this is presumably having (and will continue to have) the effect of raising the average age at marriage--at least in L1 and M2. In fact, in the last 10 years or so, age at marriage has apparently already risen substantially in M2, which is consistent with the large increase in the proportion of 141 girls in M2 attending school during the past couple decades. The pro- portion of young women with some formal education in L1 are still low but can be expected to increase markedly during the next lO-15 years. The age at marriage equation for L1 indicates that this should result in a fairly large increase in average age at marriage in L1, although I suspect the NOMEDUC coefficient for L1 overstates what the likely effect on age at marriage will be when a very high proportion (e.g., over 80 percent) of the women receive some formal schooling. For all areas except L2, WOMEDUC is negatively related to reported length of lactation, although it is significant (at the .05 level) only in Ll. To the extent that the negative relationship between NOMEDUC and BRSTFED and the positive relationship between NOMEDUC and AGEMAR are valid in L1, they will have the opposite effects on fertility although they will not necessarily by fully off- setting. Education, by increasing age at marriage, reduces fertility at any given point during the woman's childbearing ages, ceteris paribus. And as indicated by some of the coefficients produced by the supply model analyzed in Chapter V, the effect could be substan- tial. For example, the implications of the various equations given in Tables 21, 24, and 25 are that an increase in average educational attainment of about 6 years for a group of girls in L1 would be as follows: Age of marriage would increase about 4 years (based on results in Table 24). This would imply a reduction of about 1 birth at any given age while these young women are in age group 20-29 (based on results in Table 21). On the other hand, an increase of 6 years of education would imply a reduction of about 10 months in 142 in breastfeeding (from Table 25) which would in turn imply an increase of about 0.3 births during ages 20-29 (Table 21). In this particular example, the net effect of an increase of 6 years of education on the number of children born to women ages 20-29 in L1 would be a reduction of about 0.7 births per woman. In summary, the three models analyzed in this chapter do not explain very much of the variations in survival rates among children within families, age at marriage of women, and the reported length of lactation. Although the t-ratios and the coefficients for the independent variables are in most cases low and not statistically significant, in most cases the relationships between the independent variables and the various dependent variables are in the directions hypothesized, but due to the lack of statistical significance, most of the hypotheses about relationships between the variables--although not contradicted--are not statistically supported. CHAPTER VII ANALYSIS OF DEMAND FOR CHILDREN Two different techniques will be used to analyze the effects of demand variables on number of children born. The first will be multiple regressions, for which the independent variables will be proxies for demand factors. Results will be compared with those obtained in Chapter V for supply models. The second technique will be the use of discriminant analysis to analyze relationships between various characteristics of women and their husbands and their reSpec- tive responses to the question, "do you want more children?" Only those who answered either "yes" or "no" will be included in the analysis (about 95 percent of all respondents). Specifications of the Basic Models One of the regression models is specified below. Alternative formulations of the model will include various combinations of these independent variables. Regressions are also run substituting for CROPVAL (a proxy for current income) its natural log--(LN)CROPVAL-- to determine if this nonlinear function has greater explanatory power than the linear form. (5) NBORNij = f (CROPVALij, BLDQUALij. HUSB-NAGEij. WOMADVLRij. NOMRELCij. NOMEDUCij. e) 143 where: NBORNij CROPVALij BLDQUAL]:j HUSB-NAGE.. lJ WOMADVLRij NOMRELC.. lJ NOMEDUCi‘j 144 the number of children born alive to woman i in age group j the estimated value of cr0ps produced by the household of woman i during the 12 months preceding the survey (as a proxy for current income) an index of the quality and value of housing of the household of woman i (a proxy for per- manent income or wealth) a dummy variable whose presence indicates that the husband of woman i has a wage-paying job (a proxy assumed to indicate that the price of children yis-a-vis other goods is rela- tively high) a dummy variable whose presence indicates that woman i thinks there are advantages to having a large family (a proxy for the price of children being relatively low) a dummy variable whose presence indicates that the religion of woman i is Christian (a proxy for preferences being relatively in favor of other goods vis-a-vis children--relative to preferences for non-Christians) Number of years of formal schooling for woman i (a proxy for tastes as well as for the rela- tive price of children) random error each currently married woman, married once only, in age group j woman i's age group The directions of expected relationships between each indepen- dent variable and the dependent variable (NBORN) are indicated in the 1 This is on the further assumption that the children of men with wage-paying jobs have relatively less opportunity to contribute to household income through their work on the household's agricul- tural holdings. 145 parentheses as follows: CROPVAL (+), BLDQUAL (+), HUSB-NAGE (-), NOMADVLR (+), NOMRELC (-), and NOMEDUC (-). In discriminant analysis, one attempts to select independent variables which measure characteristics on which the two groups are expected to differ. In this case the two groups are those who responded "no" and those who responded "yes" (to: "do you want more children?"). Discriminant analysis is a method for creating linear combinations of the independent variables which attempt to maximize the separation of the two groups. The discriminant functions are of the following form: D. = di 1 Z + d1 Z + . . . + d. Z l l 2 2 1p p where Di is the score on the discriminant function i, the d’s are weighting coefficients, and the Z's are the standardized values of the p discriminating variables used in the analysis [Klecka, 1975]. These dij's are called standardized discriminant function coefficients and are analogous to standardized beta coefficients in multiple regression. Each coefficient represents the relative contribution of that variable to the function. The sign of the coefficient indicates whether the variable is making a positive (toward the "yes” group in this case) or a negative contribution to the function [Klecka, 1975]. Several measures (to be discussed later) are avail- able to indicate the extent to which the two groups are differentiated by the set of independent variables. Specification of the basic discriminant model for women (currently married, married once only) is given below. A comparable model will be tested for husbands. 146 7 = - '0 no, (6) MORE.ij f (CROPVALij. BLDQUALij, NOM NAGEIJ, NOMADVLR13 where: NOMADVSMij. WOMEDUCiJ. WOMRELCij. PCNTSRVij, OPMORTDN.., NSRVNG ., NOMAGE.., AGEMAR. ) 13 13 1J 13 7 MORE.ij CROPVALij BLDQUALij NOM-NAGE.. lJ NOMADVLR.. lJ NOMADVSM.. lJ NOMEDUCij WOMRELCij PCNTSRVij OPMORTDNij a dichotomous (0,1) variable indicating a response of either "no" or "yes“ to the ques- tion, "do you want more children,“ for woman i with j surviving children estimated value of cr0ps produced (shillings: 00) by the household of woman i during the 12 months preceding the survey (as a proxy for current income) an index of the quality and value of housing of the household of woman i (as a proxy for per- manent income or wealth) a dummy variable whose presence indicates that woman i has a wage-paying job (as a proxy for opportunity costs of the woman's time and hence, a proxy for price) a dummy variable whose presence indicates that woman i thinks there are advantages to having a large family (as a proxy for the relative price of children) a dummy variable whose presence indicates that woman i thinks there are advantages to having a small family (proxy for relative price of children) number of years of formal schooling of woman i (proxy for preferences) a dummy variable whose presence indicates that the religion of woman i is Christian (proxy for preferences) the percent of children born to woman i who are still living (as a proxy for "hedging" against future deaths among her children) a dummy variable whose presence indicates that woman i thinks infant and child mortality rates have been declining (as a proxy for 147 hedging against future deaths among her children) NSRVNGij the number of surviving children woman i has NOMAGEij the age of woman i AGEMARij the age at which woman i married do ll each currently married woman, married once only with j surviving children j number of surviving children, e.g., 5 or more It is hypothesized that the desired number of surviving children will be positively related to income (first and second variables), negatively related to a wage-paying job (third variable), positively related to WOMADVLR (a proxy for children having a rela- tively low price), negatively related to NOMADVSM (a proxy for children having a relatively high price), and negatively related to NOMEDUC (on the assumption that education increases the range of desired goods which compete with children for parental resources).2 It is hypothesized that Christians have a relatively weaker prefer- ence for an additional child than do Moslems or Traditionalists (seventh variable). Finally, it is hypothesized that the desire for more children will be negatively related to the survival rate among one's own children (eighth variable) and negatively related to the 2The price of children may rise with the level of a mother's schooling if educated women have higher "quality" standards-- including food, shelter, and clothing (as well as educational standards)--than non-educated women. The price proxies used in the above formulation are intended to reflect perceptions of relative prices, and hence sidestep ques- tions about the objective economic cost of children and whether it exceeds economic returns. Of course the fourth and fifth variables in formulation (6)--w0MADVLR and WOMADVSM--reflect perceptions of both economic and non-economic values of children. 148 opinion that the prospects for children dying are lower these days than in the past (ninth variable). The eight and ninth variables can be viewed as indicating whether parents take account of actual or antici- pated mortality of their children in determining the number of surviving children they want. A "pure" demand model would presumably not include age vari- ables. However, it is hypothesized that, even among women who are (presumably) still capable of given birth (i.e., women under age 45), the desire for more children will be negatively associated with age. 0n the other hand, the older the age at which women marry, the fewer their numbers of births and surviving children, ceteris paribus. Hence, it is hypothesized that AGEMAR will be positively related to the desire for more children. One version of the model for women will include these age variables. For purposes of comparison, they will be excluded from a second version of the model. Finally, it has been hypothesized that number of surviving children is negatively related-- at least above some number surviving—-to the desire for more children. Number of surviving children will be included as a variable in both versions of the model for women. It will be included in one of the two models for husbands. Husband's current age will be included in both versions of the husband model. None of the independent variables are ideal proxies for the desired independent variables, but they are the best available from the data. To recapitulate the general hypotheses about demand for children: It is assumed that demand will be positively related to income (CROPVAL and BLDQUAL), negatively related to the relative price of children, and negatively related to preferences which are l “ I. All M I I! ll 11 II I lllll I l I 'V i 'l l It'll-ll. I I III. 1‘11" 149 relatively in favor of other goods as compared to children (for which both NOMEDUC and WOMRELC are assumed to be proxies). In addition, it is assumed that demand for children also includes an upward adjustment caused by women (or husbands) taking account of the actual mortality experiences among their children (PCNTSRV) and their expectations about chances of future mortality among their children (OPMORTDN). The directions of the overall relationships between each independent variable and the response of "yes" (i.e., the respondent wants more children) is indicated in the parentheses below: CROPVAL (+), BLDQUAL (+), MOM-WAGE (-), WOMADVLR (+), NOMADVSM (-), WOMRELC (-), PCNTSRV (-), OPMORTDN (-), NSRVNG (-), WOMAGE (-), and AGEMAR (+). Descriptions of the Variables Descriptions were given earlier of the variables for income, occupation, education, religion, number of surviving children, and age at marriage, and they will not be repeated here. Table 26 pre- sents information on responses of both women and their husbands to the question, "do you want more children?" Table 27 gives percen- tages of both women and their husbands who think there are (respectively) advantages to having a large number of children, advantages to having a small number of children, and that the survival prospects of infants and children have been improving in recent years. Among women ages 25—44 with more than four surviving children, the percentages who said they wanted more children were 49 in M1, 64 in M2, 62 in L1 and 69 in L2. Among husbands ages 25-59 (monogamous 150 No NNNNV om NNNNV oN NNNNV mo. NoNNV No NomNV NN NNNNV oN NNNNV NeooN oN NNN V NN NoN w oo NNoN om NNoNV oN NNNNV om NoNNV oo NN.N ox oN NNN V No Non oo Noe oo Noe V mo NNN V oo NNoNV No NNN, o-o oo NNN V No No_ V No NNN V No Nom V mo NNN V No Noe V om Noo V NA oo. No V oo NNN V No NNN V on No. V No NNN V No NNN V NN NNN V N oN 2o V oo No, V oN NNN V mo NNN V mo NNN V oo Noo V oo NNN V o oo 2N V No NNN V NN NoN V No NoN V No NNN V No 2mm V MN Noo V o NN NNN V oo NoN V No NNN V mN NoN V No NNN V oN NNN V oo Nmo V o oo AoN V mo NNN V No NNN V mo NNN V oo NoN w oo Noe W oo NN w o oo. NNN V No ANN V oo NNN V ooN NNN V oo NNN No NNN mo No my me mm< can» mmmp ow mew: .mooEomocoE on “mm-o~ mmm< .mvcoomoz NN NNoNV oo NomNV NN NNoNV oo NNoNV oN NNoNV oo NoooV NN NNoNV NoNoN oo NoN V No NNN V oo NoNNV oo Moo V No NNNNV om NNNNV No Nomw ox oo Noe V No NNo V oo Noo V oN NN V No NNNNV oN NNNNV mo NNN o-o No NNN V oo No V mo NNN V N. NNN V .o NNN V on Noe V oo Noo V NA mo NNN V om NoN V mo NNN V No NoN V om NNN V om Noe V No ANN V N oN NoN V om NNN V No NNN V oo NNN V No NNN V om NNo V oo Noo V o oo NNN V oN 2mm V oN NNN V _N NNN V NN 2mm V mN NoN V oN NNNNV o No NNN V mo NNN V oo .Noo V oN NNN V oo Now V No NoN V mo NNNNV o oN NNN V _o Non V oo NNN V oo Moo V No NNN V oN Nmo V oN NNNNV o oo NNN V No ANN V oo Noo V VN on V No Noe V mo NNN V No NNNNV NV kumm mwm< .>._.:o muco Umtgmz .UmTsz Nag—5.230 29:03 omoN. NoNV NNNV NNNV NNNV .NoNV NoN, NNV NNV NoV NNN, any NNV NNN oNV N 22V N 22V N 22V N 22V N 22V N 22V N 22V No N4 N2 N2 oooomoo Nomo2 oooooeooN eoLoNN2o mcw>w> Newcupwgu woos “so: no» on "cowumwoo mcu ou omm>= mcwucoommm mmmopcmugmo team .oz m=N>N>gom »_ucwgeau cmcupwcu No Nongoz an um—Noeucou .cmcnpwzu woos pee; on: muconmox new case: we mmmmucmucma .om «Fame 151 husbands are included only if their wives are under age 45), the percentages of those with more than four surviving children who said they wanted more children were 53 in M1, 66 in M2, 77 in L1, and 74 in L2. Data in Table 27 are given for women ages 20 and over (cur- rently married and married once only) and for husbands ages 25-59. The proportions who think there are advantages to having large families were--for women and husbands respectively, 37 and 40 in M1, 55 and 59 in M2, 66 and 64 in L1, and 58 and 61 in L2. Percentages (of women and husbands) who think there are advantages to having small families were 52 and 56 in M1, 37 and 44 in M2, 23 and 31 in L1, and 32 and 38 in L2. Finally, the percentages who said they think that infant and child survival prospects have been improving in recent years are 78 and 86 in M1, 61 and 69 in M2, 20 and 31 in L1, and 25 and 38 in L2. Regression Results Regression results for the basic supply models were reported and discussed in Chapter V. In this chapter results will be shown and analyzed for what are referred to as demand models. Demand models have proxies for income, relative prices and tastes as the independent variables. The R25 which are produced by various formula- tions are given in Table 28 for age groups 20-29, 30-39, and 40-49 at the aggregate, district and area levels. A few of the regression results of the demand models are reported in Tables 29, 30, and 31 and compared in more detail to the results obtained with the basic supply models. '152 .m>Nmo_uxw appoouoe woo woo < “Loo :_ mmmcoommg m>wuosewwwo uco M Neon c_ mmmcoomoe w>wuoELNNN< — oM-oN moon mN NN2N .Nooeomo2 Loo -- +oN MM-OM memo MN own» .mocoomaz Lou .. +oM Mo one some: MN oNNI .moosomocos NH .cosoz vowecoe Mpacmccau No muconmax u : MNco muco uwNNNos .cmeoz umwgcoe M_ucmgeou u 3 emu;m_mzc:« "mmuoz MM MN PM ON MM PM Mm MN 2M mm NN MM MM PM +o~ MM MN No MN MN MM mm MN MM MN om «M NM Me +OM MN MM MM MN MM PM mm MN NM MN MN MM MM NM melee No MN MN MN MM MM mm NM NM om 2N MM MM OM MM-oM MM NM MN om MM MM mm MN mm MN MN NN MM NM m~-o~ meom> acoumm 2N ac_>ocoEH comm o>ox muumomoea No>N>Nom upwgu coo ucooca gown» 02M acmucmo .u MM NM NM MN ee NM Mo NM NM NM oM 2e Me NM +o~ MN NM MN mm «M MN NM Me mp mm em MM NM NM +OM mM MN MM om cc M2 M2 MM MM mm me me _e um meuoc MM MN mm MN Me PM MM NM MM MN MM we Mo «M MM-OM co 02 we MN 2M we oM NM Me NM om NM NM Ne mmuow NmmVNNsou V4ox op momouco>u< ago «cock x:_2k 0:3 “cmuewo .m NM MM 2M MM MM MM co NM NM NM oM Ne MM MM +o~ NM NM MM MM mm MM NM o2 we MM 2M MM MM NM +oM MM MM MN NN MM PM me MM MM NM MM 22 NM MM melee NM MM NM MM MM MM OM MM NM NM No we MM MM mM-om co mo mM MM MM MM MM MM NM NM MM 22 Me me mmuow mmVNNEoN Mo3<3 mcw>o3 op wmmouco>u< ago News» chMF o23 acmuemo .< NMNV NNNV NNVV NNNV NNNV NoNV AoV NoV NNV NoV NMV NoV NNV NNV NNV : 3 3 3 z 3 3 3 x 3 z 3 z 3 .586 No No N2 N2 oooomoo Nomoz «oooooeoo2 oo< .mcmm> ucmumm cw mcp>ogasm coma m>m2 muumomocm No>w>gom upwsu uco acomcH push AuV ecu .mmNNNEou Nposm ou mmmouco>u< AMV .mmNNNEoN wmcoV cu mmmouco>c< AqV ego mews» ch2u o2: mmmoucmugmo ”muconmzx new coeo3 mo meowcwoo .NN mpnoh 153 Table 28 shows that R25 are very low for the demand models. R2 does not exceed about .04 at the aggregate and district levels (except for ages 40-49 in Lushoto where R2 is about .14). At the area level, R25 range from less than .01 to about .13 except for ages 40-49 in L2 where they are around .2. R25 are generally larger when BLDQUAL is a proxy for income than when CROPVAL is the income proxy. Four of the lines in Table 28 give the R25 produced when the natural log of CROPVAL--(LN)CROPVAL--was substituted for CROPVAL to see if this non-linear version of the proxy for current income per- formed better than the linear version. In most cases (LN)CROPVAL did perform better but in no case was its coefficient or associated t-ratio very large. In all cases the R25 for these two alternative formulations are either identical or differ only slightly. In Table 29 comparisons are made of the regressions for age group 20-29 at the aggregate level for eight of the regression models reported on in Table 28. The first is the initial demand model listed in Table 28 with CROPVAL, BLDQUAL, and HUSB-NAGE as income proxies; R2 is .035. The second is the same model but with (LN)CROPVAL replacing CROPVAL; R2 is .038. In the former, CROPVAL essentially enters as zero while the coefficient for (LN)CROPVAL in the second model is positive but low. Coefficients for BLDQUAL are positive and significant at the .01 level in the first formulation and at the .05 level in the second. The only other statistically significant coefficient is for NOMEDUC (significant at the .01 level); the sign of the coefficient is negative. In the third formulation only BLDQUAL and HUSB-NAGE are included as income proxies. The coefficient for HUSB-HAGE is now 154 .oooozo3 oz< .ooomzo3 .mo ochm .ooVoz .oomNmmm .>ooozo mopo owumVV AmVo_oo_co> wcpp mom. mMo. NMo. moo. ooo. ooo. Moo. NNo. Moo. oMo. Mmo. woo. NMN. MNo. Mpo. MNo. moo. oMo. MMo. moo. mNo. mo<3iomozi NMN. MoN. NMo. Mmo. Moo. MMo. woo. Mmo. oNo. oMo. omo. ooo. ooN. Nmo. MNo. oNo. MNo. ooo. NMo. MVo. omo. ooomoflzoVi N.N. ooo. NMo. mmo. NMo. oMo. oNo. omo. ooo. mmo. MNo. MNo. MNN. NNo. oNo. m_o. Npo. MNo. MNo. moo. MVo. 3<>oomoi Pop. o—P. Moo. MMF. oNo. MNo. oMo. MMo. oNo. MMo. Moo. mop. mop. mMo. MNo. oMo. MNo. MMo. oMo. oNo. MMo. oooao z; i N.N. ooN. NMo. moo. ooo. oNo. MNo. NMo. oNo. omo. omo. oop. opp. Mmo. o_o. mmo. MVo. Noo. MMo. MVo. mmo. oaomoi NFN. Moo. Moo. MNN. MNo. Noo. Foo. omo. «No. oMo. NMo. pop. NMP. oMo. opo. mNo. moo. ooo. MNo. NNo. omo. oo<3iomoz N o<>ooNoNzoV- Now. mMo. NMo. ooN. moo. Moo. Moo. Mmo. ooo. oMo. Mmo. moo. Pop. MNo. NNo. mNo. NNo. oMo. Mmo. moo. mNo. oo<3immox o o<>oomoi N_~. mNN. NMo. mmo. mNo. oMo. oNo. oMo. MNo. NMo. mmo. NNN. NoP. Nmo. o_o. opo. MNo. MMo. oMo. MNo. MMo. oo<3immox a oooxoi Nmeooz ccoEmo Mop. omN. oNo. NoM. moM. oNo. Now. ooM. MNM. mop. moo. mNM. Now. owo. moo. NNN. Nmo. MNM. ~m_. moo. NoM. mozmm>i oNN. oMM. Noo. MNM. ooo. woo. MoM. oNM. MNM. mop. mmo. MMM. NNN. oMo. ooo. MNN. Foo. oNM. NNN. oMo. NoM. mozoo< M moozozi Fmpmuoz x_mm=M NNNV N.NV NoNV NoNV NNNV NNNV NoNV NmNV NoNV NNNV NNPV NNNV NoPV NoV ANV ANV NoV AmV NoV NNV NNV NNV mo-oo mMioM mmiom ooioo mM-oM mmiom moioo mM-oM m~-o~ moioo mMioM mmiom ooioo mmioM mmiom moioo mMioM m~-o~ moioo oMioM omiom No N3 N2 N2 oooomoo Nome: oooootooo _wooz cowmmocomm mo-oo use .mM-oM .mmiom moooco mo< co» mNmooz ucmEmo oco Mpmooz MNQQ=M mooVLo> cow mmm mo comweoosoo .wm mpaoh 1555 No>oN oN. Ne ooooe-o ooeo_cooonN _o>oN mo. No owooe-o Neoo_o_ooomo No>oN No. No o_oee-o ooooNo_oon. NooN.NV No_N.oV Noom.oV Nomm.NV Noo.N_V NNV o_o. NNo N_oo. - Noo.o moo. - ooN. - .ooo.m ”opooo NNNN.NV NoNo.oV Ammo.oV AoNo.NV Noo.NVV NNV NNo. NNo ooNN. - NNN. - NNo. - Nm_N.o .ooo.N ”ooooo AVoN.NV Nooo.oV Ammo.oV Nooo._V NNoN.oV NNV NNo. NNo Nmoo. - NFo.o NNo. - ooo.o .oom.N ”Ncooo NNoN.NV NNNN.oV NNoo.oV Nooo.oV NNo.o_V NNV mVo. NNo Nooo. - NNo.o NNo. - ooo.o .oNo.N "moooo N_oN.NV Nomo.oV AVNo.oV NoNN.NV NooN.NV NVNN.oV 23V moo. NNo .oo_. - oNN. - NNo. - Noo_.o oNo.o .omo.N "ooooo NNoo.NV Nooo.oV Non.oV NNNN.NV Aooo.NV NoN.NNV NNV moo. NNo oNoN. - NNP. - Noo. - NoVN. - .NNN.o NNoo.M ”ooooo Nooo.NV NoNo.oV Non.oV Noom.FV Nooo.NV NNNN.NV Ammo.oV 33V oMo. NNo .ooN. - NNN. - moo. - ooN. - NoN_.o Noo.o .ooo.N “ooooo Ammo.NV Nooo.oV NNom.oV NooN._V NoNo.NV NoNN.oV NNo.oFV ANV oN-oN . moo. NNo .NoN. -. oNN. - Poo. - NFN. - .NN_.o Voo. - «ooo.o ”Noooo zoooz N z oooozoz oooo2o3 No>oo2o2 oo<2-omo2 ooooooo o<>ao2oflzoV o<>oo2o m Homo n .Nm> mopoo_eo> osmocwomuco icmuco .owo mm-om mmo< .Axpco coco owwgeoz .oowgeoz MpucmcgooV :oeo3 .ccom :mLucho mo Noosoz mo mucocwscmpmo we m—mooz cooEmo nouumNmM Low mupommm No com_eooeoo "mupzmmm cowmmmgmwm Fm>mV mummmcoo< .om «poo» 156 significant at the .10 level (and the sign is negative). In the fourth formulation in Table 29, (LN)CROPVAL is included and HUSB-WAGE omitted; R2 drops slightly to .033. In the fifth through eighth sets of results only one income proxy is used; they are, respectively, CROPVAL, (LN)CROPVAL, BLDQUAL, and HUSB-NAGE. Consistent with the earlier results, only the coefficients for BLDQUAL is statistically significant (at the .05 level). In Table 30, comparisons are made at the district level for age group 20-29 for the third, fifth, sixth, and seventh formulations shown in Table 29 (i.e., formulations in which income proxies are, respectively, both BLDQUAL and HUSB-NAGE, only CROPVAL, only (LN)CROPVAL, and only BLDQUAL). Results are generally similar to those at the aggregate level except that BLDQUAL is shown to have a negligible effect on NBORN in Lushoto but a considerable effect in Moshi district (significant at the .01 and .05 levels respectively in the two relevant formulations). In fact, none of the coefficients for the income proxies are even close to being statistically signifi- cant in the models for Lushoto district while for Moshi district the coefficients for both HUSB-NAGE (negative sign) and (LN)CROPVAL (positive sign) are significant at the .10 level. In all four formu-w lations for Moshi district the coefficients for NOMEDUC are statis— tically significant (and negative). Nevertheless, the largest R2 in Moshi district is only .055 while in Lushoto district the largest R2 is only .017. Area level results are given in Table 31 for the first two of the four formulations shown in Table 30 (i.e., with income proxies being, respectively, both BLDQUAL and HUSB-NAGE, and only 1557 NosoN oN. No oNNee-N oeooVoVoonN FosoN mo. Ne OVNNL-N NoooNoN=o_mo No>oN No. No oVoeL-N NoooNNNooNN. NoNo.NV Noom.oV NNom.oV NNoN.oV NoN.MNV NNV NNo. oN. Poo. - NoN. - NmN.o Noo.o .oNo.N ”ooooo oNoImoo Nooo.oV NoNN.oV NNNm.oV NooN.oV Ammo.oV Ammo.oV 23V oNo. oN, NNo. - oNN. - NoN.o oNN. - Npo.o .moo.N "ooooo oNo2moo NNmo.oV NooM.oV on.oV 2Noo.oV NNMN.oV NNN.VNV NoV NNo. oNN oNo. - NNN. - NmN.o ooN. - Noo.o .ooo.N "ooooo oNo2moo Nooo.oV NNNo.oV NNoo.oV Nooo.oV Nooo.oV NNo.N_V NoV . o_o. oNN ooo. - NmN. - oo_.o oop. - moo.o .ooo.N "woooo oNozmoo Nomo.NV NNoo.oV Nooo.oV Noom.NV Nooo.oV NoV ooo. mmN .e.N. - NNN. - NNN. - NoNF.o .NoN.N "eoooo N2mo2 Nooo.NV NooN.oV Nooo.oV ANoo.VV Nooo.NV NNmo.oV NNV ooo. moN Nooo. - mom. - NNN. - ooN. - 2NN_.o .Noo.N "oeooo N2No2 NooN.FV NooM.oV NNoo.oV NooN.NV ANoo.oV Noom.oV NNV oNo. mmN NoNo. - ooN. - ooF. - NNN. - moo. - .Moo.m "woooo N2mo2 Non.NV NmNM.oV Nooo.oV Nooo.NV Noom.NV Noo_.oV NHV oN-oN moo. NMN NmoV. - oom. - NoN. - NNoo. - .oNN.o .oNo.M "ooooo Homo2 zoooz 2 2 oooo2o2 ooo22o: No>oo2o3 oo<2-omo2 oNoNoNzoV o<>ao2o N name n .Le> me_oewgo> uceeceemeco iceuco .eeo m~-o~ mem< .Ampco euco covecex .eeoggez mpuceNNooV ooze: .ecem cecop_2o we Loganz we mecocwseeueo we mpeoez eceemo oeueepmm New mupemem me compeeeeeo ”mapamem :ermmegmea Pe>e3 uuNNumwo .oM epoch 158 .o>o. o.. No o.ooe-o Neoo.o.oo.mx .o>o. oo. oo o.oee-o .eoo...oo.mo .e>e. .o. um eNuegiu u:oo.o.co.mo ..om..V .NNo.oV .NNN.oV .om..oV .o.o.oV .ooo.NV .NV NNo. No om.. - oNN.o omm.o ooo. - ooo. - .Noo.N ”ooooo No .No...V .ooo..V .moo.oV .omN.oV .ooo..V ..No.oV .oV NMo. No o... - Noo.o Noo.o o...o oo.. - .Noo.N ”o.ooo No .ooo.oV .ooN..V .oo..oV .ooN.oV .oom..V .Nom.oV .NV Noo. No Noo.o Nmoo..