Welfari t'iU UMiiuiic umzwle Headship in arl,ovi-a:.r -lotls ls,4 v ,t 3 U - - _ _ _ _ _ . _ . . ... _ _ ... ._ Y ._ .___ __ . __ ............................. _. _ _ . .......... .~... .. . . . . . .. . . --t~~~~4a _iX tw y f . . ~~~~ ~~~ ~~1 .. .. ..A..= -~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~- -'-- '-.!i_=.=.------- tint ~~~~~~~~A IiA2 _ [ w J fi ~ ~ ~ tr ______BEfl liSi9S& _ U i _- 1& i t' p ^ . i " , l i i' ^ > l i ; . W i L n , r .. .. ....... _i - ____ _- LSMS Working Papers No. 24 Adeasu ring and Analyzing Leiels of Living in Developzng Coxniies: An Annotated Questionnaire No. 25 7he Demandfor Urban Housing in the Ivory Coast No. 26 The Cted'Iwire Living Standards Surve: Design and Imploeenttion No. 27 TheRol of Empfoyment and Earnings in Analyzing Lvels of Living: A General Mehodology with Applications to Maysia and Thailand No.28 Analysis of Househod Expenditures No.29 7The Distribtion of Welfare in C8te d'Ivoire in 1985 No.30 Quality, Quantity, and Spatial Variaton of Price: Estiating Price Elasticitiesfrom Cross-Sectional Data No.31 Financing the Halth Sector in Peru No.32 Informal Sector, Labor Markes, and Returns to Education in Peru No.33 Wage Detrmiats in Cite d'oire No.34 Guidelines for Adapting the LSMS Living Standards Questiomzaires to local Conditions No.35 The Demandfor Medald are in Developing Countries: Quantity Rahoning in Rural GOted'loire No.36 Labor Market Actity in Cote d'lvowire and Peru No.37 Health Care Financing and the De and for Medical Care No. 38 Wage Determinants and School Attainmt among MA in Peru No. 39 The Alloation of Goods witkin the Household Adults, Chidren, and Gender No.40 The Effects of Household and Community Charctristics on the Nutrition of Preshool Children: Evidencefrom Rural COted'Ivoire No. 41 Public-Prkvte Sector Wage Doifrentials in Peru, 1985-86 No.42 The Distribution of Wdfare in Peru in 1985-86 No.43 Profits fm Self-Employment: A Case Study of COte d'lvoire No.44 77T Liing Standards Survey and Price Policy Refjrw A Study of Cocoa and Coffee Production in C6ted'Ivoire No.45 Measuring the Wiilingness to Payfor Social Services in Develping Countries No.46 Nonagrmcultural Fmnily Enterprses in Cot d'Ivoire- A Descriptive Analysis No.47 The Poor during Adjushment: A Case Study of Cote d'lvoire No.48 Confronting Poverty in Developing Countries. Definitin, Information,and Policis No. 49 Sanple Designs forthe Living Standards Surmeys in Ghana and MAuritanialPlans desondagE pour les enquites sur e nivewu de vieau Ghana et en Mauritanie No. 50 Food Subsidies: A Case Study of Price Reform in Morocco (also in French, 50F) No. 51 Child Anthropmetry in CGte d'Iwvire Estimatesfromn Tzw Surveys, 1985 and 1986 No. 52 Public-Pivate Sector Wage Comparrsons and Moonlighting in Dveloping Countrus: Evidence fomn C6ted'wize and Peru No.53 Soioenomic Determiants of Fertility in COted'lvoire No.54 The Wilingnss to Pay for Education in Devdeloping Countries Evidencefrnn Rural Peru No. 55 Rigiditddes salai- Donnes microiconom iques et macro6moniques sur 'justement du marchi du travad dans k secteur moderne (in French ordy) No. 56 The Poor in Ltin America during Adjustment: A Case Study of Peru No-57 7he Substitutaity of Public and Prir.ate Health Carefor ffit Treatment of Children in Pakistan No.58 Identifying the Poor Is 'Headship' a Useful Concept? (List continues on the inside back cover) Welfare Implications of Female Headship in Jamaican Households The Living Standards Measurement Study The living Stardards Measurement Study (Lsms) was established by the World Bank in 1980 to explore ways of improving the type and quality of house- hold data ollected by statistical offices in developing countries Its goal is to foster increased use of household data as a basis for policy deiion g S l, the lSVMS is working to develop new methods to mnt progress m raisng levels of living, to identify the consequences for households of past and proposed gov- ernment policies, and to imnprove commumcations between survey statisticians, an- alysts, and poilicymakers. The iREs Working Paper seres was started to diseminate intermediate prod- ucts from the LwS Publications in the seies include criticl surveys coverng dif- ferent aspects of the