- .oo.o oNN. - oNo.o .ooN.N ".Nooo .o .ooN.oV .oom..V .oNo.oV .Noo.oV .ooN..V .ooo.oV .oV ooo. No Noo.o Noo. - .o..o ooN. - oom.o oooN.N "Noooo .o .ooo.oV .NNo.oV .ooo.oV Nooo.oV .ooo.NV .NV ooo. o.. ooo.o .eoomooo NNo. - o.N. - .oo.o ooNo.N ”coooo N2 .Noo.oV .NN..oV ..No.oV .Noo..V .ooN.NV ..V o.o. o.. oNo. - Noeomooo ooo. - .NN. - ooo.o .oNN.N ”o.ooo N2 .oNo.NV .o.m.oV .omN..V .oom..V .ooo.oV .NNo.oV NNV ooo. MN. ooo.. - oo.. - mom. - Noo. - ooo. - .Nmo.m “ooooo .2 .NNN.NV .ooo.oV .o....V .NNo..V .NoN.NV ..oN.oV .oV oN-oN .N.. om. ooo.. - ooN. - o.o. - Nooo. - Noo..o .ooo.M “ooooo .2 zoooz No 2 oooozoz oooN2o3 No>o<2o2 oo<2-omox ooooooo o<>aomo ueeu n .2e> we.oo_2e> oceeceoeec. iceuc. .eeo mmiom meo< ..M.:o euco eewceez .ue_ceez M.N:egcooV cmEe3 .ccem cece._;o we censoz oe mucec.ELeoeo Ne m.moez eceeeo eeuee.em Lew mu.:mem we cem.coeseo Noo.omem :ewmmmemex pm>oo ee2< ..M m.nep 159 CROPVAL). The models perform best for M1. For Ml coefficients for BLDQUAL are positive and significant at the .01 level; coefficients for HUSB-WAGE and WOMEDUC are negative and (in the first formulation) significant at the .10 and .01 levels respectively. In the other three areas, none of the coefficients for the independent variables are statistically significant (with the exception of WOMRELC in one of the formulations for L1 where the coefficient was significant at the .05 level--and negative as hypothesized.) The regression results summarized in this chapter indicate clearly that at least for these p0pulations, the demand variables used here are capable of explaining very little about fertility levels and differentials. Among the demand proxies analyzed, only BLDQUAL--as a proxy for household wealth or permanent income--and WOMEDUC--as a proxy for preferences and opportunity costs-~were consistently statistically significant, and then only in Moshi district (actually, only in M1). However, in M1 the relationships between NBORN and each of these two independent variables were in the expected directions (i.e., positive for BLDQUAL and negative for WOMEDUC). Results of the Discriminant Analysis Tables 32 through 35 present results from the discriminant analysis of both women and husbands who indicated that they either did or did not want more children. Women are included who are ages 25-44, currently married and married once only, and who currently have five or more surviving children. Husbands are included who are 16D ages 25-59 and who currently have five or more surviving children. A monogamous husband is excluded if his wife is age 45 or older. These conditions were imposed for the following reasons: Only the desires of parents who still believe themselves capable of having more children are relevant for the analysis. Hence, only currently married adults are included. Women who have been married more than once are excluded so as to eliminate the possible effects of the woman's disrupted marital experience on her desire (or lack of same) for more children. Adults who are or believe themselves to be physiologically incapable of having more children should also be excluded. Thus, women age 45 or older are excluded and (somewhat more arbitrarily) men age 60 or older are excluded. For the same reason, monogamous men whose wives are age 45 or older are also excluded. The desire for at least a few surviving children is still universal in rural Africa including in the societies studied here. On the assumption that the response "I want no more children" is less credible coming from parents with fewer than five surviving children (and who are apparently still physiologically and otherwise capable of having more children), the analysis is limited to respon- dents with five or more surviving children. The independent variables included in the analysis and the expected relationship between each independent variable and the desire for more children (the "yes" response in the dependent variable) were all identified earlier in this chapter. Tables 32 and 33 give for women and husbands respectively the averages and standard deviations for each of the independent variables for both those who responded "no" and those who responded "yes." In the upper 161 No.mo No.mo Nm.mo No.oo No.oo NN.Nm No.oo .eoo. .o N no No.- NN oN oo oo oo oo oo. .otoe oeV =oe. .oz- No No oo oo oo oN. o.N .oLoeV =mo2= .oz- .o.mV .o.mV .o.NV .N.oV .N.oV .M..V .N.oV ...NV .N..V .m.NV .e.NV .o.NV NN.NV .o..V ..N. o.N. o.N. o.o. o.o. o.N. o.N. o.N. N.N. o.o. N.o. ..N. o.N. o.N. oo.tee2 cog; eo< n.2e5e3 .N.oV .e.NV .N.oV ...oV .N.oV .N.oV No.oV No.oV NN.NV .o.oV .o.NV No.oV .o.mV .o.oV o.oo N.om N.om N.om ..MN o.om N.NN o.NN N.MM o.oo o.om o.o» ..MM o.oo ooo ucegcoo n.2o5e3 .oo..V .oM..V .No..V .oo..V .N...V .oo..V .oo..V .oo..V .NN..V .NN..V .N...V .oo..V .oN..V .Mo..V oN.o mm.o oN.o N..o o..o oN.o oN.o NN.o oo.o No.o No.o .N.o No.o oo.o eoeo..2o oe.>.> -NeM ue guess: .oo.oV .om.oV .oo.oV .NN.oV Noo.oV .om.oV Noo.oV ..o.oV Noo.oV .om.oV Noo.oV .No.oV Noo.oV .oo.oV o..o o..o oN.o No.o oo.o No.o oN.o oN.o oN.o N..o .N.o No.o No.o om.o ezoo No..eoeo2 o..oo Noo.o.oo No.o.V .o...V 3o...V .m.o.V No.o.V .N.o.V .N.o.V .o.N.V .....V .N.o.V .N.o.V .o.N.V No.o.V No.N.V o.oo N.oo o.No N.oo N.No o.oo o..o o.No N.No o.oo ..No o.oo o.No N.NN oo.>.>eom coo -o..2o oo N "Neo2 .No.oV .oo.oV .om.oV .NN.oV Noo.oV .oo.oV .NN.oV .oo.oV Noo.oV .oo.oV .o..oV .oN.oV .oo.oV .oo.oV .N.o NN.o o..o N..o oo.. oo.. .o.o mo.o o..o oN.o No.o No.o oo.o .N.o oN.o“.Loo m. ce.oo.o3 M.ceEe3 .NN..V Noo..V .oN.oV ..o..V .oo.NV .NM..V .NN.NV .NN.NV .om..V .NM..V .o..NV Noo..V .No..V .No..V oN.o NN.o N..o om.o oo.. No.o oo.. oo.. mo.o No.o oo.. N... No.. No.o oo..oo2om *0 row» m . c853 .No.oV ..M.oV Noo.oV .oo.oV .No.oV ..m.oV .oo.oV Noo.oV Noo.oV .oo.oV .oo.oV .oM.oV .No.oV Noo.oV .N.o oM.o o..o oN.o .N.o oo.o om.o mo.o oN.o oN.o oo.o No.o No.o .m.o N..EN2 ..NEN 0. . .muc>e< ”cewcwoo Noo.oV .om.oV .No.oV Noo.oV .oo.oV ..m.oV Noo.oV ..o.oV .mo.oV .om.oV .oM.oV .No.oV Noo.oV Noo.oV oo.o .o.o oN.o .N.o .o.o oo.o om.o .N.o .N.o No.o oN.o oN.o No.o .o.o N..Ee2 ooeeo oo .ouc>e< "ceNcNQo .oo.oV ..N.oV .o.oV .o.oV .NN.oV .M..oV .mN.oV .NN.oV ..N.oV No..oV .NN.oV Noo.oV .NN.oV .oN.oV o..o Mo.o o.o o.o oo.o No.o No.o N..o oo.o NN.o oo.o o..o oo.o No.o ooo oo.2ea-ooe3 .oo..V 3o...V .No.oV .oN.oV .No.NV .oo..V ..N.NV .oN.NV .oo..V .oo.oV .N..NV .No.NV .No.NV .oo..V oo.o oM.o .N.o oo.o oo.N mo.N o..N oo.N om.o No.o oo.N oo.N oo.. oo.. Nooo. 2o..eoo . oo.o..oo o.o2\2 .N.o_V .N.oNV .N.omV .N.NV .o.NV ...oV .o.o.V .N.oV No.oNV .N.o.V No.o.V ...oV .N.o.V N..N.V N.o o.o. o.o. o.o N.o N... o... o.o. o.o. o.o. N.o. .... ..o. N.o. .oo "momV ooooo -oco moato o.ooNx .o.V .o.V .M.V .N.V ...V .o.V .oV .oV .NV .oV .MV .oV .NV .NV ..V we» oz we» ez we» ez we» ez we» ez we» ez me> ez No .3 m: .z euecmoo Nome: woooeeoo< we.eo.2e> .memocoeecoe c. oco.oo.>eo oceeceoMV coco—.2o oc.>.>eeM wee: Le e>.u no.3 .oo-M~ meoq .Ax_co eeco ee.ccez .ue.ceez M.NceggooV :eEez coo m_m3.ec< ucec.s.cem.o m2» 2. owe: we.ee.ce> acmecmeeec. ego Lou mce.uo.>eo oceeceum oce memose>< .NM o.oe» 162 NN.N. No.NN No.oo NN..o No.oN N..om No.oo .eooo oo N we oo.- o. m. on .o on oN oo. .oeoe oeV .oe. .oz- No No oo No oo No. .oN .oeoeV .moNg .oz- No.oV .N.o.V .o.NV ...oV No.oV .N.oV No.oV .o.oV .N...V .o.oV No.oV .o.NV ...NV .o.NV o.oo o.oo N.No o.oo ..oo o.oo N.No o.oo N.No o.oo o..o N.oo o..o N.oo moo oooeeoo m.eceemex Noo.oV .oo.mV .oN.NV .No.mV Noo..V .oN.NV .NN.NV .NN.NV ..o.NV Noo.mV .oo..V .oN.NV .Nm.NV .No.NV No.o No.o. No.o No.o oN.o N..N No.o NN.N oo.N oo.o No.o oN.N oo.N oo.N ooeo..2o oo...) . -Lom we seesaz .oo.oV .NM.oV .No.oV ..m.oV .oo.oV Noo.oV .oN.oV .oN.oV Noo.oV Noo.oV .oo.oV .NM.oV Noo.oV Noo.oV oo.o .o.o NN.o oo.o oo.o oo.o .o.o No.o oN.o No.o oN.o oN.o No.o NN.o ozoo 2o..eoeoz o..2o noo....ao .M...V .N.o.V .N.N.V No.oV ...N.V .o.N.V .N.o.V .o.N.V .N...V .o.N.V .o.N.V NN.N.V .o.N.V .e.N.V N.oo ..oo ..No o.oo o.oo ..oo ..No N.oo o..o o.oo o.oo ..mo N.oo N.oo oo.>.>tom cot -o..2o co N “.202 .oo.oV .No.oV Noo.oV .Nm.oV .NN.oV .oN.oV .NN.oV .oo.oV Noo.oV ..o.oV .NN.oV Noo.oV .No.oV .No.oV .o.o No.o MN.o No.o No.o oo.o o..o N..o NN.o om.o ...o N..o NN.o NN.o om.eNoN.oN e m. eceome: Noo.oV ..o.oV Noo.oV .oN.oV No.oV No.oV .Nm.oV .No.oV ..o.oV .mo.oV .oN.oV .oN.oV .oo.oV Noo.oV oN.o oN.o N..o No.o oo.. oo.. oN.o oN.o .N.o N..o No.o oo.o oo.o mo.o oa.om.eoo Noo.o aw 3m 3932.3: .oN.NV Noo.NV N.o.NV .NN.NV .NN.NV Noo.NV .oo.NV .oo.NV .oN.NV .oN.NV .mo.NV .MN.NV .No.NV .oN.NV oo.N oo.N .N.N oo.N oo.N oo.N oo.N oo.N oo.N oo.N oo.N NN.N oo.N No.N oe..oo2om .o meow» m.eceomaz .No.oV .Nm.oV .oo.oV .Nm.oV Noo.oV ..m.oV .oM.oV .oo.oV .No.oV ..m.oV .oo.oV Noo.oV .oo.oV Noo.oV .N.o No.o NN.o No.o oN.o No.o No.o .o.o NN.o No.o No.o No.o oN.o om.o 2..2o2 ..NEm oN .ouc>e< .ce.c_oo .oo.oV .Nm.oV Noo.oV .NM.oV Noo.oV ..m.oV ..M.oV Noo.oV Noo.oV ..m.oV .No.oV Noo.oV Noo.oV Noo.oV .N.o No.o .N.o No.o NN.o No.o .o.o oN.o .N.o No.o oo.o No.o NN.o oN.o 2..2oo ooeo3 co .muc>e< "ceNcVQO Noo.oV ..M.oV .NN.oV .oN.oV .oN.oV Noo.oV .oN.oV .oo.oV Noo.oV .Mo.oV .oN.oV .oo.oV .om.oV Noo.oV oN.o oo.o No.o No.o oo.o o..o o..o oN.o N..o NN.o oo.o oN.o N..o oN.o ooo oo.NeN-ooax ..o..V ..o..V Noo.oV Noo.oV .oo..V .oo..V .oo.NV .NN.NV .oo..V .No..V .oo..V ....NV .NN..V .No.NV oN.o No.. oN.o oo.o No.N NN.N oo.. oN.N om.o No.o NN.N No.N oo.. oo.N Nooo. No..eoo ooVo..oo o.o2N2 .N.N.V No.o.V .N.o.V .N.oV .m.o.V .o.oV No.o.V No.oV .N.o.V No.o.V No.o.V .o.NV .N.o.V No.oV N.N. o.o. o.N o.o o.N. o... o... o.o o.o. o.o o.N. N.o. .... ..o. .oo "momV ooooo -eco Meeco e.e2\3 .o.V .o.V .N.V .N.V ...V .o.V .oV .NV .NV .oV .mV .oV .MV .NV ..V we) ez Me» oz we» ez we» oz we» ez we» ez we» ez N3 .3 N: .z eoe2m33 .omez euooeeom< we.eo.co> ego: 2e e>.o 2a.: .eMiMm meo< .cez Leo Ammmezocecoe c. oce.oe.>wo eoeeceoMV Amereoecez ». Mo em< Leno: eo.3 eceV :eLe.V;o m:.>.>2:M o.o».ecq u:ec.E.cem.o on» 2. new: we.ee.2e> oceeceoeocV ecu Leo ocewue.>oo eceoceum nee meoece>< .MM o.oa» 163 portions of Tables 34 and 35 the standardized discriminant function coefficients (equivalent to standardized beta coefficients in regres- sion analysis) are given for each of the independent variables. A positive sign means the variable is positively associated with the desire for more children (the "yes" group) and vice versa for a negative sign. The expected sign for each independent variable is indicated in column 2. Various criteria are available for determining the sequence with which variables enter the discriminant function. The one chosen (somewhat arbitrarily) for this analysis is a stepwise method which maximizes the Mahalanobis distance between the two groups [Klecka, 1975: 447]. Actually, the technique selected is important only if the contribution of one or more independent variables to the function is sufficiently weak that the variable(s) would be excluded had some other criteria been used. (Conversely, a variable disregarded by the selected technique might be included in the function by another technique.) In the results reported here, there were only a few cases in which one or more independent variable was eliminated from the function by the Mahalanobis method. Moreover, results obtained when a few runs were made (for comparative purposes) using other selection criteria [see Klecka, 1975: 446-48] did not differ much from results reported here. Several indicators of the discriminating power of each function are given in the lower portions of Tables 34 and 35. The first indicator is the percent of all cases correctly classified (by the computer) subsequent to the creation of the discriminant functions. This is compared to the percent of all cases that would 164 .mexV N eoeco u N: woo .ecV . eoeco u .: oco o fiMN - N: N on + .cV.N: + .cV .N: .e.- u o econ; wuw o o No 3e. ee. .e>e. eecoce.eu emooeeo o.o».oce «so see. eeoo.exe u -- mm.ne.2e> eon 2a.: N o.o».ece u N< "Mo.ee.2e> woe oaezo.3 . M.mm.ec< u .< "Meuez Mo.. oo.. MoN. NMN. OON. oo.. NNM. oNN. oM.. NM.. oNN. oo.. .M.. mM.. Na- M.o.. Oom. MNN.. .o... mNo.. Moo. MoM.. NNN.. .MO. on. ooo.. .om. ooo. N.o. .eeceNeNN.OuVo- Moo.- oMo.- oNN.- oNN.- MMo.- moM.- ooo.- mmM.- NMM.- MMM.- oNo.- OMM.- o.M.- ooo.- .mexV N eoeco ooM. ooM. Moo. M.o. oNM. M.M. Noo. oNo. omN. MoN. woo. ..o. ooM. NNM. .ecV . eoego "me.e2uceo- oNo. OmN. oNo. mMo. Moo. .No. .oo. .OO. MNO. .MO. OOO. ooo. ooo. ooo. eeceu.o.co.m- N. O. .. m .. m N. m .. m N. O. .. o Eeuoecu Ne meecmmo- N.N. m... N.o. o.o. o.oN M.o. o.MM o.ON o.o. o.o. M.MM. o.oo o.oM M..M eceoem-.zo- m..N o..N N.mo N..N o.MN o.oo o.MN M..N N.mo o.oN N.MN N.NN o.oN M.ON oe.w.mme.u V«See-:5“. n- m.oM N.MM o.MM o.om M.oM ...M o.NM eeeecee= N. peecceo N- o.Mo o.Mo M.Mo m.oo o.oo M.NM o.oo .eueu No a we ewe»:- NN oN oo oo om oo. .eces ecV =ez= Nongoz- No No oo oo oN. O.N .eceeV 2mm»: L3252- .N..- oo.. M.N. NMo. ooo. .mM. Mom. + eme.cgez we em< mM..- ..M. MoM.- ooo.- ooo. omM.- ONM.- - eo< aceegao No..- mm..- ONM.- omo.- ooM.- moo.- omo.- mMo.- NoN.- NoN.- oMo.- ooN.- M.M.- ooo.- - cecn..cu oc.>.>eom.ez N.N. No.. Moo. moo. Mo.. MM.. ooN. mON. ooo. NMo. oON. NM.. NOM. ooN. - czeo mu..eu2ez e..2o H..o....eo M.o. mMo. .oM. MoM. oN.. oMo. MM.. MoM. MMM. oNM. mM.. moo. OMM. mmo. - oc.>.>2:m cw; -o..:o No a ”use: MNo.- omN.- oo..- .N..- -- --- MoN. MoN. moM.- .NM.- MNN. mMN. .o..- om..- - co.um.ezo ..o..wm ..o. ooo. oo..- Mo..- Mo.. ONN. oNN.- -- .o.. .o.. ooo.- No.. moo. .M.. - oc..eegem No we» .M..- MM..- NMo.- oNM.- MoM.- o.M.- MNO. Mo.. NmN.- oOM.- M.N.- Mm..- MOM.- moN.- - >......B... ..esM e» .muc>u< ”:e.c.eo NmM. ooM. oMo.- Noo.- oMo. omO. N.o. ooo. N.M. NNM. ooM. M.o. mMM. NoM. + >............ em.23 eu .muc>u< ”cewc.eO oo.. MON. -- -- ooo. .o.. .MN.- oNN.- mMM. ONM. oo..- No..- Omo.- ooo.- - aeo o:.Mee-eoe3 ooN.- MNM.- NNM.- wMM.- ooo.- Moo.- mo..- MmO.- Noo.- Noo.- oNo.- ONO.- omo.- Noo.- + xeec. Mu..e:O oe.o..oo o.o2\2 MoN.- MMN.- .mM. o.o. oON.- MmN.- oM.. MMO. -- -- oMO.- N...- ii. -- + .OO ”msMV eeozu -ego meegu u.e2\z .o.V .M.V .o.V .m.V .N.V ...V .o.V .oV .oV .NV .oV .mV .oV .NV .NV ..V N< .< N< .< N< .< N< .< N< .< N< .< N< .< N3 .3 N: euezmo3 .omez woeoecoo< :m.M mm.ne.ce> mace.e.owweo ceNHecom pcec.e.cem.o ee~.e2eecoum ouuexm uceeceeeec. mce.uueaN oo.».ammm ecu we Lezeo mc.oe2.5.cum.o ego we meeuoe.ec. mee.co> oce mace.u.ooeeo oo.Necou ucoc.e.cem.o oe~.o2eecoum House: New oo.:mem o.o».oc< ucoc.e.eum.o .oM m.oeh 165 .MemV N eaeco n N: ecu .ecV . eoego u .2 use O fllbw i N: + .cVAN: + .cV .Ne.eV 30. co» .m>m. euceee.eu wmseumo M.Mm.ece ecu eecw emu:.uxu N on 9 ages: .< .mwuez o.o».ece ezu :. ceee..;o o:.>.>N:M we Nessaz n N< .m.mm.oce :. we: coco..2o oc.>.>eam we consaz u mMM. oMM. MoN. .NN. m.N. om.. mON. NON. mMN. Mm.. mN.. Mo.. mN.. mM.. Na- moo.. ooM.. .oo.. mMN.. Mo... oMo.. o.O.. omo. mmN.. MM... oom. oom. Nom. M.m. .euceeeww.o uVo- OMN..- ON...- NM...- NNo.- mNN.- Moo.- ONM.- mOM.- Mom.- oMo.- OMM.- NMM.- Noo.- Noo.- .wemV N ooeto oMo. o.o. oMM. NoN. oNM. MMM. ooo. Moo. o.M. oNN. ooM. ooM. MMM. ..M. .ecV . oeeco “mo.eeoceo- .NO. NMO. Noo. MN.. M.O. N.o. M... omo. .oo. moo. .oo. ooo. ooo. ooo. eucoe.w.co.m- .. .. .. O. o. m N. .. N. .. N. o. N. .. Eeeeecm we memeoeo- o.NN N.oN M.ON N.M. M.NN N.ON o.o. M.N. o.NM N.MN M.oM o..M m.oM o..M ecoeem-.2o- N.oo N.oo ..Mo o.NN M.MN M.oN O.mo N.oN M.Nw ..Mo N.oN M..N o.oN o.MN ue.w.mmo.u 25.2.2.28 u- NN..o NN.oo N..MM No.oo NM.Mo NN..M NO.MM eeeoceez w. Neegcee n- NN.MN “o.NN No.oo NN..M ro.MN a..mM No.Mo .oueu we a mo :me.=- M. M. .o OM oN oo. .ecee ecV =ez: Noose:- No No No oo No. .oN .oeoeV =ooN= eooEoz- o.o.- ooo.- NNM. oo..- Mmo.- oNo.- ONM.- N.o.- oo..- ooN.- OmM.- ooM.- omN.- Noo.- - eo< ucecceo oNM.- moo.- MmM.- MoN.- NmM.- .NM.- ooo.- - cege..2o m:.>.>2:m .ez ooM.- oOM.- -- -- ooM.- moM.- N.o. oMo. oM..- NO..- Moo.- oNo.- MON.- oN..- - E..eo m3.3.8: o..oo Noo.o.oo OM.. OM.. NM.. oo.. .oM. NoM. m.N. oo.. ON.. oM.. oNM. .OM. .oM. oMM. - o:.>.>N:M cog iv..2o we a "wee: N...- MON.- mMO.- oOM.- M.M. .NN. oMN. OMF. N...- o.M.- .mN. oo.. oo.. Noo.- + no.2oom.eo oNM. moo. MNN. oMM. -- -- oMo. Moo. ooN. .mN. oNN. omN. moo. Noo. - oo.wacgu N3.8V ONM.- .NM.- Mmo. Mo.. -- -- oN..- oo..- ooo.- NNo.- oMO. omo. ooo. - m:..ee2eM we we. .oo.- Moo.- Nm..- MMM.- MM..- Mo..- .om.- NMm.- oNo.- M.M.- MOM.- mmo.- oNo.- mMo.- - N.NEeN ..eEM eo .ouc>u< Noo.o.eo MoM. oNM. moN. NoM. om.N oNN. MoN.- omN.- ooo. MMM. MmO. ooo. OMM. omN. + >......Bw moo.33 eu .muc>e< "ce.c.eo NNN.- NoM.- NN..- mo..- .MM.- MmM.- N.M.- .OM.- O...- OMN.- oMM.- MNM.- moM.- Moo.- - new m:.3ee-eoo3 --- moo.- MNo.- M.o.- ooN.- NOM.- oMM.- NOM.- oo..- MNN.- NM..- om..- mMN.- o.M.- + News. mu..oao oo.o..oo o.ooNz o... .... omN. NMM. Mmo. No.. ONo. ooo. o... N... o.M. oMM. m.N. mNN. + .OO ”msz ueuee -eco woego e.ez\: .o.V .M.V .o.V .N.V .N.V ...V .o.V .oV .NV .NV .oV .mV .oV .NV .NV ..V N< .< N< .< N< .< N< .< N< .< N< .< N< .< N3 .3 N: .z euegmo3 .omez euooegom< co.M Me.ee.ee> mucm.e.wweeo oo.NUoou ucoc.ewtem.o ee~.o2oecoom ouuoxo eceeoeeeoc. mce.uecow oo.N.omem use we Lezeo oc.ue:.E.Lem.o we» we mgeooe.ec. moo.eo> nee mucm.u.wwmeo oo.Necau ocecwe.cum.o oe~we2oecoum use: New mu.omem o.o».o:< ucoc.e.eem.o .MM o.oe. 166 be classified in a strictly random process if the proportions "yes" and "no" were known in advance. Below this are given chi-square, degrees of freedom and signi- ficance levels for the function. Next, the centroids are given for each of the two groups followed by the absolute difference between the centroids (D). R2 is given on the bottom line. Each case has a separate discriminant score which is computed by multiplying each variable (for a given case) by its corresponding coefficient and adding together these products. The group centroid is obtained by averaging the discriminant scores for each case in a group, thereby producing a group mean or group centroid. (For all cases from both groups combined, the score will have a mean of zero and standard deviation of one.) The difference between the two group means indi- cates how far apart the groups are on that dimension; with only two groups there is only one dimension [Klecka, 1975]. This difference is identified as distance (=0). According to Namboodiri [1974: 47], R2 (the square of the multiple correlation coefficient) is related in the following way to 02 (often referred to in discriminant analysis as Mahalanobis 02): 2 _ cb _ ("1"2) 2 R ‘ l + o "here 4” (n] + n2)(n1 + n2 - 2)D with nI and 02 being the numbers of cases in each of the two groups Because about 60 percent of all cases are from Moshi dis- trict, the means and standard deviations at the aggregate level are biased toward characteristics of Moshi respondents. Most of the 167 discussion and analysis will focus on area level results. As noted earlier, only respondents with five or more surviving children are included in the analysis. Some of the independent variables are associated with the "yes" response in the dependent variable in the direction hypothesized while others are not. Income was expected to be positively associ- ated with the desire for more children; the mean values in Tables 32 and 33 were expected to be larger for the "yes" group than for the "no" group. More often than not the means are larger for the "no" category than for the "yes" category, and in most cases the coeffi- cients are negative, especially for BLDQUAL. The holding of a wage-paying job was assumed to be a proxy for the costs of children being relatively high and hence the sign was expected to be negative. For husbands it is, without exception. For women the sign is negative in M1 and positive in M2 and L2. This variable has very low discriminating power for women in L1 and does not enter the analysis. (Less than five percent of the women in L1 and less than ten percent in the other three areas had wage- paying jobs.) It was assumed that being of the opinion that there are advantages to large families indicated that the price of children was relatively low for that reSpondent, while conversely, the opinion that there were advantages to small families indicated that children were relatively costly for that person. Thus, the coefficient for advantages of large families should be positive while the coefficient for the advantages to small families should be negative. Similarly, the means for the advantages-large variable should be larger for the yes" group than for the "no" group while the means for the 168 advantages-small variable should be the opposite. This is in fact the case without exception for both men and women in all four areas although the absolute and relative differences between the means for "yes" and "no" vary substantially among the areas. In most cases the two variables enter the functions with the expected signs; the excep- tions are Ll for women and M1 for both sexes. Even for these cases, however, the difference in the size of the two coefficients is con- sistent with the expected relative relationship between them. For example, for men in M1 both variables enter the functions with nega- tive coefficients, but the coefficients for the advantages-large variable are about -.27 while the coefficients for the advantages- small variable are about -.95, indicating that the latter opinion is much more highly negatively correlated with a "yes" ("wants more children") response than is the former. Thus, the discriminating power of these two variables is rather good and--at least relative to each other-—in the direction expected. Years of schooling was assumed to be a taste proxy; it was expected to be negatively related to the desire for more children. In fact, it is a rather poor discriminating variable with no consis- tent pattern. Except for women in L2 (where the coefficient is high and positive), the coefficients are low and inconsistent among areas and between men and women. Among men, Christianity tends to be positively related to the desire for more children--contrary to what was expected. Among women in the Lushoto areas the signs are negative. Polygyny was hypothesized to be positively related to the desire for more children. It is in the Moshi areas but not in the Lushoto areas. 169 It has been noted at several points that it is frequently hypothesized that parents adjust upward the number of children they want as a hedge against actual or possible mortality among their children. It was expected that the desire for more children would be negatively related to the mortality rate among a respondent's children; that is, parents whose children have experienced a relatively high mortality rate would be more likely to want additional children than parents whose children have experienced a lower rate of mortality. Similarly, it was hypothesized that parents who are of the opinion that mortality rates among infants and children have been coming down in recent years (i.e., that survival prospects for children have been improving) would feel less reason to "hedge" against future deaths among their children and hence would be less likely to want more children. The expected signs for both of these mortality-related variables are negative. In fact, with the exception of women in Ll, survival rates (at least based on births and deaths reported by women and births and surviving children reported by men) are higher on average for those who want more children than for those who don't want more children. This may be in part because both average age and average number of surviving children are consistently higher for those who don't want more children; hence they would have experienced relatively greater cumulative risk--both in terms of numbers of children and years--than those who want more children. Among men, the sign for the mortality- opinion variable is consistently positive. It is also positive for women in M1 but negative for women in M2 and L2 and doesn't enter the 170 function in L1. Thus, there appears to be no support for the "hedging” hypothesis in these data. It was expected that the desire for more children would be negatively related to the number of surviving children. This was consistently the case, and generally the coefficients are relatively large. It was also hypothesized that the desire for more children would be negatively related to respondent's age. This is the case except for women in L1. Finally, it was hypothesized that for women the desire for more children would be positively related to age at marriage. It is in M1, M2, and L1 but the sign is negative in L2. In sum, the hypotheses for proxies of income, preferences and mortality-related responses are not supported. The hypotheses for the price proxies do appear to be supported (except for the variable for women's wage-paying jobs) and the results for the variables on opinions about advantages of large and small families respectively are particularly interesting. Independent variables which most con- sistently provided relatively good discriminating capabilities in the functions for women were number of surviving children and advantages to a small family and--somewhat less frequently--percent of children surviving and woman's current age. Variables which most consistently made very little contribution to the woman functions were value of crops produced, building quality index and woman's education. There was less consistency as to strong and weak variables in the functions for men. The most frequent strong variable was advantages to a large family, with wage-paying jobs, man's age, and number of surviving children entering less consistently as strong variables. Man's 171 education was the variable which most consistently made a weak contri- bution to the functions. The indicators of the discriminating power of the functions (given in the lower portions of Tables 34 and 35) suggest that while the level of discrimination is generally not high, these independent variables do pennit some discrimination between the two groups of responses. The discriminating power is greatest for men in the two Lushoto areas and for women in Ml. There are very few Lushoto men who want no more children, and apparently the characteristics of these men differ relatively more from characteristics of those wanting more children than is the case for Moshi men. The differences in the group centroids (for men) are above 1.25 in L1 and about 1.60 in L2 com- pared to differences of about 1 in the Moshi areas. In L2 the R25 are around .35, compared to R25 of about .2 in the Moshi areas and .22 to .29 in L1. For the women, the addition of current age and age at marriage (Analysis 2 in Tables 32 and 34) substantially increases the discrimi- nating power of the functions in all areas but L2. Current age was included in both analyses for men but number of surviving children was included only in Analysis 2. Inclusion of number of surviving children in the functions for men increased the distance between the group centroids and the size of R2 only slightly in M1 and M2 and actually resulted in a reduction in the number of cases correctly classified. Addition of information on number of surviving children substantially improved the discriminating power of the function for L1 and L2, however, especially for L1, as measured by the increases in 172 the percentages of cases correctly classified and the relative increases in all indicators given in the table. In summary, the results of the discriminant analysis are perhaps not as striking as might have been hoped, but they are interesting and are generally consistent with both the results of the regression analyses discussed earlier and with the main hypothesis of this thesis. They suggest that at least for most of the proxies of demand characteristics used in this study and for the stages of social and economic development and transition represented by these four rural areas, there is not much support for most of the generally hypothesized relationships between desired number of children and conventional measures of income, preferences and mortality-related experiences and perceptions. There does seem to be support for the hypothesis that the desired number of children is negatively related to the perceived relative prices of children, at least based on the somewhat novel proxies for relative prices of children used in this study. The next chapter will attempt to draw some conclusions about the effects of rural development on fertility, at least in these four study areas, as implied by the results of the analyses performed in Chapters V through VII. CHAPTER VIII EFFECTS OF RURAL DEVELOPMENT ON FERTILITY Data were presented in Chapter IV on a number of indicators of development in the study areas as of 1973 together with some discussion of the extent to which social and economic development occurred in the study areas between about 1900 and 1973. Further data on social, cul- tural and fertility characteristics and changes therein in the study areas were presented in Chapters V and VI. The general conclusion was that considerable change and development has occurred in all four areas ~during this century, but that there are now substantial differentials both among and within the study areas in the extent to which individual families have participated in and benefitted from the changes. By most criteria, M2 is the area which has experienced the greatest amount of socioeconomic change as well as the area which apparently has the least unequal distribution of economic and educational improvements. L1 is the area which has experienced the least amount of economic and educa- tional gains. The major period of rapid economic gains in the Moshi areas probably occurred prior to the 19605, and this is especially apparent in the levels of past expenditures on improved housing. This contrasts to the pattern of rising incomes in L2 (and to a lesser extent L1) which by some criteria has the highest level of current average income per household among the four areas. Economic gains in 173 174 L2 have been concentrated much more within the last 10 years and especially the last few years. Hence, levels Of wealth appear to be considerably lower in L2 than in the Moshi areas; they are of course lowest of all in L1. There has been a similar pattern in the expansion of educational Opportunities for women in recent decades as well as in the demise Of polygyny, the reduction in the average length of breast- feeding, the utilization of medical facilities, and to some extent, an increase in the age of marriage. The approach in this chapter will be to draw on the analyses and results in Chapters V through VII to discuss the effects of some aspects of development on fertility, first through develOpment-caused changes in supply variables, then through changes in characteristics affecting demand for children, and finally to comment on possible net effects of the various components of development on fertility. The Impact of Rural Development on Supply Variables and Fertility The supply variables analyzed in Chapters V and VI were woman's current age, her age at marriage, number of years married, whether or not the woman was married to a polygynous husband, length of lactation, and her number of deceased children. In most cases, there were high levels of statistical significance between each Of these variables and number of children born. Current age, years married, and number of deceased children were positively related to fertility, while age at marriage, polygynous marriages and length of lactation were negatively 175 related to fertility. Woman's age is probably the only variable unaf- fected by develOpment.1 Age at Marriage 0f the independent variables in the supply model, the one which has the biggest overall effect on number Of births (excluding woman's age) is age at first marriage. The effect of development on age at marriage is not entirely unambiguous. In the model in Chapter VI, BLDQUAL--assumed to be a proxy for household wealth--was generally nega- tively related to age at marriage implying that the independent effect of rising incomes over a long period of time would be a tendency for earlier marriage. However, other aspects of development may have the opposite effect on age at first marriage, particularly woman's educa- tion. Results showed that in both M2 and L1 the relationship between the woman's amount of formal education and her age at marriage is posi- tive, fairly strong and statistically significant. Women completing primary school married on average about two years later in M2 and 5 years later in L1 than women with no schooling. This implies a decline of about half a birth per woman in M2 and about one birth in L1, even net of the offsetting effect of shorter breastfeeding among educated women. However, there was essentially no relationship between education and age at marriage in M1 and L2. 1Even age is likely to be affected by develOpment if one conse- quence of develOpment is declining mortality. In particular, to the extent that female adult mortality declines, this will increase the aver- age number (and total number) Of childbearing years of a group of women passing through those years. A higher proportion of those who reach age 15, for example, will also reach age 45. Hence, ceteris paribus, the average number of children eventually born to that group of women will be higher than would be the case under conditions of higher adult female mortality. 176 The evidence indicates quite clearly that age at marriage has risen considerably in recent years in M2. M2 has the highest overall levels of educational achievement for women, and the biggest gains for women have been rather recent, apparently paralleling rises in average age at marriage. M2 is also the area which--based on most indicators of development available from this study--has experienced the most overall development while at the same time achieving probably the least unequal distribution of the benefits of development. Number of Deceased Children Improved living conditions, education, greater access to health services, and better health care can all be expected to contribute to general declines in mortality. The only variable in the regression model of survival rates among children (in Chapter VI) which was statistically significant was whether or not a woman's last baby was born in a hospi- tal (as a proxy for quality of health care received by the woman and her children). It was concluded that luck was probably the most impor- tant cause of differential child mortality among families. However, it is quite possible that overall mortality rates could decline in a com- munity with luck still playing a dominant (although not exclusive) role in determining which children die and which survive. Mortality rates have certainly declined substantially in recent decades in all four areas. Analysis of reported age distributions pro- duced estimates of female expectations of life at birth of 50 to 55 years in all four areas. These are exceptionally high levels for rural Africa. Crude death rates are estimated to be in the range of about 12 to 15 per thousand; mortality rates have probably declined by 50 percent or more during this century. Reductions in mortality rates for 177 infants and young children have certainly had the independent effect of reducing fertility somewhat, although it is not possible to estimate very precisely how much. The survival rates implied by the reported number of children born and number deceased are not helpful since they show that mortality rates are if anything higher among children of younger women than among children of women in their forties and Older. This is certainly not the case and indicates, if nothing else, that older women failed to mention a substantial proportion of their deceased children. To the extent that mortality rates for infants and young child- rent continue to fall as the result of on-going improvements in health and living conditions, the independent effect will be to lower fertility somewhat.2 However, while declining mortality results in somewhat lower fertility, it also results in a somewhat higher average number of sur- viving children. For example, a coefficient of 0.7 for NDIED implies that for each child that does not die, the number of births will be reduced by 0.7; it also implies that the number of surviving children for that woman will be increased by 0.3, ceteris paribus. Hence, while rising fertility unambiguously implies a rising average number of sur- viving children ceteris paribus, fertility decline caused by declining infant and young child mortality rates also results in a rising average number of surviving children. 2Rising incomes and education and improved living conditions will not necessarily cause infant and young child mortality rates to decline. To the extent that these changes lead to substitution of bottle-feeding1%":breastfeeding--especial1y if it's unhygienic bottle- feeding which it often is in low income areas--mortality rates of young children could rise. This may be the principal explanation of the nega- tive relationship between woman's level of education and the survival rate among her children which was produced by the model in Chapter VI. 178 Duration of Breastfeeding The model in Chapter VI showed that woman's age and BLDQUAL are both positively related to length of breastfeeding while woman's educa- tion is generally negatively related to length of breastfeeding, and that women of husbands who have wage-paying jobs also breastfed for shorter periods than other women. Husband's wage-paying job may be a proxy for income (with which to buy breast-milk substitutes) as well as a proxy for outside or non-traditional influences, including the notion that it is fashionable to wean earlier and use breast-milk substitutes. Although the relationship with BLDQUAL is positive--perhaps because women in households with better housing are also more likely to be older and more traditional, almost all other evidence indicates that the secu- lar trend in these four areas is for substantially shorter average lengths Of lactation. In addition to the negative relationships with education and wage-paying jobs, it is also probably the case that many other types of outside and/or non-traditional influences encourage shorter periods of breastfeeding. Again, the shortest reported breast- feeding--particularly for younger mothers--is in M2 (with M1 a close second and L2 a close third for women under age 30). Moreover, breast- feeding Of 30-35 months, as reported by many Older women, is so long that it seems almost inevitable that average duration of breastfeeding in the future will be considerably less than this. In most cases, the relationship between duration of breastfeed- ing and fertility is negative and statistically significant. Most Of the coefficients imply that the reduction of about a year in breast- feeding is associated with an increase of up to half a birth per woman for women in their 30's. 179 Polygyny In the Moshi areas and in L2 polygyny is negatively related to fertility and statistically significant for women in their thirties.3 Coefficients are -1.1 in M1, -2.5 in M2 and -0.73 in L2. The incidence of polygyny has declined greatly in the Moshi areas in recent years; there were no women ages 20-29 in M2 married to polygynous husbands, and only 7 percent of those ages 30-39 were married to polygynous husbands compared to 19 percent in M1, 29 percent in L1 and 41 percent in L2. Certainly in the Moshi areas the spread of the Christian religion has been a major cause of the decline in polygyny. However, some aspects of development—-particularly education of women--but also non-traditional influences in general contribute to attitudes which view polygynous marriages increasingly less favorably. In any event, the evidence is clear that the incidence of poly- gyny has declined sharply in the Moshi areas and that this has had the independent effect of causing a rise in average number of children born per woman. For example, if women married to monogamous husbands have on average 1.5 more births than women married tO polygynous husbands, and if in the past about one-third of the women were married to polygynous husbands while now there are no polygynous marriages, the net effect is an average increase of one-half birth per woman. In summary, the probable effects Of rural development on the supply variables--at least in these rural areas--is some rise in average age at first marriage, a decline in the average length of lactation, a 3According to Caldwell [1975: 12] most evidence from West Africa also shows that fertility is lower in polygynous than in monogamous marriages. 180 decline in the incidence of polygyny (with relgion probably being, how- ever, the single most important determinant of the rate at which polygy- ny becomes less common), and a net decrease in infant and child mortality although this may be partially Offset by the adverse consequences of unhygienic bottle-feeding of babies. It is really not possible to know what the net effect on fertility will be because this depends largely on the "mix" of the various changes taking place. Rising age at mar- riage_reduces both fertility and number of surviving children, ceteris paribus. Earlier weaning increases both, except to the extent that it causes somewhat higher young child mortality (i.e., due to bottle-feed- ing), in which case average number of surviving children would be lower. Declining incidence of polygyny increases both fertilityand number surviving, ceteris paribus. Finally, reduction in average number of children dying (per woman) reduces fertility but increases average number of surviving children. It seems probable that the net effect of the above changes in supply variables (resulting from development) is that average number of surviving children will rise somewhat, since of the above changes only rising age at marriage definitely reduces number of surviving children. This conclusion--that the net effect of changes in supply variables is likely to be an increase in the average number of surviving children while the net effect on fertility could be change in either direction or no change whatsoever--is of course what was postulated in the theoreti- cal framework developed in Chapter II and incorporated into Figures 3 and 7. 181 Rural Development and Changes in Demand Characteristics Parental demand for children is expected to be positively related to income, negatively related to the relative price of children, and positively related to tastes or preferences which are relatively in favor of children vis-a-vis other consumer goods. Average household incomes have increased in the study areas during the past few decades, although in the Moshi areas they may not have increased during the past 20 years or so. Average incomes have without doubt increased substan- tially in L2 in recent years. It is not certain what has happened to the relative price of children in recent years although it seems likely that the price Of children vis-a-vis other goods has risen, at least in the Moshi areas, and in L2 (at least in comparison to L1). The main price proxies used were respondents' Opinions about whether there were advantages to having large and small families respectively. Over half of the women under age 50 in M1 said there were advantages to having small families; this compares to about 40 percent in M2, 30 percent in L2 and 20 percent in L1. Percentages for men under age 50 were almost 60 in M1, almost 50 in M2 and about 40 in L2 and 35 in L1. It is certainly the case that Moshi residents have spent much more on other consumer goods--particu- larly expensive housing—-than have Lushoto residents, and hence perhaps feel relatively more competition between children and other goods for their financial resources than do Lushoto residents. For similar reasons, it is tempting to speculate that inthe Moshi areas tastes have shifted relatively in favor of other goods vis- a-vis children in recent years. Most (perhaps all) of the very substan- tial rise in average household income occurred prior to the 19605 in 182 the Moshi areas, while the big jump in average household income has occurred only within the last few years in L2, and average income is still much lower in L1 than in the other three areas. This earlier increase in incomes in the Moshi areas means that residents there have been earning and spending more money for a longer period of time (then have Lushoto residents) and have experienced the acquisition of certain modern consumer goods-—especially good h0using--by many of their neigh- bors and relatives, if not themselves. Aspirations to acquire non- tradition goods have had more time and opportunity to develop in the Moshi areas. These include aspirations for (and by) children (such as for education and better clothing) which increase their price and (eventually at least) make the price of children positive and probably continuously rising. These changes have probably also affected aspira- tions for non-traditional consumption goods which compete with children for the household's financial resources. Put another way, reference group standards have for some time been rising and are undoubtedly now much higher inthe Moshi areas than in the Lushoto areas (at least L1). Not all households have fine homes and/or children with secondary education in the Moshi areas, but enough do that these seem potentially within the reach of nearly all families. This has not been the case in the Lushoto areas, although the relatively high incomes which have recently started to accrue to some residents of L2 may soon result in increases in types of expenditures similar to those made earlier inthe Moshi areas. Hence, incomes have certainly risen in recent decades in all four areas, although much less so in L1 than in the other three areas, and among the other three areas the timing and distribution (within the 183 areas) of the rises, and the effects on permanent incomes (wealth) are certainly different. Probably both relative and absolute prices of children have risen, for at least some parents in the Moshi areas, although this conclusion is more speculative than that for incomes. It is even less certain, however, what has happened to preferences for children vis-a-vis other goods, although it seems probable that in the Moshi areas preferences may have shifted relatively in favor of non- child goods vis-a-vis children. Thus, in the Moshi areas (and perhaps to a lesser extent, L2) incomes have risen, the relative and absolute prices of children have probably risen, and if preferences have changed, it is likely they have shifted away from children. Rising incomes should increase the demand for children while the latter two effects should reduce demand for children. There is certainly evidence that demand for children is present and operative among parents in these areas. Data presented in Table 26 (Chapter VII) on proportions of parents who said they want more child- rent and proportions who say they don't want more make sense in several ways. First, for both mothers and fathers the proportions who say they don't want more children are very small among those with relatively few surviving children, and the porportions wanting no more rise systema- tically (with a few interruptions in some groups) as the number of sur- viving children increases. (Only parents presumably still capable Of having children are included in the analysis.) For women and husbands, respectively, the percentages with three surviving children who said they want more children are 71 and 100 in M1, 96 and 86 in M2, 92 and 91 in L1 and 96 and 100 in L2. This compares with the following 184 percentages for those with more than seven surviving children who said they want more: 12 and 48 in M1, 43 and 41 in M2, 60 and 63 in L1 and 62 and 64 in L2. Second, as might be expected, in most cases for a given number Of surviving children a higher proportion of women than men say they want no more children. Third, L1 is usually the area with the highest percentages of both men and women who say they want more children. All the data available on development indicate that L1 is the area in which the least amount of development has taken place. It is therefore to be expected that residents of L1 would hold relatively more traditional values than residents of the other areas, including the traditional value of wanting as many children as possible.4 Finally, the results of the discriminant analysis presented in Chapter VII show a high level Of consistency between both respondents who said they wanted no more children and those who said they wanted more and their respective responses to questions as to whether they thought there were advantages to large families and advantages to small families. That is, a much higher percentage of respondents who want more children thought there were advantages to large families than for those who want no more children. Similarly, a much higher proportion Of those who want no more children thought there were advantages to small families as compared to those who want more children. 4The converse does not hold however. By most indicators, M2 is the most highly developed of the four areas yet in most cases M1 has the highest proportions of both mothers and fathers who want no more children, with M2 and L2 between M1 and L1. 185 Altogether this seems to be fairly strong evidence that demand for children (as measured by whether parents want more children) is present for individual parents, and that within a given community, a given age group and for parents with a similar number of surviving children, there are some who do and others who don't want more children. Thus, the patterns of responses to the question, "do you want more children?" and the relationships between these responses and the proxy variables used for relative prices iniflwediscriminant analysis model in Chapter VII were all quite plausible and in general what would be expected. However, analysis of the relationships between the responses to this question and both income and taste proxies was much less satisfactory. It was expected that both current income and perma- nent income (wealth) would be positively related to the demand for children. However, the relationships were in some cases positive and in others negative with no consistent patterns. When demand proxies were included in regression models Of fertility, only BLDQUAL (proxy for permanent income) consistently had the expected positive sign and only in a few formulations was it statistically significant, and the coeffi- cient was invariably small. Inclusion of proxy variables for relative preferences in both the regression and the discriminant models resulted in no significant or even consistent results. Inclusion--in the discriminant analysis model of demand--Of the responses of parents to actual or anticipated mortality among their children produced, if anything, results opposite to what were hypothesized. This is in contrast to the findings Of some other studies which indicated that parents seem to respond to actual or expected mortality among their children by increasing the number of 186 children wanted, as a hedge against future deaths. This was implied, for example, by a study of Taiwan [T.P. Schultz, 1971] and a study in the Philippines [Harmon, 1970] although in these studies the responses to mortality were not formulated in a strictly demand context as they were here. In summary, results of analysis of demand for children are much less satisfactory than results of analysis of supply;5 There is evidence that differential demand for children is present among parents and that they can express it in terms Of whether or not they want more children. It is also certain that incomes have risen in recent decades; it is likely that relative prices of children have risen; and it is possible that preferences have shifted somewhat away from children. Theory indicates that rising incomes should increase demand for children while the price and preference changes should reduce demand for children. The higher proportions Of both mothers and fathers who want no more child- ren in L2 and the Moshi areas compared to L1 are consistent with some 5This is in contrast to the conclusion of Snyder [1974: 625] that the results Of his regression analysis of data from Sierra Leone supported the demand theory of the determinants of fertility. While it is true that the coefficients of his proxies for income (husband's education) and prices (wife's education) had the expected signs, the t-ratios were small and most were not statistically significant even at the .10 level. In the same model he had the "supply" variables of woman's current age, her age at the birth of her first child, and the mortality experience among her children. For all age groups almost without exception, the t-ratios for all three of these variables were considerably larger than those for the demand proxies, and they no doubt accounted for much of the R25 (which ranged from .16 to .49). It appears that from his results one could justifiably conclude that supply characteristics are more important that demand characteristics in determining fertility in his sample of Sierra Leonean women. 