L%ws data collection program and reports on improved methodologies for using Lving Standards Survey (Ls) data. More recent publica- tions reommend specific survey, questionr and data processing degns, and demonstrate the breadth of policy analysis that can be carried out using Ls data. ISk Worlkng Paper Number% Welfare Inplications of Female Headship in Jamaican Households FrWdic Louat Margaret E Grosh Jacques van der Gaag I The World Bank Was,ington, D.C Copyright 0 1993 The International Bank for Reconsuction and Development/THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433, USA. All rights reserved Manufactured in the United States of America First printing May 1993 To present the results of the Living Standards Measurement Study with the least possible delay, the typesaipt of this paper has not been prepared in accordance with the procedures appropriate to formal printed texts, and the Worid Bank accepts no responsibility for errors. The findings, interpretations, and condlusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliaed oraizations, or to members of its Board of Executive Directoxs or the countries fiey represent. The World Bank does not guarantee the accracy of the data induded in this publication and accepts no responsibility whatsoever for any consequence of their use. Any maps that accompany the text have been prepared solely for the convemence of readers; the designations and presentation of material in them do not imply the expression of any opiuion whatsoever on the part of the World Bank, its afffliates, or its Board or member counties concerning the legal status of any country, territory, city, or area or of the authorities thereof or concering the delimitation of its boundaries or its national affiliation. The material in this publication is copyrighted. Requests for permission to reproduce portions of it should be sent to the Office of the Publisher at the address shown in the copyright notice above. The Warld Bank encourages dissenation of its work and wil normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to copy portions for cassroom use is granted through the Copyright Clearance Center, 27 Congress Street, Salem, Massachusetts 01970, USA. The complete backlist of publiations from the World Bank is shown in the annual Inde of Pubibatims, whch contains an alphabetical title list (with full ordering information) and indexes of subjects, authors, and counties and regions. The latest edition is available free of charge from the Dastribion Unit, Office of the Publisher, Department F, The World Bank, 1818 H Street, N.W., Washingtom, D.C 20433, USA., or from Publications, The World Bank 66, avenue d`Iena, 75116 Paris, France. SSN: 0253-4517 Frdric:c Lauat is an economist for the Soc2&6 Gdn&rale in Paris, France. Margaret E Grash is an econamist in the Poverty and Human Resources Diviion in the Policy Research Departmnt of the World Bank. Jacques van der Gaag is the division chief of the Human Resources Operations Division in the Latin American and the Carnbbean-Country Departnent m of the World Bank. Librazy of Congress Catalogig-Publicatton Data Louat, Frederic F. Welfare implicatios of female headship inJamaican households / Fr6&dic Louat, Margaret E Ch, Jacques van der Gaag. p. can - (ISMS working paper, ISSN 0253-4517; no. 96) Includes bibliographical references. ISBN 0-8213-2384-9 1. Women heads of households-4amaica-Economic conditions. 2 Poor women-Jamaica. L Crash, Margaret E. IL Gaag, J. van der. Title IV. Seres. HQ1517L68 1993 306W59097292-dc2O 93-21834 CIP v Abstract In this paper we first compare the economic status of male- and female-headed households. We then analyze differences in the use of resources (time and money) between the two groups. Finally, we focus on the relative well-being of the children in these households. Our findings show that poverty and female headship are weakly linked. For instance, if we draw a poverty line that labels 10% of the Jamaican population as poor, 9.0 percent of people living in male- headed households are poor versus 11.1 percent of people living in female-headed households. This result is based on per capita consumption as the welfare indicator. If other indicators are used, or poverty