187 decline in demand for children in recent years, but the evidence is far from conclusive. SummaryObservations Thus, the most likely effects Of development on supply of and demand for children would appear to be as follows: Changes in the supply variables have probably on balance resulted in some increase in the average number Of surviving children per woman although the net effect on fertility is unknown. Changes in demand variables have pro- bably had the net effect of reducing the desired number Of surviving children. There is no evidence in these data of a strong positive relationship between incomes and the demand for children. There is evidence of a negative relationship between the price of children and the demand for children, at least as the price of children is perceived by mothers and fathers, and there is evidence of probable increases in the proportions Of parents wanting no more children in both Moshi areas and, to a lesser extent, in L2. It seems likely that if preferences have changed, they would have shifted in favor Of other goods vis-a-vis children, although there is no real evidence that they have done 50. Hence, absence of much relationship between incomes and demand for children means a weak or non-existent income effect, even though incomes have been rising in the last few decades in the study areas. On the other hand, the apparently stronger negative relationship between price of children and demand for children suggests that on balance, if demand for children has been changing in response to development, it has been declining. 188 The theoretical formulation presented in Chapter II postulated that in response to develOpment, supply would rise and demand would fall although the timing and rates of changes might differ. This seems to be what has been happening in these areas. It was also postulated that when a mother's (or father's) number Of surviving children (supply) equaled or exceeded the number desired (demand), the person would pre- fer to have no more children and would indicate so in response to the appropriate question. Certainly the patterns in the responses to this question are consistent with this expectation. If the responses are valid, it implies that a substantial number--alth0ugh in most cases not a majority--Of parents with more than about four surviving children are at the point where supply equals or exceeds demand. The patterns further suggest that a smaller proportion are in an excess supply situation in L1 than in the other three areas. All of these results are consistent with the theory. It was postulated further in Chapter II that because of the do- minant role of supply variables in determining actual fertility, parents and rural communities generally would probably experience a considerable period of excess supply before demand variables would become important enough to motivate parents to actually do something about preventing future births--e.g., practice contraception or abortion. Again, although questions about practice of contraception or abortion were not put to the respondents, there is no evidence from non-survey sources that either are found in these areas. In fact, modern contraceptives were not available to residents of these areas, and abortion is consi- dered taboo. It is not known whether any respondents have experienced sufficient socioeconomic transition and have sufficient excess supply of 189 children that they would actually want to use modern contraceptives if they were readily available (and hence, the person would then be in Stage 4 of the transition model in Figure 3). However, the analysis strongly suggests that a sizable proportion of respondents have exper- ienced some excess supply and are in Stage 3 of the theoretical frame- work shown in Figure 3. CHAPTER IX SUMMARY AND IMPLICATIONS Summary The received thoery of the determinants of fertility, and espe- cailly the theory as refined by the so-called Chicago school of econo- mists, assigns a dominant role to demand variables--incomes, prices, preferences, and the value Of parents' time (as it affects their opportunity costs in bearing and rearing children)--in determining fertility. In contrast to received theory, the reformulation of the theory of the determinants of fertility in Chapter II assigns a dominant role to supply variables in the early stages of development while demand characteristics dominate only in late stages of development and in high income, modern societies. This reformulation postulates that socioeco- nomic transition affects fertility through a rather complex sequence of changes which initially cause the average number of surviving children (supply) to rise, somewhat later bring about a decline in the average number Of children desired (demand) and eventually these two sets of changes cause parents in their later years of childbearing to find them- selves in excess supply rather than excess demand situations. Continued development causes relative prices and preferences to shift further against children vis-a-vis other consumer goods and the absolute amount of excess supply of children (per parent) typically continues to increase until parents are motivated to do something about it--meaning 190 191 practice contraception or abortion to try to prevent further births. The main purpose of this study was to test that part of the reformula- tion of the theory of the determinants of fertility which postulates that in traditional and early transitional stages of socioeconomic deve- lopment in rural societies, supply variables are more important than demand variables in determining fertility. These four rural areas were chosen for the study (1) because they were among the sample areas of the 1973 National Demographic Survey of Tanzania and (2) because of both similarities and diversities in their cultural, demographic, agricultural and other economic characteristics. The data for this study were obtained by interviewing adult members of about 1500 households in these four areas during the latter part of 1973. At the end of the 19th century these areas were still socially and economically traditional societies. Birth and death rates were high, polygyny was common, and Christian missionaries, European settlers and a foreign (German) administrative authority had only recently arrived in these areas and had as yet made no real impact on their health, educa- tion, religion, agriculture or other socioeconomic characteristics. While these were not completely egalitarian societies prior to European influence--at least in terms Of social status and political authority, differentials in material well-being of residents both within and among these areas were probably small [Feierman, 1974; Iliffe, 1971]. Since the beginning of this century, death rates in the study areas have declined by over 50 percent and the populations have increased four to five times in size. A host of major social and economic changes have occurred and have resulted in large differentials in the social and material well-being of the residents, both within and among the four 192 areas. These differentials include access to educational Opportunities, utilization of health care facilities, participation in commercialized agriculture and in the non-agricultural wage labor force, and the acquisition of non-traditional forms of wealth including good housing and modern consumer goods. These differentials have been in part caused by and in part have caused accompanying social and cultural changes and differentials. These cultural and social changes include religion, the incidence Of polygyny, age of marriage of women, and breastfeeding prac- tices. The theory of the determinants of fertility as reformulated in Chapter II posutlate that the socioeconomic changes which have occurred should have also affected fertility, and that the socioeconomic differ- entials which have emerged should also result in fertility differentials. These expectations are confirmed in the analysis of the determinants of fertility in Chapters V through VIII. The evidence shows that, as expected, supply variables dominate while demand variables have as yet had very little to do with levels, differentials and changes in ferti- lity in these rural areas. This dominant role assigned to supply varia- bles in determining fertility is in contrast to a large and growing body of literature coming from the Chicago school of economists who have concentrated their efforts on improving theory and on developing metho- dology for measuring the determinants of demand for children while largely ignoring supply variables. Since this analysis indicates that it is supply and not demand variables which primarily determine fertility at least in the traditional and early transitional stages of development represented by these four rural areas of Tanzania, further refinements of demand theory by Chicago school economists hold little promise for 193 adding to knowledge about the relationship between socioeconomic and demographic transition in societies at these stages of development and modernization. The analyses here show that demand determinants also change as a consequence of socioeconomic development but that the expected posi- tive relationship between income and desired number of children is weak or non-existent. Most of the evidence indicates that development pro- bably causes relative prices--and perhaps preferences--to shift in favor of other goods vis-a-vis children. Thus, the probable net effect of development on demand for children is to cause desired number of children to decline. The net effect of development on number of surviving children in the rural areas analyzed here is probably some increase although the net effect on fertility could be either increase, decrease or no change depending on the mix and amount of changes in the various supply varia- bles. Hence, results of the data analyses with regard to both supply and demand variables are for the most part encouraging and support this reformulation of the theory Of the determinants of fertility. Poligy Implications Two types of policies can be distinguished. One is those which enhance the prospects and process of develOpment, including all three development criteria identified at the beginning of Chapter IV. The second is population policies, in particular those which are specifi- cally intended to speed the onset and subsequent rate of fertility decline (e.g., family planning programs). The first set (above) encompasses a whole host of policies, including education, health, and social services and amenities generally, ii 1.} II! I III! I i l 194 incomes, prices, employment, fiscal, foreign exchange, urban development, geographic and ethnic priorities, research, and many others--all of which impinge in important ways on the process of development in rural areas, including the distribution of development benefits within and among rural areas. All of these policies can together be viewed as being also, in part, a population policy to the extent that they bring changes in rural areas which in turn cause changes in mortality, fertility and migration. When the goal is fertility reduction, it is not necessarily helpful to distinguish between those policies which affect demographic variables through the intermediaries of education, health, incomes, con- sumption, etc., from more narrowly construed population policies--such as family planning programs--which presumably affect fertility more directly. The remaining brief comments will refer to aspects of both types of pOpulation policies. Analysis of data in this thesis indicates quite clearly that at least for current and earlier stages of development in these areas, policies which affect determinants of the supply of children potentially have more immediate effect on fertility than policies intended to affect demand for children. That is, policies which bring about rising age at first marriage, discourage early weaning and reduce infant and young child mortality rates will all contribute to lower fertility. Presum- ably education--especially for women--and health care which reaches essentially the entire pOpulation (again, especially women and child- ren) would be likely to have the greatest favorable effect on these three supply variables. But educational content is perhaps as important as school enrollment itself. Curricula for upper level primary school 195 students should include instruction on the value of prolonged breast- feeding, on simple and effective family health care, and on family nutri~ tion, particularly maternal and young child. Moreover, at least in some areas increased education is likely to cause some offsetting rise in fertility, even with proper educational content, because it seems proba- ble that better educated women will be less enthusiastic about being part of polygynous marriages, and to the extent that the incidence of polygyny declines as a consequence of increased education of women and other changes taking place in these rural areas, this will probably have the independent effect of some rise in fertility. Despite this, the demise of polygyny would presumably represent an enhancement of the status and roles of women--which would contribute to the long term process of development and fertility decline--and hence is desirable in its own right. Policies whose objective is to reduce fertility by causing changes in supply variables should be considered essentially short run in nature and probably have the characteristic of being "one-shot" in that there are limits to the extent to which marriage age is likely to rise, mortality to fall and weaning to be delayed, and these limits are such that the maximum resulting decline in fertility is rather modest (e.g., in the range of 10 to 15 percent). This means that any long run, sustained and substantial fertility decline--e.g., a reduction of half in average number born--can only come about through reducing demand for children accompanied or followed by parental use of effective means for more nearly matching actual number born with number desired--e.g., the effective use of modern contraceptives. 196 The analysis in this thesis indicates that demand for children apparently does decline as the consequence of development, although desired number of surviving children probably now averages about 5 or 6 even in the Moshi areas. But the effect of development on demand and eventually on fertility is almost certainly a long run effect—-at least in comparison to the effect of development on supply. These data sug- gest a lack of direct relationship between income and desired number of children. But they indicate that higher incomes are associated with changing perceptions of the advantages of large and small families which are used here as proxies for relative prices. The results suggest that rising incomes and development generally cause relative prices and per- haps preferences to shift against children (at least, against quantity of children desired). That is, the analysis indicates that development does cause demand for children to decline, but that it probably takes a rather long time for this decline to result in parental action to prevent further births. Policies which raise educational levels, raise incomes and improve access to consumer goods which compete with children for parents' time and income will all speed the decline in demand for children. These results reinforce the view that development is a very long term process and that the time required for development to cause sus- tained fertility decline is probably at least several decades and neces- sarily encompasses at least two or three generations. Development is complex and multifaceted, and it does effect determinants of fertility-- primarily supply variables initially, and later and more importantly in terms of long run fertility decline, demand variables. These results suggest that we must more carefully specify the stage of socioeconomic 197 and demographic change or transition which the target population is in before deciding what types of policy interventions may be most suitable. Can this long term process be circumvented or speeded up by introducing into these communities programs which provide family plan- ning services and promote the values associated with their use? The tentative answer suggested by this analysis is, "perhaps yes, but only to a limited extent and the timing has to be right." It appears that some development would have to take place before changes in supply of and demand for children are sufficient to cause a sizable proportion (e.g., one-fourth to one-third) of parents in later childbearing ages to experience excess supply and want no more children. Family planning programs introduced into rural areas which have experienced little or no socioeconomic transition will probably find little demand for their ser- vices and, because their purposes run counter to traditional values, they may generate an adverse response and perhaps be counterproductive by making it more difficult to successfully promote their values at a later, more suitable,time. On the other hand, once a considerable amount of socioeconomic transition and develOpment--which is likely to be causing supply to rise somewhat and demand to decline--has taken place in a rural area, there may very well be substantial and increasing numbers of parents--particu- larly women-—who could be persuaded to use contraceptives. A well designed and implemented program which results in a satisfied first group of participants might find the demand for services increasing as a growing proportion of women with excess supply (of children) want to prevent additional births and are persuaded by the success other women are having with family planning. 198 It seems unlikely that a family planning program could create feelings of excess supply of children. It seems quite possible that a well designed and implemented program could exploit existing feelings of excess supply. Such a program might very well now find acceptors inthe Moshi areas, and perhaps even in L2. It seems unlikely that there would be much market for services in L1. However, even a program promoting the advantages of family plan- ning for proper child spacing—-as Tanzania'a family planning program now does--would have the salutary effect of increasing birth intervals over what they would otherwise be, thereby causing lower fertility. If maternal and child health care services which successfully promote the values of breastfeeding (and discourage bottle-feeding), later weaning and better maternal and child nutrition practices are coupled with these services, this will have the further salutary effect of reducing infant and young child mortality--and thus reducing fertility--while increasing the average number of surviving children. This in turn would cause parents to reach the point of excess supply of children relatively earlier, thereby making it more likely that they would eventually want to use contraceptives to prevent future births. APPENDICES APPENDIX 1 DESCRIPTION OF DATA COLLECTION AND PROCESSING APPENDIX 1 DESCRIPTION OF DATA COLLECTION AND PROCESSING Origins of the Project . This study was initiated in late 1972 when I was a Research Fellow in the Bureau of Resource Assessment and Land Use Planning (BRALUP) at the University of Dar es Salaam and a staff member of the Population Council. The (then) Ministry of Economic Affairs and Development Planning of the Government of Tanzania wanted to obtain updated demographic estimates for the country; the most recent demo- graphic data available at that time were from the 1967 census and there were no prospects for obtaining new estimates until the early 19805 when the results of the next census would be available. The Ministry had instructed its Bureau of Statistics to investigate the possibility of conducting a national demographic survey. Mr. J.J. Mpogolo, Commissioner of Statistics, had worked closely on other pro- jects with the Director of BRALUP, Professor A.C. Mascarenhas, and with Professor R.A. Henin, Director of the Demographic Unit in BRALUP and the Population Council's Resident Representative for Eastern Africa. Professor Henin was particularly interested in the opportun- ity to further investigate fertility differentials in Tanzania; his analysis of the 1967 census data had shown the probable existence of substantial fertility differences among large parts of the population. It was agreed by Mr. Mpogolo and Professors Mascarenhas and Henin that 199 200 efforts should be made for BRALUP and the Bureau of Statistics to jointly undertake a 1973 National Demographic Survey. The Government of Tanzania was prepared to make available substantial resources. The International Development Research Centre (IDRC) of Canada subsequently provided a sizeable grant to BRALUP to augment staff, computer and other resources available at the University of Dar es Salaam for carry- ing out the survey and analysis of the data. The Population Council was also very supportive of the undertaking and was prepared to provide consultative and technical support in addition to the existing large grant to BRALUP in partial support of salaries, physical facilities and equipment and on-going research and training activities. It was agreed that the 1973 National Demographic Survey (NDS) should have the two objectives of (l) obtaining demographic estimates for regional and national levels and (2) providing the Opportunity to study fertil- ity levels, trends and differentials in Tanzania. The first half of 1973 was spent in organizing the demographic survey. Field work commenced in August and was completed in December. About 65,000 households were included in the NOS, about 45,000 of which were in rural areas and 20,000 in urban areas. All of 1974 and the first half of 1975 were spent in data processing; the second half of 1975 and the first part of 1976 were spent in data analysis, and the 1 first substantive results were published in Dar es Salaam in 1976. As a staff member from August 1972 to August 1975 of both BRALUP 1These include volumes on statistics for regions and the mainland (Vol. 1), tables for socio-economic groups, modes of life and tribes (Vol. II), data for individual clusters (Vol. III), description of the methodology (Vol. IV) and the Training Manual for interviewers and super- visors (Vol. V). 201 and the Population Council, I was involved in all phases of the NOS commencing with initial discussions between BRALUP and the Bureau of Statistics in late 1972. I was one of four senior BRALUP staff members who worked on the project; in addition to Professor Henin and myself, Mr. I.D. Thomas worked full-time on the NOS until June 1974. Dr. D.C. Ewbank joined the project in November 1974. I was particularly interested in studying relationships between socioeconomic and fertility and morality changes in rural areas. The occasion of the National Demographic Survey seemed to be a unique oppor- tunity to investigate some Of these relationships in a few selected rural areas of Tanzania. With encouragement from Professors Mascaren- has and Henin, I prepared a proposal to undertake a study in Kiliman- jaro and Tanga regions in Northeastern Tanzania in conjunction with the N05. The Population Council subsequently provided a grant covering the period January 1974 through July 1975 to supplement BRALUP re- sources used in this project. The project was officially titled, "A Study of Rural Development and Demographic Change in Four Densely Settled Areas of Moshi and Lushoto Districts" (hereafter referred to as the Study). It was agreed that my time should be divided about equally between the Study and the NOS. Two principal sources of data were used for the Study; one source was data collected by the NOS questionnaire. The second source was a questionnaire designed to gather more detailed information on economic, educational, social, cultural and health characteristics of households and individuals in the study areas; this was called the Socioeconomic questionnaire. Both questionnaires were originally drafted in English and subsequently translated into Swahili for use in the field. (English 202 versions of the two questionnaires are included as Appendix 2.) Organization of Data Collection Sample Selection Because the Study was carried out in conjunction with the N08 and depended on the N03 for some of the data, sample selection for the Study was limited to the sample for the NOS. Due to the dual objectives of the NDS--i.e., to obtain regional and national demographic estimates and to study fertility trends and differentials, it was decided that sampling for the NOS should provide clusters in rural areas which would be large enough to permit investigation of fertility trends and differ- entials within and among these rural clusters. Earlier work by Profes- sor Henin and others had shown that at least 1000 women age 12 or older were required to adequately study fertility differentials by age groups, so it was decided that each rural cluster selected for the study of fertility trends and differentials should include this number Of women. It was further decided that a minimum of four rural clusters would be required for each region in order to provide adequate regional demogra- phic estimates. Available resources were insufficient to permit this number of interviews to be carried out in each of Tanzania's (then) 18 rural regions. It was thus decided to select eight rural regions, each of which would have four "large" clusters (about 900 households) for the study of fertility trends and differentials, while the remaining ten regions would each have four "small" clusters of approximately half that size, and the objective in these latter ten regions would be limited to Obtaining basic demographic estimates. The eight regions with large clusters were selected because they represented a variety of important 203 geographic, social, cultural, economic and agricultural characteris- tics.2 Random numbers were used to select four Enumeration Areas (used for the 1967 census) in each region, each Of which became the locus for one of the four clusters (for a region). (An average of about 500 people resided in each Enumeration Area in 1967.) Clusters were built up around each Enumeration Area (EA) by including adjacent Enumeration Areas located within ever larger concentric circles which had as their approximate midpoint the geographic center of the randomly selected EA, until the targeted number of households for the cluster was ob- tained. Since these were only estimates of household numbers based on the 1967 census maps, actual household lists were later constructed during preparatory visits to each cluster prior to the field work. Kilimanjaro and Tanga were two of the eight regions selected to have large clusters. Two of the four clusters in each of these two regions were located in densely settled highlands areas. In Kiliman- jaro region the two highlands clusters were in Moshi district while in Tanga region the two were located in Lushoto district. These four clusters were selected primarily because of their cultural and economic differences and basically similar ecological, agricultural and demo- graphic characteristics. The four were also close enough to each other to meet constraints of field work logistics and to enable me to supervise the work of both the NOS team and the team conducting the socioeconomic interviews. 2A discussion of the basis for selecting these eight regions is given in 1973 National Demographic Survey of Tanzania, Vol. IV: Survey Methodology,71976 (by Henin, Kocher, Ewbank and Thomas). 204 About 900 households were included in the NOS sample in each of the four clusters. It was decided to include in the Study every second household in each Of the two Moshi districts and every third household in each of the two Lushoto districts; that is, the target number of house- holds for the Study was to be 450 in each of the two Moshi clusters and 300 in each of the two Lushoto clusters. These would then represent about one percent of the total number of households in each district. As it turned out, 290 and 300 households respectively from the Lushoto clus- ters and 442 and 460 households from the Moshi clusters were included in the eventual data set. Appendix Table 1 lists the numbers of households, ever-married women, never-married women age 12 or older, husbands of currently married women, and surviving children age 7 or older (regard- less of current residence) of ever-married women who were included in the Study in each of the four areas (clusters). Appendix Table 1. Numbers of Households and Persons Included in the Study. Number of Cases Category Study Areas Districts Grand M1 M2 L1 L2 Moshi Lushoto T°taI (1) (2) (3) (4) (5) (5) (7) (8) Households 442 460 300 290 902 590 1492 Ever-married Women 505 502 312 312 1007 624 1631 Never-married Women age 12 or older 249 325 142 128 574 270 844 Husbands of currently married women 411 432 267 260 843 527 1370 Surviving children age 7 or Older (regardless Of current residence) of ever-married women 1400 1651 850 867 3051 1717 4768 205 Questionnaire Construction and Design The N03 Questionnaire: The main Objective of the NOS question- naire was to obtain information which would permit the estimation of fertility and mortality rates. As such, the central feature of the questionnaire is a detailed pregnancy history for each woman age 12 or older (see Block SC on the lower portion of the second page of the NOS questionnaire in Appendix 2). This provides details of the outcome of each of the woman's pregnancies, including the current status of each child born alive. Information was also collected on deaths within the household during the 12 months preceding the interview (see Block 2 on the first page of the questionnaire). Selected social, cultural and economic information for all women age 12 or older and for all husbands of currently married women were also obtained (Blocks 5A and 58 respectively on the second page). Because of the large NDS sample, it was important that the NOS questionnaire be as parsimonious as possible while Obtaining as much basic information as possible. In its final design, the questionnaire consisted of two pages; the first (Blocks 1 through 4) provided informa- tion on the household. The second page (Block 5) provided information on one woman in the household and her spouse (if she had one). A separate page 2 was used for each woman age 12 or older in a household, so that a questionnaire for a single household frequently required two or more copies of page 2. Questionnaire books of 100 pages each were printed. They consisted of 25 copies of page 1 of the questionnaire, each copy of page 1 was followed by 3 copies of page 2. Additional copies of page 2 were added for those households which had more than three women age 12 or older. 206 Interviews using this questionnaire were conducted by members of the NOS team of interviewers. Each NDS team consisted of about 25 interviewers, 5 supervisors, a team leader and two assistants. The team leader and his or her two assistants were all permanent staff members of the Bureau of Statistics. The interviewers and supervisors were employed only for the duration of the field work. Most of them were recent secondary school leavers who had not yet begun or found permanent employment although a few had not attended secondary school. The N03 team which worked in the Moshi and Lushoto clusters also did all the interviewing in Coast region (another "large" cluster region) as well as the interviewing in the remaining two clusters in both Kili- manjaro and Tanga regions. The senior staff members of BRALUP who worked on the NOS project had overall responsibility for the NOS field work. I was responsible for supervising the work of this particular NDS team in addition to making brief supervisory visits to teams work- ing in central and western Tanzania. (There were a total of five rural and two urban teams.) The Socioeconomic Questionnaire: The purpose of the Socioecon- omic questionnaire was to Obtain additional information on character- istics of households and their members. Detailed information was obtained on the following characteristics: education (including that of children and parents of respondents), selected health practices, household production, consumption and wealth (including housing), attitudes toward selected health practices and social norms (including advantages of large and small numbers of children), beliefs about child survival rates, prenatal, delivery and breastfeeding practices, 207 indicators Of household progressivity, and selected intertemporal (or intergenerational) data. For this questionnaire, all male heads of households were interviewed as well as all ever-married women in each of the households. In its final form, the Socioeconomic questionnaire consisted of two parts, the first of which gathered information on characteristics of the household and of the head of the household. The second part gathered information on characteristics of each ever-married woman in the household; a single four-page section was used for each women (see Appendix 2). Each questionnaire was stapeled across the top so that the inter- viewer could conveniently turn the pages by flipping them up (rather than to the side or across a corner). Each questionnaire was also stapeled to a cardboard sheet. The combination of the cardboard plus the top-stapeling made the questionnaires much more durable and mini- mized the problems of pages tearing off. This can be an especially annoying and serious problem during coding and editing as the question- naires get repeated use. Questionnaires were color-coded by using a different color of cardboard for each of the four areas. This again greatly facilitated organization and filing of questionnaires during Office coding and editing. Constraints Of Political and Cultural Sensitivities Population issues were quite controversial in Tanzania in 1973. The Family Planning Association of Tanzania, a private organization affiliated with the International Planned Parenthood Federation, had been making family planning services available in some urban areas of 208 Tanzania for several years. Some family planning workers together with many health workers both within and outside Government wanted to expand family planning services and extend them to outlying areas as part of general health services. Some others were voicing criti- cism Of family planning, both as doing a disservice to Africans and as an extension of foreign interests. There was heated criticism of family planning in the press; critics sometimes associated it with and at the same time condemned teaching and research on population-related tOpics--together with their sources of funding and resources used (including staff members), particularly at the University of Dar es Salaam. Articles were also in the press defending family planning services for their contribution to the health and well-being of mothers, their children, and ultimately the nation: some also defended teaching and research on population-related topics. At the time the N05 and the Study were being organized, the outcome of this debate was uncertain, but it was prudent for those of us working on these projects to attempt to minimize potential controversy about them. For example, the contribution of improved demographic estimates to sound planning was emphasized. One implication for the Study was that questions should ngt_specifically address issues of contraceptive knowledge, attitudes or practice (the conventional KAP information).3 Questions eliciting information on attitudes about family size and advantages of small and large numbers of children were thought suffi- ciently distinct from family planning topics that they were included 3The validity of responses to KAP-type questions, particularly among rural residents of low income countries (and particularly in Africa), has been questioned by various analysts. See, for example- Hauser (1967). 209 in the questions, and there was no subsequent adverse reaction to having done 50. Field Testing and Evaluation A preliminary version of the Socioeconomic questionnaire was drafted in April and May 1973. This version was tested in the field by conducting about 40 interviews in Moshi district during the first week of June. The interviews were conducted by four University stu- dents (during their long vacation). Evaluation of the completed questionnaires resulted in substantial revision of many questions and deletion of several which appeared not to have a good chance of Obtaining valid responses. A revised version of the questionnaire was then used for about 30 interviews in Lushoto district during the last week of June. Both the N05 and Socioeconomic questionnaires were used in this second field test. Evaluation of the second field test resulted in final changes and production of the final version of the Socioeconomic questionnaire (as given in Appendix 2). Language Tribal languages are used by nearly all people in ordinary home, village and community life in rural areas of Tanzania. There are about 130 different tribes in Tanzania, most with distinct (although in some cases, related) languages. Some tribes even speak different dialects which are mutually unintelligible to speakers of the differ- ent dialects; this is the case for some Chagga dialects (in Moshi district). However, most adults in Tanzania know some Swahili. It is 210 Tanzania's national language. It is the language of instruction in primary schools, and it is the language used in interactions among individuals of different tribes or even among those of the same tribe who speak different dialects. For example, a Chagga whose homeland is in the Rombo area (on the eastern slope of Mt. Kilimanjaro) and one whose homeland is in the Machame area of Moshi district would use Swahili when speaking together because they would probably not under- stand each other's dialects, even though their home areas would be located not more than 100 kilometers apart. Because Swahili is widely known and spoken in Tanzania and in the study areas while usually only people with at least some second- ary education understand much English, it was decided that Swahili should be the language in which both questionnaires were printed and the language used by interviewers in the field. In those relatively few cases in which a respondent's knowledge of Swahili was inadequate, it was necessary for another local person (often the Ten-House Cell Leader) to serve as translator unless the respondent and the inter- viewer happened to be of the same tribe and spoke the same language or dialect. Most of the translation of the two questionnaires from English to Swahili was done by a few students at the University of Dar es Salaam and a few staff members Of the Bureau of Statistics working together with myself and other senior staff members on the project. Translations were submitted for review, corrections and recommenda- tions for changes to a few Swahili experts, including staff of the Institute for Swahili Research at the University of Dar es Salaam. 211 Field Operations Cogperation of Government and Party_0fficials: The Study was a project of BRALUP and the NOS was a joint project of the Government of Tanzania and BRALUP. Because of its Government-sponsorship, NDS staff had access to the cooperation and assistance of Government offic- ials at all levels. In addition, the NOS had the official approval and support of the Tanzania African National Union (TANU) which is the only political party on the mainland. The smallest TANU unit is called the Ten-House Cell, and locally-elected leaders at that level are called Ten-House Cell Leaders. The actual number of households which are members of a single Ten-House Cell may vary from about half a dozen to 25 or more in rural areas and up to 50 or more in urban areas. The Ten-House Cell system is generally well-established throughout the country, particularly in rural areas, and almost with- out exception rural Ten-House Cell Leaders know who all of their households are, and adult members of each household in turn know who their Ten-House Cell Leader is. One individual in each local commun- ity is head of the Ten-House Cell Leaders; that person is the lowest- ranking Official appointed by the Party (TANU). Because the NOS was Officially supported by both Government and Party, the Party issued instructions to officials at all levels to give all possible support to the project. In almost every instance, Ten-House Cell Leaders gave generously of their time in helping to prepare lists of house- holds and adult members, in urging respondents to be cooperative and candid in responding to questions, and in generally making the in- terviewer teams welcome in the local communities. 212 Staff Recruitment: As noted earlier, there were five rural and two urban NDS teams. Each team had a Leader, two Field Assistants, about five Landrover drivers and about 25 interviewers and 5 super- visors. Interviewers and supervisors were nearly all temporary workers employed only for the duration of the field work. Most were recruited somewhere within the areas in which a particular team was to work. Those on the team which worked in Kilimanjaro, Tanga and Coast regions were recruited in Kilimanjaro region with the assistance Of the Regional Labour Officer in Moshi. Those recruited underwent a week's training near Moshi which included both classroom instruction and practical interviewing experience followed by careful evaluation of the interviewing experiences. The training was in Swahili and was lead by Mr. S. Ngallaba, Assistant Commissioner of Statistics and one of the team leaders (in another part of the countrY); Mr. Ngallaba had previously completed an M.A. in demography at the University of Dar es Salaam under the supervision of Professor Henin. During the week in which the NOS interviewers and supervisors were trained, interviewers and supervisors for the Study team were re- cruited, again from Kilimanjaro region with the assistance of the Regional Labour Office in Moshi. About 15 potential interviewers were recruited for the Study team; the target was to select eightinter- viewers and one supervisor. A second supervisor was a junior staff member of BRALUP. During the first week of NOS interviewing, the Study interviewer-candidates participated in an intensive week-long training program in Moshi (which I conducted), including both class- room instruction and field practicals. All candidates were first trained as interviewers. At the end of the training session, one was 213 . selected to be a supervisor and was given supplementary training. Appendix 4 presents the l6-page Interviewer's Manual followed by the 4-page Supervisor's Manual. The two had been prepared prior to the training sessions and the training was organized around the Inter- viewer's Manual (which was in some places revised during the train- ing itself as a result of training experiences). Both Manuals were designed to be used as references throughout the field work. At the end of the training period, one supervisor and eight interviewers were hired (one subsequently resigned). The Study team followed the NOS team into each of the four study areas by about one to four weeks. It usually required the NOS team 7 to 10 working days to complete all interviews in each area (about 900 households and about 1500 women over age 12 in each). The NOS team worked in Kilimanjaro, Tanga and Coast regions in that order, covering four areas in each region. The Study team worked in study areas M1, M2 (both in Moshi district in Kilimanjaro region), L2 and L1 (both in Lushoto district in Tanga region) in that order. The Study team worked about 15 to 20 days in each area. During the field work, about half of my time was spent supervising each team. Throughout this period, the two teams were never more than a 3—hour drive from each other. I also made three trips of two to seven days each to other parts Of northern Tanzania to check the progress and work of one urban and two other rural NDS teams. Matching Households and Individuals: A particularly critical requirement for the socioeconomic interviews was that each house- hold (and all individuals interviewed within each household) had to 214 be matched without error with those previously interviewed in the N03. The main procedure adopted to insure this was that each Socioeconomic questionnaire was specifically tailored to a particular household and set of individuals within that household. Prior to the Study team's actually commencing interviewing in an area, certain identifying infor- mation was transferred from the previously completed NDS questionnaires onto the Socioeconomic questionnaires. This information was as follows: name of the Ten-House Cell Leader, name of the head Of the household and the names of each ever-married woman in the household (all of whom were to be interviewed individually), and names and ages of all children Of these women who were 7 years of age or older (as reported in Block 5C of the NOS questionnaire). Further, all NDS questionnaires had previously been numbered, and these numbers were then transferred to the Socioeconomic questionnaires. This enabled interviewers to correctly and easily identify households and individ- uals in the field, and it permitted accurate matching from two separ- ate data sources during data processing. (See Block 1 of both questionnaires.) Time Constraints: For several reasons it was particularly critical that the timing or phasing of the Study be carefully coordin- ated with that of the NOS. It was desirable that both interviewing teams be working during the same time period so that I could reside close to the study areas and spend full time supervising field work rather than commuting the long distance from Dar es Salaam for a period of several months. It was essential that the NOS team conduct their interviews before those of the Study team, but it was desirable that 215 the Study team begin work in an area as soon as possible following the departure of the NOS team. It was also desirable that all field work be completed by November (or before December at latest) so the teams would not be in the field during the rainy season. (This would have caused enormous logistic and morale problems and would have greatly lowered interviewer efficiency.) All of these objectives were accomplished, but it required careful planning and allocation of time and resources at all stages, particularly in drafting, pre- testing and finalizing the Socioeconomic questionnaire, recruitment and training of interviewers and supervisors, and printing and prepar- ation of Socioeconomic questionnaires as well as during the actual fieldwork itself. Data Processing Editing and Coding The Socioeconomic questionnaire was designed to be coded by the interviewer during the interview itself insofar as possible. A little over 75 percent of all responses were actually coded during the interview. Every completed questionnaire was checked in the field by a supervisor. Errors and omissions detected by a supervisor were to be corrected immediately by the interviewer responsible, and when necessary the interviewer returned to the household to get addi- tional information. Some additional coding was also done by the super- visor at that stage. "Open-ended" questions and a few whose answers were particularly complicated to code were left to be coded at a later time. The entire Socioeconomic questionnaire was also designed to 216 function as a coding sheet. Boxes corresponding to the number of digits in each coded answer are located in the left margin Of each page adjacent to the respective questions. The presence of an "I" next to a box or boxes indicates that the question was to be coded by the interviewer during the interview. The number above each box indicates the proper column on a computer card (see Appendix 2). After all questionnaires were completely coded,computer cards were punched by reading directly off the questionnaires. It is believed that this procedure avoided introducing a potentially large number of errors due to transferring data from questionnaires to coding sheets before punching. The NDS questionnaire was not designed to be coded in the field, and it was necessary to establish and supervise a relatively more costly and time-consuming editing and coding operation following com- pletion of the field work. The questionnaires themselves were first hand-checked (edited), and then in a second procedure the responses were transferred to separate coding sheets where they were recorded in coded (numerical) form. Even taking into account the far greater size Of the NOS undertaking, there was a considerably greater problem of errors being introduced during both the editing and coding pro- cesses with the NOS questionnaire than there was with the procedure used for the Socioeconomic questionnaire. Acceptable codes for all questions on the Socioeconomic question- naire which were to be coded in the field by either the interviewer or the supervisor were given in the questionnaire itself next to the cor- responding question, so that the possibility of either the interviewer or supervisor using an unacceptable code was minimized. Codes were 217 not prepared for the open-ended questions until after the field work was completed. The procedure for determining codes for these questions was to review a large portion (about one-quarter to one-half) of all questionnaires and prepare an inventory of a fairly lengthy number of response categories. These categories were then collapsed into the requisite number of major categories, usually determined by the number of columns reserved on the computer card. For example, if one card column had been reserved, seven codes were available for up to seven categories of responses (excluding "doesn't apply," "don't know" and ."notappliicable" which, in addition to the other seven codes, exhausted the ten available for a single card column). Three criteria were used in selecting the maximum of seven major categories of responses: (1) Those types of responses in the "longer“ lists which were basic— ally similar to each other were potentially combinable. (2) Categor- ies of responses with a large number of cases should be preserved. (3) Categories of responses of Special interest should be included, even if the numbers of cases were relatively small. Coding instructions for the Socioeconomic questionnaire are given in Appendix 3. Coding instructions for the NOS questionnaire were prepared by senior staff members of both BRALUP and the Bureau of Statistics working in some cases in small groups and in other cases individually. Use of Computer Programs for Validation (Editing) After the data from both surveys were coded and punched and verified on computer cards, computer programs were written to further validate (or edit) the data. The main functions performed were, 218 (l) to make sure that each code was "acceptable"--i.e., the code was among those that could be correct for that particular variable (al- though Of course not necessarily correct for a particular case), and (2) to make all possible internal consistency checks. The programs caused all variables and cases which either had unacceptable codes or failed a consistency check to be printed. The original questionnaire and/or the computer card was then examined to determine the source of error. A new card with the corrected code(s) was then punched, and after all errors discovered during that run were likewise corrected, the deck of cards was run on the validation program again until even- tually all errors which could be discovered in this way were corrected. This undertaking was useful although time-consuming in both the writing and testing of the validation program and its use. Preparation of Data Files in Tanzania Two types of computer data files were produced on the University of Dar es Salaam computer from the Socioeconomic survey data. One was a file of household data and the second a file of ever-married women data. These are the first two files listed in Appendix Table 1, although at that stage the ever-married woman file did not include either household data or NDS data. Frequency distributions and some crosstabulations were produced on the Dar es Salaam computer for each of these files. Four files were produced from the NOS data: household file, death file, child file, and woman file. Only data from the latter has been used in the Study. For purposes of the Study, a special (all) woman file was prepared containing data on only women in the Study households (about one-half of all households in the Moshi areas and 219 one-third of all households in the Lushoto areas). Preparation of Data Files at Michigan State University Two data tapes were brought from Tanzania to Michigan State University. One contained all data from the Socioeconomic survey; the other contained NDS data for all women (age 12 or older) in the households in the Study. At Michigan State University, the all-woman file was split into a separate file for all ever-married women and another file for never-married women. Data from the socioeconomic woman file (these were only ever-married women) were then merged with the NOS ever-married woman file. Data from the socioeconomic house- hold file were then merged with the data on both the new ever-married woman file and the never-married woman file. These procedures created ' two new woman files, one for all ever—married women and the other for all never-married women. Each file contains all the NOS data for each woman on the file as well as all household data (from the Socio- economic questionnaire) for each woman on each file. The ever-married woman file also contains data for each woman from pages 8-12 of the Socioeconomic questionnaire. The household file was not altered. Two other files were then derived from the newly created ever- married woman file. One was the husband file. The husband file was created because the use of the ever-married woman file to analyze data on characteristics of husbands produced multiple inclusion of polygynous husbands. That is, a man with two wives would be included on the file twice, once for each wife. Data on husbands were available for currently-married women only. For each currently-married woman who reported that her husband had only one wife, the data were simply 220 transferred from the ever-married woman file to the husband file. In those households in which one or more currently—married woman reported that her (their) husband(s) had more than one wife, the data (on characteristics of the husband) from only the fjr§t_of these women was transferred to the husband file. While this procedure did not guarantee that there would be no cases of multiple inclusion of hus- bands, about the only way multiple inclusion could occur would be for wives of a single husband to be resident in different house- holds, and while this was not unknown in the study areas, it was rare. Thus, the husband file is a considerably superior source of data on characteristics of husbands than is the ever-married woman file. The second file derived from the ever-married woman file is the children file. Data on all surviving children age 7 or Older (regard- less Of current residence) of all ever-married women were transferred to a separate file in which each child is a "case" instead of each mother being a "case." Data on all other characteristics of the child's mother and the mother's household (and husband) are attached to each child on the child file. This file greatly facilitates analysis of characteristics of children. Appendix Table 1 lists all five data files together with the number of cases in each file by study areas, districts and the overall total. APPENDIX 2 THE QUESTIONNAIRES 1973 NATIONAL DEMOGRAPHIC SURVEY FOR MAINLAND TANZANIA NDS1 Block 1. Identification Name Code Confidential _Region / / Z This survey is Cluster / / / authorized by Ten-cell Leader . / / / law. All in- Head of Household77 / 1 / formation will Su ervisor _[.Z / be strictly Interviewer / /,j§confidential. Date of Interview (day/month): lst visit: 2nd visit: 3rd visit: Block 2.Household Members: Relationship Sex: Age in Is with the (Male=M com- father Head of the Female pleted alive? Household = F rs Yes No mother alive? Yes NO To be completed for all normally resident persons. Seen (S) Not seen NS 15 Block 3. H0usehold Mortality: Deaths in the household during the last 12 months preceding the date of the interview. Rela- Sex: Date of Age at Death in: Place Of Health Name tionship Male Death Completed Completed Death: Per- with the = M Months if Years if Home or sonnel Head Female Month Year 24 months more than Dispen— pre- = F or less 24 months sary sent* 1 2 3 4 5 6 8 9 10 *(a) Meg, (medical personnel), (b) Local (local doctor), (c) None Block 4. Comments of the Interviewer. 2231 222 Block 5. To be completed by every female aged 12 years and over listed in Block 2 Block 5A. The Woman Herself. Her name Her number 1. Date of Birth: Day Month Year 2. Tribe 3. Religion 4. Marital status: __ (a) Never married [_/ __ __ .(b) Currently nhrried [_/ by: (i) Civil Ceremony [_/ __ (ii) Religious Ceremony [_/ ___ (iii) Other: (a) Traditional Cerencgy [_j (b) Consegsual Union __[_ ls husband present [_/ or absent [_/ If absent, when did she last see him? Date Of present marriage: Day___ Month Year (c) Widowed [_[__ (d) Separated [_/ by: (i) Legal Divorce:__ .__ __ (a) Civil [_j (b) Religious [_/ (c) Traditional [_/ (ii) Not Divorced 5. How many times have you been married? ___ 6. Have you ever been widowed before? Yes [_/ NO __/ If yes, age when first widowed:____ 7. If ever married. date Qf_first marriage: Day Month Year 8. Are you: (a) literate [_/ _1f 50, number of years formal education completed: (b) Not literate [_j 9. Place Of birth: (a) If born in Tanzania, state district (b) If born outside Tanzania, state country 10. Occupation ___ 11. Employment: (a) Self-employed [:77 (b) Hired labourer [:7 (c):Not working [_/ 12. Cash income: Have you received any cash income during the last 12 months? Yes [:7 No [:7 If yes, (a) from wages [:7 (b) from products sold [:7 Block 58. The Woman's Husband. His Name 1. Date of Birth: Day Month Year 2. Tribe 3. Religion 4. Number Of wives___ 5. Are you: (a) literate Z 7 If_so, number years formal7education completed: (b) Not literate [_/ 6. Place Of birth: (a) If born in Tanzania, state district (b) If born outside Tanzania, state country Occupation __ Employment: (a) Self-employed [:7 (bliHired labourer [:7 (c) Not working [_/ . Cash inggme: Have you received any cash income_ggring the last 12 months? Yes [_/ No / If yes, (a) From wages [_/ (b) From products sold [_/ \DCDV Block 5C. The Woman's Children Preg- Name of Child LB=live Midwife Date of Sex: Is child still alive? birth; present Birth Male=M If yes, If dead, nancy . order (If any) SB=still at * Month Year Female where now date Of death birth birth =F living? Month Year 1 2 3 4 5 6 7 8 9 10 Reasons for pregnancy gaps of more than 2 years: *(a) Meg. (medically-trained midwife), (b) Trad. (traditional midwife, (c) None 223 A2 . 2 SOcio-economic Questi onnai re Seem-economic Questionnaire 1 August 1973 ' Block A: Identification C :ddtr 0 Echo d Suocrv sor n ever to O nterv10w: st 51C! v s t: 3rd v 5 ti tleaso put an "x" on the map below at the approximate place where this homesite is located within the cluster. (um cm 4) 11 0 kilometers I approximate CIA-’5’” boundary.'-o-.—.-o— 224 31001: 3: Questions to ask of the :{D-D 0F Til: HOUSEHOLD [INTEAVIE‘Cxi (a) For Block b (pages 2-8), interview the person listed as "dead of Jousehold" on N05 Form 62. (b) If the Head of Household is a woman (for example, a widow), ask her questions 29-55 (pages 4-7) in Block 0 and Ell questions in Block C. 12 [ I1 (1) Have you had any vocational schooling or training? Yes = 1 NO 2 2 ‘ “ IlNTEAVIEiEfi} If 33, go to Cuestion 6; if yes, go to Question 37 (2) Describe the type Of vocational schooling or training: (a) . (b) ‘5 1‘ 17 1o 19 20 ....v.................. ......C.......OCOCOOCCCCOOOCCO....... ["I’ I I/ I I: I (3) For how long did you attend (weeks, months or years)? (a) . .0...... 2122 2324 . (b)oooooeooooeooooooo .1: I I/ II:I (h) Jhat skill did you acquire? (a) . ......COOOOOOOCCOOCC......OOCOOOO (b) 2 26 . 0.00.00.........OOOOOOOOOIOOO.... I I I (5) What was the name(s) of the training institution(s)? (a) (b) ......O.........O...0.0... ......CCCCCO......O...O... I I1 (6) Did your father ever attend primary school? Yes 8 1 HO ' 2 Don't know 8 8 _ IINTEHVIEJER: If yes, go to Cuestion 1; if 22 or don't know, go to 8] 13 2 t 2 I1 (7) what was the highest standard or form your father completed? [IMTEHVIE:ZA: write number of years in box_at left, ._ [on Don't know = 88/ [ II (8) Did your mother ever attend primary school? Yes = 1 NO 2 2 __ Don't know = 8 _ [INTERVIEflfixz If £33, go to Question 2; if 22 or don't know, go to 19] l 2 l 2 I1 (9) Rhat was the highest sgandard or form your mother completed? [on Don't know = 887 - [ II (10) Have you ever been selected (chosen) to be a leader (for example, Ten- House Cell Leader, Leader of Ten-douse Cell Leaders, member Of_COOpcratlve executiVe committee, water officer, member of any other committees, etc.)? Yes = 1 Mo = 2 , INTENVIEJHR: If yes, go to Cuestion 11; if 22, go to Cuestion £27 £7 (11) If yes, describe the position(s): [ II (12) Uhat was the total number of gives your father had in his lifetime? lor: Don't know = 87 IINTEAVIL:L.: If the answer Ts only one wife, Eo_to Question 4; for all other answers, go to (uestion lg] ‘—_ 30 . [ II (13) :hat was the largest number of wives your_fathcr had at any one time? [era Don't know =‘§/ :_’ ‘l::Z:EI (14) any do you think your father had that number of wives? ......‘OCOOOOOCOOIOOIO0..........OOOOOCOOOOIOOI00.000.00.00.0............ ‘0 ......COO......O.......0...............OOOOOOOCOOOOOOOOOOO..0............ [ I1 (15) How many wives do you want to have? 41 42 I (16) Uhy do you want to have that number of wives [Th (uestion 157? '- -' o l 0.0.0.0000.........OOIOCOOOOOOOOOOOOI.....OOOOOOOOOOOOOOOOOO0.0.0.0000... 000000000000000000000000000000.000.OOOOOOCOOOOOOOOOIOOo.o-oooeooooooooooo H A S. H 48 fiiv :37- 61 62 l :30 225 (17) How many children were born alive to your father (from all his wives)? [era Don't know = 88 (18) Of all these children born alive to your father, how many survived past childhood to about age 10? [or: Don't know = 887 (19) How many children have been born alive to you? (20) How many of your children are still living? (21) If a man had the following numbers of living children, would you say he had a few children or many children? INTEJVI?Y.N: Ask the respondent about each of the numbers below (2 through 14) and for each number, tick whether the respondent thinks that number would be a few children or many childreflj dren (a) Few children (b) many children (i) (2) (3) (4) (61. (7) (22) DO you think there are any advantages in having a large number of children? Yes = I NO 8 2 Don't know = 8 ‘_ IHTaNVIalon: If yes, go to uuestion‘gz: if 23 or don't know, go to 23/ (23) If yes, what are the advantages of having a large number of children? (24) Do you think there are any advantages in having a small number Of children? Yes = l :—-——- Ho = 2 Don't know = 8 - [INTERVILHJ i If £33, go to r'uestion 32; if 22 or don't know, go to 267 (25) If es, what are the advantages Of having a small number of children? (26) Do you want to have any more children? (a) wants more = l (b) Does not want more = 2 (c) Other (specify: (27) Do you think children these days are more likely to die, less likely to die, or have about the some chance of dying as in the East? (a) Sane ‘ l (b) Bore likely = 2 (c) Less likely = 3 (d) Unacrtain = a (28) ihy do you think so [3h Question 217? (no cum 4) (neon CARD 5) [:71 LE7 226 (29) Please indicate how many of the following items your household owns and also indicate how many of each item were purchased within the past.l2 months: Type of item icyc e mo rcyc 0 car {‘0 0 watch c oc crd ar the r eas Gnalr so a table made by car c 0t “5 oox clothes c UOurC wood bed metal be cotton, kapok or ruooer mattress erosene stove ordinar lantern essure lantern torch as ir.t umbrella sew mac inc Kumber owned now ."l number purchased in the past 12 months (30) Please provide the following information about your livestock: anber number slaugh- Sumner sold' Type of you tered for home for cash in livestock now consumption in last 12 own last 12 months months (15 ‘3“. (2) Jan (3) 6:1. (a) (a) cattle 39‘4“: '“ 4243 (b) goats “"5 46-47 48.49 (c) sheep 50-51 52 53-54 (d) pigs 55-5'3 57-58 59-60 c) chickens 5562 _ and ducks 63 64-65 “’67’60 lNTfiJVIEwEx: If the respondent has 22 cattle, go to Question 21; if the respondent now owns cattle, go to Question 3j7 (31) Did your father own cattle? Yes 8 l , [0'0 3 2 . linemen-m If yes, go to l‘uestion _3_2_; if 23, go to Guestion 317 (32) Uhy is it that your father had cattle but you don't have any cattle? 227 IINTEnVIEHLR: If the respondent has no cattle, now go to fuestion 21; if the re5pondtnt now owns cattle, now go to tuestion‘227 g: (33) How many of your cattle are "grade" ("exotic") cattle? (34) How many of your cattle are "local" cattle? (35) Do you think you have more cattle or fewer cattle than your father had when you were a child in your father's home? (a) I have more cattle than my father had = l (b) I have fewer cattle than my father had = 2 . (c) I have about the same number as my father had = 3 H s: \H pa E” (36) th do you think this is so 13h Question 357? (mum 5) ......OCOCOOOOCO......OCOOOCOIOCOOOOO.......OCIOIOOOOOOOOOOOOI (8‘01.qu ......OOI.....OOOOOOOOOOt...IOIOCOOOCOOICOOOOOC00.00.00.000... (3?) Crops: Please provide the following information about the crops you grow: . * (1 2) * ~__§3 4 5'5.") (=} ° (1) (2) (a) (h) 1 Ticn if now mucn did you now much lame , _ of grown harvest last year sold for Cro by re- Lbags, kilo, or.;'cash(cedes 1 3 3 4 5 5 7 9 p :pondent .......(spccifyl/ bclow) I T I I I / /. / I11—1o 4a) 01=coffce 7—'I I / I I I. / 7 fi-R (b)Ld=tu; ‘7 I I I I I 7. I I 3%93 (c) 03=wettle barks IF I I I I I I. I;_I 4b48 (d) 3£=cardamon > fl I I I / I./ I51_-se \cJU5338531’1dS , . ,-'//////////////,// [1'72 77 I I I Io / I 5b40 (f) 06=maize I, I I I I I I. I 47 7343 (a) b7=beans 7“,] I I I [2,]. I‘ll H-LJ 1h) 08=finser millet I I I I 7 / /. I I21-23 u) 0>~=c;t)bf.zc I I_7I2—7 I If I; I I 3%85 (J) lt=onions 'F I I 77 I I. I [4148 (k) 11=Carrots I I I / I I.. I Isms (1) 12=bepQ-‘:r‘ ' Ill/i I I I L I I Gbé' (m) 13=tozato s I I I I I I. I I'll-73 (n) lu=l-...,k (“091110 6) (atom cm: 7) None = 2 fl ‘2 1 1" Very 12.:th = 3 I Z 2 2 II 38 how man coffee tr- d ADO“: one-quarter = a ( ) y Les 0 you have? About one-half = 5 1 16 17 10 about three-quarters 8 6 I (39) How many tea trees do you have? leost all = 7 19 20 21 22 "11 = 1 I (40) How many wattle bark trees do you have? 23 24 25 26 l 2 I 2 II (61) How many cardamon trees do you have? (42) Counting all your gardens (farms) together, would you say they are bigger, smaller or about the same size as all of your father's _ gardens (farms) when you were a child? I 71 (a) Father's gardens (farms) were bigger than mine = l (b) Father's gardens (farms) were smaller than mine 8 2 (c) Father's gardens (farms) were about the same size as mine a 3 (AJ) dhy is this so 13h Cuestion @27? m 2, O.......OOOOIIOCIOOOIOOOO00.0.0000... LII 0000000000000coo-cocoon000000000000.ococoon0000000000000oo'oocoocoooo fin Iii/i"? 73 74 228 (44) Do you have any sons of less than age 20? Yes i 1 Ho ' 2 IINTERVIEIEX: If yes, go to Cdestion‘gzs if no, go to question g§7 (65) If '03, what plans have you made for providing land for your sons? the following farm implements you own; 0 lement a) tractor b s a r C .003 C cart ~ vas macn w or grin lLfi maizt nncn he or p n; S CVQ or O ; r; i; v s :\AOV 11.: bye. r9 .v coffee fernentinc vat co ee ermontznz x runs Of mu 1715 an stor beer owned now 90. umber (46) Farm Implements: Please provide the information requested below about Number purchased in the t 12 months 44 45 43 l. t 54 so (47) How do you think your family is doing financially as compared with 5 years ago? 15 your family doing much better now, a little better now, (a) huch better now = l (b) A little better now = 2 (c) chh worse now = 3 a little worse now, or much worse now? (d) I little worse now = a (e) about the same = S (f) Uncertain = 6 (48) Please provide the requested information about the following changes you might have made in your farming operations during the last 2 years: have you done If yes, do this within you still If no in "armi —— £ract2§e the last follow this column 3, 5 years?. practice? why not? (yes = 1) (yes = 1) (no = 2) (no = 2) (1) Col. L7 (2) Col. (3) Cal. (a) a) planted f hybrid maize 60 6‘ 62 (b) aceuired grade cow(s) 63 64 65 (c) sprayed coffee ‘6 60 (d) use artiiiCial fertiliZQr 69 7o 71 IifilunVIidEn:i7h9/‘Uo you think the respondent answered (uestions 29-a8 1217 truthfully? Yes = 1 No 3 2 Don't know 3 8 Your comments: 0 229 [INTERVIEJEA: Puestions >1 and 52 (below) are for LUSHUTO DISTnICT only7 75 l 7! (51) Are you a member of an ujamaa village? Yes = 1 No 8 2 IINTLRVIauLA: If es, go to (uestion‘égz if no, go to 23/ (52) If yes 13h question 517, what is the name of the E7 gglgAg‘ERpa) ujmaa V111a887'0000000000000000......OIOCOOO 1. l 71 (53) now many buildings does your household use altogether (including latrine and granary)? / XIZAVIeALn: Jrite the number in the Defy (54) Please provide the following information about each building that your ~ family uses: [Imfballenbnz Use the codes given below the table] dal s noo number ing Shape Uses made made Door windows Floor of Number of of . rooms 2 s I 9 ll lidfl 2.2—31 314! 4161 52$: czq: 6 CODE NUMEEkS: [2] Shape /3-57 Uses L737 'Aalls made of [27 Roof made of 0=round O=residential 0=grass O=grass l-cquare or (sleeping) l=thin sticks l=other thatch rectangle 1=store =uncut wood (banana, etc.) (granary) 2=wattle & mud uncovered 2=reed mats 2=kitchen 3=wattle & mud covered 3=flattened tin 3=latrine 3=mud brick containers 4=animal a=corrogated metal sheets 4=roofing paper building S=cut (sawed) wood S=corrogated (shelter) 6=stone metal sheets S=others 6=concrete block (cement) 6=tiles (or (specify) 7=fired brick asbestos) 7=other uses 8=others (specify) 7=wood (sawed) not mentioned 8=others (Specify) [87 Door 12] Windows [167 Floor [117 Number of Enone O=none [Edirt . rooms Issack l=wood frame uncovered 1=rocks or l=one room Z’grass 2=metal frame uncovered stones ’2=two rooms Z'reed 3=wood frame 2=lime 3=three rooms - 3=wood (uncut) with shutters 3=wood a=four rooms 4=flattened tin a=metal frame a=concrete S=five rooms containers with shutters S=other 6=six rooms 5=corrogated S=wood frame with glass (specify) ( . . metal sheets 6=metal frame with glass (brite 1" 6-wood frame 7=others (specify) total number - of rooms) 73glass 8=other (spec —a— ify) INlthiLkax [53/ consider all the buildings you have seen in this cluster so far. Please indicate how you think the buildings of this household conpare with the buildings of the other households in this cluster. [37 relatively Root (poorest one-quarter of all households) = l 737 relatively 53”“ (best one-quarter of all households) 3 2 ZE] about aVoranf-zmiddle'hilf of all households) = 3 14 75 M =3 '8 B (m cum 8) 230 IINTEKVIEJLA: If’tfie Head of neusEhold was a women, now ask her all QUestions in Block C (pages 9-12). If the Head of household was a man, you now answer questions 56-59 below] INTEAVIE§£nx [30] If the head of dousehold was a man, did his wife (or wives) listen to the interview? Yes = 1. No 3 2 [577 If es, explains ......OIOOOIOI00.00.00...OIOOOOOOOOOOOOOOIOOIOOOIOOII... 0.000....OOOOOOOIOOOOOOOOI.IOOOIOOOOOOIII.......OOOOOOOO ......OOOOOOOIOOOOOO‘O0.0....OOOOOIOOOOIOOOOOCOOOIOOIOOOO [587 Did anyone else listen to this interview (e.g., the Ten- flouse Cell Leader or children)? Yes = 1 No 8 2 1227 If es, explain: ......C.........OOOOOOOOOCOOII‘OIOOOOOOOCOOOOOOOO0.0.... .IlNTEdVIEHXn: Now thank the man for his cooperation and ask if you can now interview his wife (or wives) alone -- if possible7 231 Block C: Questions to ask of each EVEn MARRIED WOMAN in the Household INTEAVIETaJ: (a) if there is more than 1 ever married woman in the house- household. Then, using Separate (additional) schedules, interview each of the remaining women in the household who has ever been married (using only Block C of each additional schedule). (b) Try to interview the woman alone without anyone else being present. uster en-House e O USu. n n CL‘V ewer te o nterv ew: st 5 t: d v Sit; v s t: [1371 (60) Do you have any children of 7 years of age or older? Yes = l -—- No 8 2 (CARD 9) [INTERVIENBM If yes, go to Question 21: if 23, go to Question 937 (61) If es, please provide the following information about the education of all your own living children who are of‘z years of age or older: 1 - D .- 2 Age sex tten nest CVe at- or eac ch - Guano) of of school tained if no longer in school, how Name of Child ild 1d this year in school; current much did you pa~ in (h=1) (Yes 1) level if presently in school fees ars F=2) (no 2) attendinr school this ast vezr __ .. 1. c 1&40 2%60 3140 {$60 SLJO 61-70 1h£o span OFGO 3h40 than SLAO ( ITIIUI cinn 9) .[féTi (62) Have you had any vocational schooling or training? Yes = I No 8 2 ’INTERVIEHEAI If.22' go to Question 67; if yes, go to Cuestion 637 1357' (63) Describe the type of vocational schooling or training: ” 15 ‘6 .....OOOO0.0.0.0.......0.00.0.0....OOOOOOOOOIOIOOOCCOOOO (64) For how long did you attend (weeks, months or years)? 17 16 (65) What skill did you acquire? 00.0.00.........IOOOOOOOCOOOO.30.... 1 1:55' (66) what was the name of the training institution? 232 Yes 8 1 No - 2 Don't know 3 8 (67) Did ybur father ever attend primary school? INTEAVIEUER: If yes, go to Cuestion 28; if 22 or don't know, go to £27 (68) What was the highest standard or_ form your father completed? lora Don't know = 88/ Yes 9 1 N0 = 2 Don't know = 3 (69) Did your mother ever attend primary school? ’INTEK/IEJEKI If yes, go to Cuestion 703 if no or don't know, go to 7f7 (70) what was the highest standard or form your mother completed? _/_ors Don't know = 827 (71) What was the total number of wives your father had in his lifetime? Iorl Don't know = 8/ IINTERVIF:JL{: If the answer is only one wife, go to Question_ 73; for all other answers, go to Question 72/ (72) Hhat was the largest number of wives your father had at any one time? lpr: Don' t know = 8/ (73) How many children were born alive to your mother? (74) How many of your mother' 5 children survived past childhood (to about Mg 12)? Zpr: Don't know = 887— (75) How many children were born alive to you? (76) How many of your children are still living? (77) If a woman had the following numbers of living children, would you say she has a few children or many children: (INTERVIEJER: Ask the respondent about each of the numbers below (2 through 14) and for each number, tick whether the respondent thinks that number would be a few children or many children7 Number 0 L v 23 ldren i (a) Few children (b) children 2 (78) Do you think there are any advantages in having a large number of children? Yes = l -_——_' No I 2 Don't know = 8 INTERVIEUER: If Xii' go to Cuestion 12: (79) If yes, what are the advantages of having a large number of children? (80) Do you think there are any advantages in having a small number of children? Yes 3 1 No 8 2 Don't know = 8 INTERVIEHEA: If (81) If es, what are the advantages of having a small number of children? 13%: Don't know = 887 if no or don't know, go to 827 ts, go to Question 81; if 22 or don't know, go to 877 233 (82) Do you want to have any more children? :71 (a) Wm more = l (b) Does not want more = 2 (c) Other (specify: ....‘0.......0...0..............CCOOOOIOCOCOOOOO. (83) Do you think children these days are more likely to die, less likely to die, or have about the same chance of dying as in the past? (a) Same chance as in the past =‘l ...-—_- (b) Hore likely = 2 (c) Less likely = 3 (d) Uncertain = to (84) Why do you think so 13h Question 83]? £371 ‘5 ‘7 oooooooooooooooooooooooooooooooooo ‘ 2 J ......OOIOIIO0.00.....CO...OOOOOOOOOOOOIOOOIOIIOOIOIIOOOOOOOOOOOOOOOOOO 48 [‘71 00000000000000.0000cocoons-00000000000000.000.00000000000000.0000...coo (85) Have you ever given birth? Yes = 1 N0 = 2 INTEnVIEIE.Iz If yes, go to Question £2; if 22' go to Cuestion 227 (86) Did you attend health facilities during the last 2 months of your last pregnancy? 49 (ES—Eb: at all = l I 71 (b) Services of local (traditional) doctor = 2 - (c) Attended .edicai facilities (specify below): ‘ (i) once_= 3 (ii) monthly = a (iii) twice a month = 5 (iv) weekly = 6 (v) other (specify: ......‘0.00.00.........OOOOOOOOOCC (87) Where was your act. baby delivered? so (a) hospital or modern facility = l I 71 ‘ (b) at home attended by modern midwife = 2 --. (c) at home attendei by a traditional midwife = 3 (d) at home attended bya relative = u (e) other (specify; 5, 52 (88) How long did you breast- feed the last baby that you weaned (age I of child in months or years at the :ir.e of weaning)? . [1%. Never breast fed a baby up to weaning= 777 II FNVII‘JK‘ If the Orespond ent is currently breast-feeding a ..,--..-.. which the woman weaned. ‘tke sure that the woman's answc-r does not refer to a child who died while still being breast-fed] INTERVIHTEI: If the anszer in Cuestion 88 is never bret st- -fed a baby 32 to weanina, go to Cues tion 92; otherwise, ask 6:7 53 (89) How long did you breast- ~feed the last baby you weaned before I the baby just referred to /in Question 887 (age of child in months or years at the time of .;eaning)? Iota No previous baby was breast-fed up to weaning = 727 (90) In your opinion, do women these days breast-feedtheir babies for a longer period of time, a shorter period of time, or about the same 1:37: length of time as did your mother and other women of her age? (a) Women today breast-feed for about the same length of time _as in mother' s day=1 IINTIHxV IE_T2: If this is the answer, go to Question 927 (b) Women today breast-feed for a shorter period of time 2'3 (c) Women today breast-feed for a longer —period of time = 3 (d) Unce§:_in = 8 IINTEMVILIE{: If the answer is (b), (c) or (d), go to Question 917 (91) Why do you think this is so? 22i4 (92) Which of the following foods has your household eaten in the last month? INTdRVIFiéin: Yes = l "— No = 2 DOn't know = 8 _22. 60 61 62 Hilk___/> tieat_/__/ Poultry_/_—__7 Eggsfl Fish_/_:_7 (93) which of the following foods has your household eaten in the last week? ’ fiilklfi Meatfi Poul tryfi Eggsfi Fish_/:Zf n E nwtd: u was an interpreter re Uire at any time during t.e entire fI‘T avr” TJ‘F‘ ' q ' a ' *1 I 71 interview of this woman? Yes = 1 N0 = 2 7° 7‘ L9_5_ If 08' prlam'......OOOOOOOOOOOOOOO0.000000000IOOO 0.00.00.00.00..OOOIOOOIOOOdOOOOOOII0.00000000000COOOOOOO. .0.........IOIOIOIOOIOOO00.000.000.000....OCOIOOOOOIIOOIO ......OOOIC0.0......0C0.0.0.0.0..........OOOOOOCOCOOOO... 72 . . . . /-—71 [907 Did the woman's husband listen to any part of you: inter- -—- View of this woman? Yes = 1 No 3 2 73 I977 If yes, did he interfere in any way with the woman's [ 71 responses?. Yes = l 1 N0 = 2 [2&7 If cs, explains [997 Did any other person or persons (e.g., the Ten-House Cell 76 Leader or the woman's children) listen to any part of the l 71 interview with this woman? Yes ' 1 N0 = 2 [1507 If es, explain: 777° (m cm 9) IlNTERVlEUeAs Thank the woman tor her cooperation. do sure to interview all women in this household who have ever been married (using Block c, paces 9-127 APPENDIX 3 CODING INSTRUCTIONS FOR THE SOCIOECONOMIC QUESTIONNAIRE Standard Codes (but there are some exceptions as noted in instructions): l = Yes 2 = No 6 = Other 7 = DA (Doesn't apply or NA = not applicable) 8 = NS (Not stated or NR = no response) Ql. l = Yes 2 = No 7 = DA 8 = OK 9 = NS 02. (For both a and b) l = good agriculture 2 = teaching 3 = carpentry 4 = engineering mechanics 5 = masonry 6 = tailoring and weaving 7 = other 8 = DA (Doesn't apply) 9 = OK (Don't know) l0 = NS (Not stated); also the code for (b) (col l4) if only 1 vocational schooling is given; means "no second vocational training given.“ 03. 004 = l month 056 = 13 months l08 = 25 months Similarly from l60 = 008 = 2 months 060 = 14 months 112 = 26 months 37 months up to: 0l3 = 3 months 065 = 15 months ll7 = 27 months 775 = l79 months 0l7 = 4 months 069 = l6 months l2l = 28 months 02l = 5 months 073 = 17 months l25 = 29 months 777 = DA 026 = 6 months 078 = 18 months 130 = 30 months 888 = OK 030 = 7 months 082 = l9 months l34 = 3l months 999 = NS (or "continuing" 034 = 8 months 086 = 20 months 138 = 32 months --i.e. still train- 039 = 9 months 091 = Zl months l43 = 33 months ing; no duration 043==lO months 095 = 22 months l47 = 34 months given) 047==ll months 099 = 23 months l5l = 35 months 052= l2 months 104 = 24 months l56 = 36 months Note: The response for (a) should be coded in cols lS-l7; response for (b) should be coded in cols l8-20. Q4. l = agriculture 2 = teaching 3 = furniture making & roofing houses 4 = repairing out of order motor vehicles 5 = masonry 6 = tailoring & weaving 7 = other 8 = DA = (Note: if 2b--Col l4--coded "9”; code 4b--col 23--"7") 9 = OK l0 = NS Note: For 4a use only col 2l; do not code col 22. For 4b use only col 23; do not code col 24. 05. l = Teacher's Training College Singa Chini 2 = Moshi Carpentry Shop 3 = Masoka Training Centre 4 = Zamzali Contractors Ltd. 6 = other 7 = DA (Note: If 2b--col l4--is coded"9"; code 5b--col 26--as "7”) 8 = OK 9 = NS 06. l = Yes 2 = No 7 = DA 8 = OK 9 = NS 235 Q7. 08. 09. 010. 011. 012. 013. 014. 015. 016. Vain VOWVO‘U‘IDQJ N—l 236 Code according to number of years; 00 - 20 (approx.) 77 = DA 88 = UK 99 = NS 1 = Yes 2 = No 7 = DA 8 = OK 9 = NS Code according to number of years: 00 - 20 (approx.) 77 = DA 88 = OK 99 = NS l = Yes 7 = DA 9 = NS 2 = No 8 = OK = V.D.C. (Village Development Chairman) = Chairman of a Committee or Group (eg. TANU Youth League, School Committee, Parish Committee, etc.) = Ten-House Cell Leader (Balozi) = Co-operative Society Committee Member = Member of a Conciliation Board = other = DA = OK = NS; also code for col 35 if no second leadership position given ode according to number given: 0 - S = 6 or more 8 = OK = DA 9 = NS Code according to number given: 0 - 5 6 = 6 or more 8 = OK 7 = DA . 9 = NS 0 = to get more wealth (i.e. to cultivate more land, care for more cattle, etc.) l = to have more children; prevailing norms about desired number of children 2 = due to custom, religion, tradition, etc. 3 = wanted more that l wife; wealthy enough to have more than one wife 4 = divorce, death or separation of l or more wives; death of children; one or more wives childless 5 = wanted only l wife (or was in his interest to have only 1 wife); was not_ wealthy enough to have more than l wife 6 = other 7 = DA 8 = OK 9 = NS; also code for col 39 if only 1 reason given (as coded in col 38). Code according to number given: 0 - 5 6 = 6 or more 8 = UK 7 = DA 9 = NS 0 = to get more wealth (i.e. to cultivate more land, care for more cattle, etc.) l = to have more children; prevailing norms about desired number of children 2 = due to custom, religion, tradition, etc. 3 = wanted more than l wife; wealthy enough to have more than l wife 4 = divorce, death, or separation of l or more wives; death of children; 1 or more wives childless 5 = wanted only l wife (or was in his interest to have only l wife); was not wealthy enough to have more than 1 wife 6 = other 7 = DA 8 = OK 9 = NS; also code for col 42 if only l reason given (as coded in col 4l) 017. 0l8. 019. 020. 021. 022. 023. 024. 025. 026. Code 76 = 77 = Code 76 = 77 = Code 76 = 77 = Code 76 = 77 = a. N—J II II N—‘O II II II Nd somwosm-bu II II II II II II II II II omummwa-oo omwmmbwmd 237 according to number given: 0 - 75 76 or more 88 = UK DA 99 = NS according to number given: 0 - 75 76 or more 88 = OK DA 99 = NS according to number given: 0 - 75 76 or more 88 = UK DA 99 = NS according to number given: 0 - 75 76 or more 88 = OK 0A 99 = NS Code with the number at the bottom of the right-most column ticked in row (a): l - 7: corresponding to-the right-most column ticked Note the following special codes: 0 = code for (a) in all cases where (b) is coded "l" (meaning that row (b) should be ticked in the first col and row (a) should not be ticked) 9 = NA or NS 8 = Not an acceptable code for (a) .4 Code with the number at the bottom of the left-most column ticked in row (6): l - 7: corresponding to the left-most column ticked in row (b) Note the following special codes: 8 = code for (b) in all cases where (a) is coded "7" (meaning that row (b) should not be ticked at all in these cases) 9 = NA or NS 0 = N93 an acceptable code for (b) Yes 7 = DA 9 = NS No 8 = UK Provide happiness or prestige To increase the size of the family (or the population or nation) To help with work-—e.g.: cultivating shamba, household work, etc. (also: to increase family income or family wealth) To care for me when I'm old or sick To help solve my problems If some children die, some will still remain other DA DK NS (col 54 if no response given); also: code for col 55 if only one reason given (as coded in col 54) Yes 7 No 8 DA 9 = NS DK Easier to bring them up or easier to feed them Easier to educate Easier to give them land They will care for me better when I am old God's will other DA DK NS (col 57 if no reason given); also: code for col 58 if no 2nd reason given If God wishes wants more Does not want more Too old or too ill Cannot bear more children Never given birth other DA DK NS 027. 028. 029. 030. 238 l = Same 4 or 8 = Uncertain 2 = More likely 7 = DA 3 = Less likely 8 = OK 9 = NS a. If in 027 (col 60) the answer was (a)="same”=l, code as follows: Col 6l: l = God's will 2 = Medical facilities have not caused reduced mortality 6 = other 7 = NA 8 = OK 9 = NS; also: code for col 62 (no second reason given) b. If in 027, the answer was (b) "more likely to die these days"=2, code: 0 l #0)!” \OGDVO‘UI malnutrition hospitals and clinics are too few and/or too far away; or don't get proper treatment (in hospitals/clinics) = the period of breast-feeding is too short; or pregnant mothers breast-feeding children die from new diseases; or colds and fevers; or too many diseases today unwanted and unprepared for children don't get proper care; or present poor child-care practices witchcraft other DA DK NS; also code for col 62 if no second reason given c. If in 027, the answer was (c) "less likely to die these days"=3, code: Ndo somwmmbw babies are delivered in clinics these days children nowadays are properly cared for people go to hospitals and clinics; or vaccinations; or long ago causes of diseases were not known nowadays people know family planning "progress“ ("development")(including good food, clothing and medicines) "local" dangerous medicines are no longer used other DA DK ' NS; also code for col 62 if only one reason given (coded in col 6]) d. If in 027, the answer was (d) "uncertain" = 8, code as follows: \omeN—n Too young to know of the past uncertain, but there are still some children dying other DA DK NS; also code for col 62 Items (a - v) and cols 63-78 and ll-38: For each col, code according to the number given: 0 - 6 7 8 9 7 or more DK NS Items (a - e) and cols 39-40 - 66-68, code according to the number given, with the following limitations; as indicated below hicases of l, 2 or 3 cols: 0 chUMi 2cMumm 3cMumm number given 00-96 = number given 000-996 = number given 7 or more 97 = 97 or more 997 = 997 or more UK 98 = OK 998 = DK NS 99 = NS 999 = NS -6 7 8 9 Q31. 032. ‘033. 034. 035. 036. N—I O OCDVO‘U‘OUO N-I' Code 76 77 98 99 Code 77 98 \DmeN—l D) 239 = Yes 7 8 DA No 8 8 DK 9 = NS I can't buy cattle because I don't have enough money (or because I don't harvest enough cash crops to enable me to buy cattle) I was not given any cattle by my father My elder brother took (or slaughtered) all the cattle we inherited from my father I have no wife to feed (care for) cattle My cattle all died from diseases My cattle all died of hunger (due to lack of food and grazing land) other DA DK NS; also: code for col 7] if only one reason given (as coded in col 70) according to number given: 00 - 75 76 or more DA DK NS according to number given: 00 - 75 76 or more DA DK NS I have more cattle than my father had I have fewer cattle than my father had I have about the same number of cattle as my father had DA DK NS If in 035, the answer was (a) = “l", code as follows: My father could not or did not keep livestock My father paid out his cattle in dowry I bought my cattle myself other DA (if respondent has no cattle) DK NS; also code for col 78 if only one reason given SOCDNO‘wN-d II II II II II II II f in 035, the answer was (b) = "2", code as follows: = I have grade cattle instead of local cattle = Nowadays we concentrate on agriculture rather than livestock-keeping; or I cannot marry additional wives (due to religion) to look after more cattle I don't have enough money to buy more cattle My father gave me only part of (or none of) his cattle Some of my cattle died due to hunger and/or diseases There is not enough grazing land other DA (if respondent has no cattle) DK NS; also code for col 78 if only one reason given I 0 l @mNO‘U‘Ith II II II II II II II II f in 035, the answer was (c)= "3", code as follows: My father paid out his cattle in dowry other DA (if respondent has no cattle) DK NS; also: code for col 78 if only one reason given I l 6 7 8 9 240 Q37. (l) Replace the coding box to the left with the new coding box (it must be pasted onto the questionnaire). (2) 037 is to be coded onto 2 cards: Cards 6 and 7 (a) Card 6: for items (a- .g) or (01- -07); Code the following cols of this card: ll, .17; 2], .27; 3], ...,37; 41, M47 5], ..57; 6L ,67; 7L 77. Do not code the following cols (leave them blank): l8, 19, 20; 28, 29, 30; 38, 39, 40; 48, 49, 50; 58, 59, 60; 68, 69, 70; 78, 79, 80. (b) Card 7: for items (h- n) or (08-14); Code the following cols of this card: ll,...,l7; 2l,...,27; 3l,...,37; 4l,...,47; 5],...,S7; 6L 67; 7L .77. Do not code the following cols (leave them blank): l9, 20; 28, 29,30; 38, 39, 40; 48, 49, 50; 58, 59, 60; 68, 69,170; 78, 79, 80. (3) Within each of the above "sets", the following codes are to be used: cols l & 2: for “if grown and unit of measure“: 00 = Grown but nothing harvested (code cols 3-6: 0000; code col 7: 7) Dl = Grown but amount harvested is not given (code cols 3-6: 9999; col 7: 7); this is also code for bananas 02 = Not grown (code cols 3-6: 9997; code col 7: 7) 99 = Not stated whether grown or not (code cols 3-7: 99977) If grown, unit of measure given: 03 = kilo 04 = bags (gunia) 05 = tins (debe) 06 = bowls (bakule) 07 = baskets (kikapu) 08 = big baskets (matenga) 09 = bundles (mizigo) IO = trees (miti) ll = No. of "heads" ("kabeji") 12 = gallons (galoni) 13 = buckets (ndoo) 14 = 100 kilos (that is, TIMES lOO Kilos) IS = l00 heads (that is, TIMES l00 heads) Cols 3, 4, 5, 6: Amount harvested - From 000.0 up to 999.6 for specific amount given 9997 = DA (where cols l & 2 are either 02 or 99) 9998 = OK 9999 = NS Col 7: Pr0portion sold for cash: All (Proportion = l.00) None (Proportion = 0) Very little ( Proportion = 0.l0) About one-quarter (Proportion = 0.25) About one-half (Proportion = 0.50) About three-quarters (Proportion = 0.75) About all (Proportion = 0.90) DA NS \DNmQU‘OWN-d uuuuwuuuu 038. Code according to number given: 000 - 9995 9996 = 9996 or more 9998 = OK (Note: 9997 is not an acceptable code) 9999 = NS 039. Code according to number given: 0000 - 9995 9996 = 9996 or more 9998 = OK (Note: 9997 is not an acceptable code) 9999 = NS 040. Code according to number given: 0000 - 9995 9996 = 9996 or more 9998 = OK (Note: 9997 is not an acceptable code) 9999 = NS Q41. 042. 043. 044. 045. 046. $033de Code 9996 9998 9999 O! O OWN—l N—l OQNO‘U‘I b w 24) according to number given: 0000 - 9995 = 9996 or more = OK (Note: 9997 is not an acceptable code) = NS Father's gardens (farms) were bigger than mine Father's gardens (farms) were smaller than mine Father's gardens (farms) were about the same size as mine DK NS If in 042, the answer was “I", code as follows: Father had many wives (or more than l wife) to work his shamba(s) I inherited only a small proportion of my father's shambas Long ago there wasn't as much scarcity of land as nowadays Up to now I have no land of my own but am using the same land with my father Father had many plots; or father had many people working for him I have divided the land among my children other NA (if answer to 042 = 8 or 9) 0K NS; also code for col 29 if only one reason given (coded in col 28) tomVOlU'ltb wN—‘O f in 042, the answer was "2", code as follows: I purchased some more land to add to that I inherited from my father People long ago did not cultivate big enough shambas for cash crops but only small ones for food products Long ago.people did not value land as much as we do nowadays other DK NS; also code for col 29 if only one reason given I l 2 3 6 8 9 If in 042, the answer was "3", code as follows: I inherited all of my father's land other UK NS; also code for col 29 if only one reason given 6 8 9 Yes No DK NS Up to now I haven't made any plans; or: They are still too young for plans for land acquisition I have already given them land to settle; or: I have already started cultivating a shamba for them I will give this shamba to one or some of them and I will seek (or already have obtained) more land for the rest They will get this shamba; or they will have to share this shamba (for I don't have any others) = They will have to join ujamaa villages; or: The government will have to provide for them I am concentrating on their education rather than on giving them land other 0A DK . NS; also code for col 32 if only one reason given (as coded in col 31) Items (a - m) and cols 33-58: For each col, code according to the number given: 0 - 6 7 8 9 7 or more UK NS 242 Yes, I have done this within the last 5 years No, I have not done this within the last 5 years In each case where previous col (i.e. cols 60, 63, 66, or 69 respectively), If previous col coded "l", code as follows: Yes, I still follow this practice No, I do not still follow this practice In each case where previous col coded l, 8 or 9, code here: 7 = DA If previous col coded "2", code as follows: Not enough money Hybrid maize doesn't yield as well as local varieties other UK NS Not enough money Grade cows are too delicate to stand adverse conditions other UK NS Not enough money Have not yet obtained insecticides other UK NS Not enough money other UK NS 047. 2 = Much better now 3 = A little better now 4 = Much worse now 5 = A little worse now 6 = About the same 7or9=Uncertain l0 = NS 048. a. Cols 60, 63, 66, 69: 1: 2: 8 = OK 9 = NS b. Cols 61, 64, 67, 70: coded 2, 8 or 9, code: 7 = DA 1: 2: 8 = OK 9 = NS c. Cols 62, 65, 68, 71: 048a,_col 62: l = 2 = 6: 8: 9: 048b, col 65: l = 2: 6: 8: 9: 048C, col 68: l = 2: 6: 8: 9: 048d, col 7l: 1 = 6: 8: 9: 049. l = Yes 8 = OK 2 = No 9 = NS 050. If in 049 the answer given is "l", code as follows: very well other DK NS 100001014:- DON-d He was very cooperative He showed me some of the things I wanted to see without hesitation I was accompanied by the lO-House Cell Leader who knows the man His financial position and his circumstances matched with his answers I feel quite sure he answered truthfully (or: "I have no doubts”) 051. 052. '053. 054. 055. 056. 057. 058. 059. Interviewer Name: \JO‘U‘IOWN—‘O \Dm NN-fl ‘0 II II 243 Yes No NA (This code is for all households in both Kilimanjaro clusters but it is an unacceptable code for all households in both Lushoto clusters) DK NS Funta Ujamaa Village Balanghai Ujamaa Village Bawa Ujamaa Village Manga Ujamaa Village Kilombero Ujamaa Village Mkumbulu Ujamaa Village other NA (a) Code for all households in both Kilimanjaro clusters (b) Code for those households in Lushoto clusters for which the answer in 051 was "2", "8" or “9" OK or NS Kisiwani Ujamaa Village Code according to number given: 0 - 6 All OmUN—i ‘OmN-d 7 = 7 or more 8 = OK 9 = NS codes are given on page 7 of Questionnaire (However, on tape for cols 2-l0, all codes have been increased by l-—e.g., 0 has become l; l has become 2; etc. For cols 2-10, ID = NS; for col 1]. 9 = NS.) Relatively poor (poorest one-quarter of all households) Relatively good (best one-quarter of all households) About average (middle half of all households) DK NS Yes No 0K NS Code only col 74; leave col 75 empty (uncoded): N II II I = Hife conversed with him (sometime during interview) 2 = Nife listened to him (did not talk to him) 6 = other 7 = NA 8 = OK 9 = NS Yes 8 = DK No 9 = NS Code cols 77 and 78 as follows: Ol = Balozi present; did not interfere ll = Balozi present; interferred in some way (eg: helped answer one or more questions) 02 = Neighbors present; did not interfere l2 = Neighbors present; interferred in some way 03 = Mother-in-law present; did not interfere l3 = Mother-in-law present; interferred in some way 04 = Children present; did not interfere l4 3 Children present; interferred in some way 05 = Other man (men) present; did not interfere 15 =,0ther man (men) present; interferred in some way 06 = Other woman (women) present; did not interfere I6 = Other woman (women) present; interferred in some way 77 = DA 88 = OK 99 = NS 1 = S. Jim 3 = Mtangi 5 = Mlingi 7 = Mchau 9 = Shoo 2 = Lance Massawe 4 = Mtui 6 = Ulomi 8 = Anase l0 = Marandu 060. 06l. 062 .- 063. 064. 065. 066. 067. l 2 244 Yes No Read across the top of this table: Cols l, 2; 3; 4; 5, 6; 7, 8, 9. 0: For each child whose name is entered, these cols should be coded as follows: N—l \OQNO‘wN-d Cols l, 2: Code according to age given in years: 07 - 97 98 = OK 99 = NS Cols 3: (Sex): I = Male 2 = Female 9 = NS Col 4: l = Yes ' 2 = No 9 = NS Col 5, 6: Code according to highest year of formal education completed: 00 - 20 98 = OK 99 = NS (Note: Ol = Stnd l; , , , 07 = Stnd 7, 08 = Stnd 8; 09 = Form I; , , , 12 = Form IV; , , , 14 = Form VI) Cols 7, 8, 9, 0: Code according to amount of school fees reported in Shillings: 0000 - 9996 = Shs 0/- up to Shs 9996/- 9997 = DA 9998 = OK 9999 = NS (Note: In this table--06l--all children must be listed in chronological order, starting with the oldest; if this is not done, they must be re- listed correctlyf). Yes 8 UK No 9 NS Good Health Good Child Care Good Agriculture other NA DK NS Code according to number of weeks (equivalent full-time attendance): See codes given for 03, page l oomflwad —-l \OQNO‘NH N—l Did not acquire any skill Good Health Good Child Care Good Agriculture other NA 0K NS Uomboni Makomu other ‘ NA 0K NS Yes 8 No 9 DK NS 068. 069. 070. 07l. 072. 073. 074. 075. 076. 077. 078. 079. 245 Code according to number of years: 00 -20 (approximately) 77-= DA ' 88 = OK 99 = NS 1 = Yes 8 = OK 2 = No 9 = NS Code according to number of years: 00 - 20 (approximately) 77 = DA 88 = OK 99 = NS Code according to number given: 0 - 6 7 = 7'or more 8 = OK 9 = NS Code according to number given: 0 - 5 6 = 6 or more 7 = NA 8 = OK 9 = NS Code according to number given: 00 - 25 (approximately) 88 = UK 99 = NS Code according to number given: 00 - 20 (approximately) 88 = UK 99 = NS Code according to number given: 00 - 25 (approximately) 88 = UK . 99 = NS Code according to number given: 00 - 20 (approximately) 88 = OK 99 = NS N1—‘C Nd UDOQ\JO\U1¥DQJ For row (a), code with the number at the bottom of the right-most column ticked: l - 7 corresponding to the right-most column ticked (for row a). Note the following special codes: 0 = code for (a) in all cases where (b) is coded "1" (meaning that row (b) should be ticked in the first col and row (a) should 595 be ticked) NA or NS Egg an acceptable code for (a) 9 8 For row (b), code with the number at the bottom of the left-most column ticked: l - 7 corresponding to the left-most column ticked (for row b). Note the following special codes: 8 - code for (b) in all cases where (a) is coded "7" (meaning that row (b) should not be ticked at all in these cases) NA or NS This is ggt an acceptable code for (b) 0 Yes 8 OK No 9 NS Provide happiness or prestige To increase the size of the famin (or the population or the nation) To help with work--e.g.: Cultivating shamba, household work, etc. (also: To increase family income or family wealth) To care for me when I'm old or sick To help solve my problems If some children die, some will still remain other DA DK NS; also: code for col 40 if only one reason given (as coded in col 39) 080. 08l. 082. 083. 084. omwmmwa-a ..J OWWN-J CQNOU'lthd DI C. Q N—I' 246 Yes 8 = D No 9 - N Easier to bring them up or easier to feed them Easier to educate Easier to give them land They will care for me better when I'm old God's will other DA 0K NS; also: code for col 43 if only one reason given (as coded in col 42) If God wishes Nants more Does not want more Too old or too ill Cannot bear more children Never given birth other UK NS Same More likely Less likely Uncertain NS If in 6083 (col 45) the answer was (a)= same" = 1, code as follows: Co 14 l = God' 5 will 2 = Medical facilities have not caused reduced mortality 6 = other 8 = OK 9 = NS; also: code for col 47 (no second reason given) If in 083 (col 45) the answer was (b) "more likely to die these days" = 2, code: 0 = malnutrition l = hospitals and clinics are too few and/or too far away; or don't get proper treatment (in hospitals/clinics) 2 = the period of breast-feeding is too short; or pregnant mothers breast-feeding 3 = children die from new diseases; or colds and fevers; or too many diseases today 4 = unwanted and unprepared for children don't get proper care; or present poor child-care practices 5 = witchcraft 6 = other 8 = OK 9 = NS; also code for col 47 if no second reason given If in 083, the answer was (c) "less likely to die these days" = 3, code: 0 = babies are delivered in clinics these days I = children nowadays are properly cared for 2 = people go to hospitals and clinics; or vaccinations; or long ago causes of diseases were not known 3 = nowadays people know family planning 4 = "progress" ("development")(including good food, clothing and medicines) 5 = "local" dangerous medicines are no longer used 6 = other 8 = OK 9 = NS If in 083, the answer was (d) "uncertain“ = 4 or 8 code as follows: Too young to know of the past Uncertain, but there are still some children dying other DK NS ommN—a II II II I. ll 247 085. Yes N0 N5. ONd 0| II II Other Not at all Services of local (traditional)doctor once nonthly twice a month weekly NA (DA) DK NS 086. OEOCDNOlU‘IAde —l 087. hospital or modern facility at home attended by a modern midwife at home attended by a traditional midwife at home attended by a relative I alone at home other DA DK NS \DmNO‘IU‘wad 088. Code according to number of months breast-fed: 00 - 75 76 = 76 months or more 77 = Never breast-fed a baby up to weaning 88 = DK 99 = NS 089. Code according to number of months breast-fed: 00 - 75 76 = 76 months or more 77 = NA (DA); or: No previous baby was breast-fed up to weaning 88 = 0K 99 = NS 090. l = Women today breast-feed for about the same length of time as in mother's day 2 = Women today breast-feed for a shorter period of time 3 = Women today breast-feed for a longer period of time 8 = uncertain 7 = DA 9 = NS 09l. a. If the answer in 090 is "l", code cols 56 & 57 with "77" (= DA) b. If the answer in 090 is "2", code cols 56 & 57 as follows: 0 = Women are more apt to get pregnant sooner these days (due to mono- ganism, sharing same bedroom with husband, excess drinking which arouses sexual drives, no wars to keep husbands away, etc.); also: Women these days value their husbands more than their children I = Mothers have taken to bottle-feeding small babies; or plenty of aids (to breast-feeding); or supplementary foods for small babies are available these days; or: babies are well-cared for these days 2 = Women are afraid that long breast-feeding is bad for their health 3 = too much prostitution 4 = Women wean babies early so as to have many children 5 = sickness (to mother and/or child); or: poor breast-feeding 6 = other 7 = NA (DA) 8 = OK 9 = NS 092. Cols 58, 59, ..., 62 should each be coded as follows: Yes No DK NS OmNd 093. 095. 095. 096. 097. 098. 099. 0100. Cols l = Yes . 2 = No 8 = OK 9 = NS 1 = Yes 2 = No 3 = NS Code only col 70; do not code col 71; leave it empty. Code col 70 as follows: 1 = Balozi interpreted 2 = Son or daughter interpreted 3 = another woman interpreted 4 = husband interpreted 5 = some other person interpreted 6 = other 7 = DA 8 = OK 9 = NS 1 = Yes 2 = No 9 = NS 1 = Yes 2 = No 7 = DA 9 = NS 00 not code col 75 (leave it empty); code col 74 as follows: 1 = helped to interpret 2 = helped answer the school fees question 3 = ordered his wife to answer the questions 4 = listened to the interview & on mention of "death" he was disappointed and asked me (interviewer) to stop mentioning it 5 = helped answer 077 0 = interferred with other questions 6 = other (type of interference) 7 = DA 8 = OK 9 = NS l = Yes 2 = No 9 = NS 01 = Balozi present; did not interfere 11 = Balozi present; interferred in some way (e.g.: helped answer 1 or more questions) 02 = Neighbors present; did not interfere 12 = Neighbors present; interferred in some way 03 = Mother-in-law present; did not interfere 13 = Mother-in-law present; interferred in some way 04 = Small children present; did not interfere 14 = Small children present; interferred in some way 05 = Other g§g_present; did not interfere 15 = Other man present; interferred in some way 06 = Other woman present; did not interfere 16 = Other woman present; interferred in some way 77 = DA 88 = UK 99 = NS 248 63, 64, ..., 67 should each be coded as follows: 249 Coding instructions for map (cover page) Place the map transparency over the cover of each questionnaire. Find the "x" which the interviewer put on the map (cover of the questionnaire: it marks the location of the household within the cluster). Code the box (col 11) found to the left above the map according to where the "x" is located between two contour (elevations in feet) lines. The codes are given below. Note: These instructions apply only to the 2 Moshi district clusters: 061 and 063. Code all households for the Lushoto clusters (162 and 163) with ”10". Codes for Moshi District (clusters 061 and 063): 3900 to 4200 feet 4200 to 4500 feet 4500 to 4800 feet 4800 to 5100 feet 5100 to 5400 feet 5400 to 5700 feet 5700 to 6000 feet 6000 to 6300 feet NS (this is for clusters 061 and 063) \DmVO‘mwa—i 11 II II II II II II II II Note: If the "x" falls exactl pp_a dividing (contour) line, code it in the upper category (e.g.: If an "x" is exactly on the contour line of 5100 feet, enter code ”4" for 5100 to 5400 feet). 10 = Code for all_households in the 2 Lushoto district clusters (162 8 163) APPENDIX 4 INTERVIENER'S AND SUPERVISOR'S MANUALS FOR THE SOCIOECONOMIC QUESTIONNAIRE Name: INTEKVIENER'S MANUAL FOR THE SOCIO-ECONOHIC QUESTIONNAIRE This Hanual is designed to assist the interviewers and supervisors in using the socio-economic questionnaire. It's purpose is to explain morezfnlly the schedule so as to enable the interviewer to do the best possible job. The manual is divided into the following three sections: A. Instructions for the Interviewer in the Schedule Itself 3. Instructions about Specific Questions and Sections C. Other Information to the Interviewer A. INSTRUCTIONS FUA Tdh INTERVIEJEK 1h THE SCdLUULE ITSELF A number of signs or signals have been built-into the questionnaire. These signs are in the form of instructions to you. They are listed below. 1. The Column of the Question Number. There are two separate columns of question numbers. Those questions for which the question number appears in the left- -hand column are to be answered for all respondents. Those questions whose numbers are in the right- hand column are—to be answered for some -- but not all -- respondents. For these questions, whether or not they are answered TSF'E—particular respondent depends on the respondent's answer to a previous question, and the instructions to the interviewer which followed that previous question. For example, on page 2, Cuestion l is in the left- hand margin. This question is to be asked to OVery respondent. ruestions 2- -S are in the right- hand column. These questions are to be asked only to certain respondents. The instructions to the interviewer following question 1 -- where it is written: IMDADISI -- indicate when these questions are to be asked. In the case of questions Z-S, these are to be asked only if the answer to question 1 is yes. The interviewer' s instructions after question 1 tell you that if the answer to question 1 is no, you are to skip over questions 2- -S and go directly to question 6. Similarly, question 6 (which is in the left-hand column) is to be asked of every respondent, but question 7 (in the right-hand column) is to be asked only if the answer to question 6 is yes. If the answer to question 6 is no or I don't know, skip question 7 and— go directly to question 8. (Notes Questions 38-43 on page S are something of an exception to this. They may appear to be in the right-hand column but actually they should be considered in the left-hand column, and they should be asked of every respondent.) 2' 1§2£21§l and If. :7' Specific instructions to the interviewer are given in many places on the schedule. One type of instruction is given within enclosed straight-line brackets -- for example, page 2 between questions 1 and 2: "lhDADISI: Kama jibu ni . . . . .7". Also for example, after questions 7 and 9, the following is given in brackets: “f—u: Sijui = 887“. 3. Instructions in Enclosed aoxes. The second type of direct instructions to theginterviewer is completely enclosed within a double-line box. One example of this is at the top of page 8. 4. Qgestions in Enclosed boxes. All questions within double-line boxes are to be answered by the interviewer himself. They are not to be asked to the respondent. [ll of these questions which are not to be asked of the respondent but are only for the interviewer himself to answer have the question number in a box -- for example, /;97. The first example of this is found at the bottom of page 6 -- questions /;97 and f__7. 5. Answer in "numbers". Wherever possible, the answer to a question is to be given as a number and the number is to be written in the box or boxes located at the extreme left-hand side of the page. For example, for question 1 the answer is either yes = l or no = 2. If the answer is es, the number "1" is to be written 1n the box on the left (which has the small number "12" given above the box). 250 25] - 2 _ 6. g. Wherever the letter "N" -- meaning "Mdadisi" -- appears next to a box (or boxes) on the left-hand side, it means that the interviewer -- mdadisi -- should himself fill in this box with the number which correSponds to the proper answer. For Bismp1e,‘33 questIgn 1, there is an "M" next to the box. This means that the interviewer himself should write the answer in the box -- he should write "1" if the answer is yes and "2" if the answer is no. Where "M" does not appear next to the box (or boxes), the interviewer should not write anything in the box but he should write out his answer in the space—provided following the question. In questions 2-5, there is SE— "M", so the interviewer should not write anything in the boxes but he should write out his answers to these questions in the spaces provided. Questions 6-10 each have an "M", so the interviewer should write his answer in the box. Question 11 does not have an "M" next to the box, so the interviewer should write out his answer in the space provided. 3. INSTRUCTIONS ABOUT SPECIFIC SECTIONS AND QURSTIONS You should find most of the questions straight-forward and unambiguous, and you should have no difficulty in answering properly. For example, the answer to question 1 (0 I) will be either yes or no, and since there is an "M" next to the box, you should either put ”1" (for yes) or "2" (for no) in the box. However, some of the questions (and other instructions in the question- naire) may require clarification in addition to what appears in the schedule itself. Some of these are discussed below. _ The first item is for completing Block A -- identification -- on page 1. In so far as possible, this information will have been completed in advance. The cluster name and code will have been written in. You may also find that the name and code number of the Ten-House Cell Leader (Balozi) and the Head of Household have also already been written in. If so, make certain that the name of the head of household which is written on the schedule is the same as the name of the person you are about to interview. (If the name and number of the balozi and the head of household have already been written in, then for each head of household on your list (NDS Form 62), there will be one socio-economic schedule already prepared for you to use. Thus, if the name of the head of household written in Block A is not the same as the name of the person you are about to interview, you are probably at the wrong household and you should ask the balozi to take you to the correct house- hold.) Before going to the balozi for assistance, however, ask the head of household for the names of all the ever-married women in his household and check these against the names of the ever-married women listed on NDS Form 62 (after the name of the head of the household). If the names given to you by the head of household (to whom you are speaking) can be matched with those of the women on N08 Form 62, then it might be possible that you are actually at the correct household, but that the head of that household is now going by a different name than that given on NDS Form 62. If by making this check you are able to satisfy yourself that you have the correct house- hold, then proceed with the interview. However, if you are still uncertain as to whether you have the correct household, then go to the balozi for assistance. If for some reason the balozi cannot locate the correct house- hold, you must inform your supervisor of the problem and let him solve it. You should go to the next household on your list (NUS Form 62). CAUTION: Under no circumstances are you to interview the members of a household unless you are -absolutely certain that the household you are about to interview is identical (the same) as the one listed on N08 Form 62 and in Block A of the schedule. Next, write in the name of your supervisor and your name (interviewer). Then write the date you are conducting this interview. If this is your first visit to this household, write today's date after "lst visit". If this is your second visit, write today's date after "2nd visit"; similarly for the third visit. If after three visits to a household you still have not finished interviewing all the necessary household members, inform your supervisor, and he should then arrange with the balozi for an appointment with the household members at which time you can complete the schedule. 252 - 3 - After completing Block A, look at the map below block a. This is a map of the cluster. As best you can, find on the map the approximate place where this homesite is located within the cluster, and put an "x" on the map at that point. Be Very careful to put the center of the "x” at the exact spot on the map where the household is located. It will be necessary for all the interviewers to walk around the cluster together with the supervisors before starting the interviewing in a cluster so as to familiarize yourselves with the geography of the area. Learn the names of the riVers and the location (place) names and find where they are on the map. defore marking the loca- tion (with an "x") for a houSehold, ask the household members (and other neighbors) the name(s) of the nearest river(s) and location (plaCe or area) names of the plaCe where this household is living. Find the nearest road, track, and/or footpath, and then carefully study the map to find these same reference points. When you are Certain that you have found the proper spot on the map, carefully mark it with an "x". If you are unCertain, ask your supervisor for help the next time he visits you. Note on the map that the approximate cluster boundaries are drawn in. In almost all cases, the homesite of the household you are interviewing will be located inside the cluster boundaries. Sometimes, however, the homesite may be located outside the cluster boundaries -- for example, on the opposite side of a river. If you find such a case, first make absolutely Certain that you are not making an error. Check very carefully the nemes of rivers and locations (areas) on the map with those given by the people in the area. If after checking carefully you still find that the homestie is outside the cluster boundaries as drawn on the map, put an "x" at the spot where the homesite is located outside the cluster. Top of pale 2: As instruction (a) indicates, you should now interview the person who is listed as the head of household on N08 Form 62 (not Form 56 as it reads in the schedule). If the head of household is a man, you should interview him using all of pages 2-8. However (as indicat3d in instruction b), if the head of household is a woman, you should ask her only the questions found on pages 4-7 (questions 29-55), and then ask her all questions in dlock C (pages 9-12). .Q_gs If the respondent has answered iii to Q 1, you should ask him 0 2. In the Space provided, write down the reSpondent's description of the type of vocational schooling or training he has had. If he has had only one form of vocational schooling or training, you should write the description of that one training after "(a)“ in the place provided for the answer to 0 2. If he has attended 2 different vocational training pregrammes, you should write the description of one of them under "(a)" and the des- cription of the other under "(b)". If he has attended more than 2 vocational programmes, write down only the two most important ones. 0 3: The places for two answers here correspond to the places provided for two-answers in O 2. If the respondent has attended only one vocational school programme, you will have written the description of it after "(a)" in 0 2. Now in Q 3, you should write how long he attended. If the respondent has attended two or more vocational training programmes, you will have written down a description of the two most important ones after "(a)" and "(b)" of Q 2. You should then match your answer "(a)" in Q 3 to the train- ing referred to in "(a)" of Q 2, and you should match "(b)“ in C 3 with the "(b)" of 0 2. 4: Write out the details of the skill(s) vauired in the training programme(s) referred to in Q 2 and Q 3. Provide as much information as possible about the skill(s) acquired. Q_Zc If the answer to Q 6 is yes (="l"), you should ask the respondent Q 7. If he does not know for how many years his father attended school, then the answer to Q 7 will be sijui (="88") and you should write the number "88" in the place provided in the two boxes to the left of Q 7, as: [8 78 “a If he does know how many years of formal education his father had, you should write that number of years in the two boxes on the left. For example, Standard 4 should be written: 0 74 75 Standard 7 should be written: [0 77 hi 253 - 4 _ Standard 8 should be written: /“7F§7fi1 Form II should be written: l-f7n5751 Form IV is: /l 72 7M and Form V1 is: /l /a In If you have difficulty translating the amount of education intofl years (for writing in the boxes), write out below C 7 the answer given (for example, if the reSpondent says his father was "certified as Grade II"). d 9: Similar to g 7. Q 11: If you find from L 10 that the respondLnt has held one or more leadLrship positions, describe tthe positi ans in the space provided after C 11. Do not be satisfied with simply describing one leadership position. If he has held at least 2 or 3 leadership positions, _write down all of them. If he has held many lLadr-rship positions -- for example, 3'31“ more, write down the S or 6 most important. Q 12: You must be Very careful with questions 12 and 13. In C 12 you should write down the total number of wives the respondent's father had in his entire lieftime. For example, if the rLspndent's father during his lifetime was first married to 2 women, then he divorced one, then he married another, then he divorced another one, and tth he married still another one, then the total number of wives in his lithime was a, even though the man was never married to more than 2 women at any one time. Thus, you should write "4" in the box to the left of Q 12. Note: write the :nswer in the box to the left only up to the answer "7". ThL answer "8" is for sijui = 8. 'fhus, if the _answ:? to L 12 is 8, 9 ,-10, 11,12 (or more) total wives, do not write the real answer in the box to the left, but instead write the real *total number of wives just below Q 12, and in the box to the left write "7" . that is, "7" is to be writeen in the box to the left for all numbers-of 7 or more, but you should in addition write down the actual total (if it's greater than 7) just underneath Q 12. L 13: If the answer to U 12 is one (that is, the total number of wives the rLspondent' 5 fater had in his lifetime was only one), then you should skip P 13 and go directly to L 14. dowever, if the answer to U 12 is 2,3, a (or more, whatever the number), or if the answer to L 12 is sijui = 8, then you should ask the respondent-T 13. The answer to L 13 should be the largest numer of wivLs the res pondent' s father had at any one time. In the example giVLn above wherL the man had a total of? wives in his lifetime, it can be seen that he never had more than 2 wives at any one time. Thus, the number you should write in the box to the left (TbE‘EnL example referred to), is "2“ (for L 13). CIUTION: You must be Very careful with ( 13 to get onl y the largest number of wives the respondent's father had at any one time. C 14: Be sure to write down all the reasons the rCSpondent gives for his father's having the number of wives-Tindicated in Q 12 and Q 13). Do not be satisfied with just writing down one reason if in the respondent's mind there is more than 1 reason. However, if the respondent is not able to provide an answer to L 14 -~ that is, if he says he doesn't know why his father had the number of wives reported in { 12 and L 13, then you should write sijui as the answer to C 14. ChUTIUNI write sijui only if the reSpondLnt is not able to give any other reasons whatsoever for his father having the number of ines reported in Q 12 and t 13. C 16: us in ( 14, probe for all the reasons the respondent says he (the respondent) wants to have the number of inLs he said he wanted (in Q 15). On this question you should not accept an answer of "sijui" because the respondent must have at least one reason -— and he probably has SLVLral reasons -- for wanting the number of wives he indicated in ( 15. dL sure to write down all the reasons he gives. L 17: "How many children were born alive to your father -- from all his wives; that is, from all the ines reported in (L12. You should write the answer in the 2 boxes_£p the left of ( 17. For example, if the answer is 8, you should writex /<>7??7h in the boxes. If the answer is 21, you should write: /2 71~7H however, if the respondent is uncertain about how 254 - 5 - many children were born alive to his father, then in the 2 boxes you should write: [8 78 7: (for sijui). That is, do not write down a "guess" about the number of children born alive to the r<:spoHUEnt' s father. urite down the number of children only if the respondent is certain he knows correctly. Otherwise, write: /3 78 h; (for sijai). Q 18: This refers to those children born alive to the respondent's father who did not die between the time of their birth and about age 10; that is, thBEE'Ehildren who were still alive when they were about 10 years old. Note: It does not mean only those children who are now age 10. Q 21: There are 7 different numbers of living children given in this question -- "2", ”4", "6", "8”, "10”, "12", and "14”. You are to ask the respondent the following series of questions: "If a man had 2 living child- ren, would you s:ty he had a few childien or many children?" ff the respondent answers 'few" children, you should put an 'x” against “few children " under the number "2' If the rCSpondent s.ys 'many children", you should put an "x" against "many children" under the number "2". You must then ask the respond- ent: "If a man had fi_living children, would you say he had a few or many children?“ Again, if the respondent answers "few children", you must put an "x" under the number "4" against "few children"; if the respondent answers ”many children", you put an ”x" under the number "4" against “many children". You continue asking similar questions for the numbers “6", "8", ”10", "12" and "la" children, and for each of these numbers you put an "x" against whether the respondent thinks that number of children would be "few children" or "many children". It is possible that for some numbers the respondent may say, "it is not {ex 1;} it is not many; it is the normal number." If the respondent gives this answer, you should not put an "x" against either "few" or "many" but you should simply leave—that particular column blank. However, this is the only situation for which you should leave a column (or columns) empty. Otherwise, you should put an "x" against either "few" or "many" according to the answer of the respondent. 0 23: If the answer to 0 22 is yes (=“l”), then you should ask the respondent Q 23: "what are the advantaTes of having a large number of child- ren?" You should write down all the advant.-ges he gives. Do not be satisfied with just writing down one advantage, but ask the respondent to tell you all the advantages of haVififi a large number of children. 0 25: This question asks: "dhat are the advantages of having a small number of children?” As indicated for Q 23 (above), you should write down all the advantages the respondent gives. igain, do not be satisfied with writing down just one advantage, but you must ask the respondent to tell you all the advantages (in his opinion) of having a large number of children. Q 26: If the answer to this question is: wants more children ("nataka zaidi"5 Z="l") or: does not want more children (“sitaki zaidi") (="2"), write either number "1" or "2" (whichever is correct) in the box to the left. However, if there is some other answer, you should leave the box empty and write out the answer (which the respondent gives) in the space provided following: "(c) Menaineyo (eleza:..............)". C 28: "why do you think so [Th Question 277?" That is, whatever the answer the respondent gave to C 27, you should now ask him why he thinks this is so. For example, if in Q 27 he said: "Children these days are more likely to die than in the past" (and you would have written "1" in the box to the left of Q 27), then for 0 28 you should ask the respondent: "Jhy do you think children these days are more likely to die than in the past?" Similarly for any of the other 3 answers to 0 27. ugain, do not be satisfied with only one answer for Q 28. Probe for all the reasons the respondent has for the answer he gave to Q 27. 255 0 29: For each item listed in O 29 -- (a) through (v) -- you should ask the questions in column 2 and column 3; that is: "What is the number you own now?" and: "What is the number you purchased in the past 12 months?" Then, for each item (a) through (v) you should write in the answer in each of the columns 2 and 3. Do not leave any blank Spaces where an answer should be. Write down the exact number the respondent gives. For example, if he says his household has one bicycle now and no bicycles have been purchased in the last 12 months, then against "Baiskeli" in column 2 you should put "1" and in column 3 you should put "0". Or for example, if against "chair" (h = "kiti”) he answers that his household now owns 12 (perhaps he could show them to you) and they have purchased 3 in the last 12 months, then against "kiti" in column 2 you should write "12" and in column 3 you should write "3". CuUTION: Do not leave any blank Spaces. arite in the proper number in every box -- even iT_the number is "O". 30: For each type of animal (cattle, goats, sheep, pigs, chickens and ducks), write in column 2 the number now owned; in column 3 write the number slaughtered and eaten at home in the —last 12 months: and in column a write in the number sold for cash in the last 12 months. CnUTlON: (a) When referring to cattle, explain to the respondent that you want to include calves and other young cattle as well as fully grown cattle. (b) Distinguish between goats and shoe ep. That is, when you ask the respondent the questions about "mbuzi", make sure he is referring only to goats, because the next thing you are going to ask him about is "kondoo" -- sheep. (c) .gain, do not leave any boxes empty. If the answer for any of the boxes is "0", EfiEh write "0" in the box. (d) If this man is a butcher, then do not include any cattle he may have purChPst to slaughter in order to 33:1 the meat to other people. 0 31: i.fter 0 30 but before 0 31 are some instructions to the inter- veiwer. us the instructions explain, if in 0 30 the respondent says he now has no cattle (including no calves or other young cattle), then ask him Q 31. If the respondent now has cattle (any number from one on up), then skip OVer 0 31 and 0 32 and go directly to Q 33. Q 32: If the respondent answers in o 32 that his father did own Cattle (yes = "1”), then ask him 0 32. rgain, be SUtL to write down all the reasons the respondent giVes for his father's having Cattle Wth he has none nimSLlf. Q 35.1f the respondent now owns Cattlc (in u 30), then ask him QUestions 33, 34, 3S ant‘ 30. In questions 33, 34, and 35, you art to write the answers in the bOXLS yourSCIf. For example, if the answer to Q 33 is "a“, thtn in the two DOXes to the left of U 33, write. /T7/()/L If the answer is "2”, write: [a 72 7h Q 36: This QUrSClUHS refers to the answer giVen to Q 35. For example, if in Q 35 the reSpondent Said: "l thL morn Cdttlu thvn my father had”, then in Q 36 you should ask him: ”Why is it that you haVe more Cattle than your father !1n!?” Wt in, do not in. sutlelL(:\4lCh just Lflh. relaxai, but writ\.(L;wn all the re sons the respondent Can give for his answfr in Q 35. Q 37: Fourteen crOps are listed in Column 1. You should ask the respond- ent whether or not ht: grows «(lch of that 14 crops. For each Cl”)p, if the answer is iii in Column 2, you should put an "x" against that crap. Thtn, for each crop for which you made an "x" (that is, for each crop tnt respondtnt grows), you should ask him how much he hirvistld l:st y.1r. (Note. You are not to ask this question for binan 5 Die USL axnerilly th rtSpondtnt will be unable to inswer how many bananas h. harvtstt..) “rite out the answer in the space provided in column 3. For example, the respondent may say he hderSCLd "2% bags of coffee". Under column 3 avainst coffee, you should write "23 bags". For cabbage the rLSpondtnt may Siy "so hté.d”; in column 3 agiinst Cabbage you should write: "60 head". For onions the respondent m