measures other than the head count index, the differences become even smaller. If the main cause of concern for female-headed households is the expectation that female headship is highly correlated with poverty, then this concem can be put to rest. Ihe study finds some evidence of small differences in resource use between the two types of households. Labor force participation data indicate that female heads are more likely to work in the market place than women with similar characteristics who are the spouses of male heads of households. Again, the differences are small: on average 64.5 percent versus 57.9 percent. The analyses of household expenditures shows that female-headed households spend no more on food than do male headed households. However when looking at more detailed food expenditures, the differences are more pronounced. For instance, female headships appear to be associated with spending on higher quality food items such as meat, vegetables, milk and other dairy products. Perhaps the most important question answered in this paper is to what extent female headship influences child welfare. The resuts show that children in female-headed households have, by an large, equal access to social services and equally good welfare outcomes as chfldren in male-headed households. vi This paper owes much to many. We would lilke to ackmowledge he coaions, n terms of critique and saggestions, of the particMpants of th seminas organized by the PERPA and LATHR divions of the World Bank, and the Population Cwuncil in New York, and the Iutrnatonal Food Policy Research Institute. Special thans go the Badbar Diallo for typing numerous vrsions of the manuscript and to Natalie Leboucha for computational supporL U vii Foreword Policymakers concerned with amelioration of poverty have singled out fenale-headed households as one of the key target groups deserving intensified attention. This paper examines the issue of female-beadship and household welfare in a specific context - Jamaica. In so doing it iliustrates how systematc analysis of the issue should be done in other countries, and the importance of comprehensive, disaggregated data in examining the link between headship and welfare. This paper is part of a broader program of research in the Population and Human Resources (PHR) Departnent on the extent of poverty in developing ccunries and on policies to reduce poverty. The research progra. is located in the Poverty Analysis and Policy Division. The data used here are from the Jamaican Survey of Living Conditions, which is one of the Living Standards Measurement Study (LSMS) household surveys which the World Bank has implemented in many developing countries. All three aud:ors have helped the Jamaicans implement their survey and have at different times been part of the Povety Analysis and Policy Division. Ann 0. Hamilton Director Population and Human Resources DeparLment ix Table of Contents L. Introduction I The Budget Constraint 1 The Time Constraint 3 Different Preferences 4 Summary 4 II. The Data S m. Female-headship in Jamaica: Who, How Many, Where? 6 How Prevalent are Female-headed Households? 6 Why Are Females Heads of Households? 6 What Do We Mean by Headship? 8 How are Female-headed Households Different? 10 IV. Poverty, Welfare Distribution, and Female-headship 17 Consumption Measures of Welfare 17 The Distribution of Welfare in Jamaica is How Does the Consumption Distribution Vary by Gender of Household Head? 18 Poverty Measures and Female-headed Households 24 The Probability of Being Poor: Multivariate Analysis 26 Female-headship as a Targeting Indicator 28 Conclusion 30 V. Differential Resource Use 31 Tnime Use 31 Consumption Patterns 32 VI. Children's Acces to Social Services 37 Health Services 37 Education 38 Nutrition Programs 39 Summary 41 VI. Child Welfare Outcomes 43 Health 43 Nutrition 44 Education 48 Summary 50 VIII. Discussion 51 Bibliography 54 Annexes 57 x List of Tables Table 111-1: Location of Households and Gender of Head 10 Table 111-2: Household Size and Household Composition by Gender of Head 1 I Table J11-3: Characteristics of Household Heads by Gender of Head 12 Table 1114: Distribution of Heads by Labor Force Status and Gender of Head 14 Table 111-5: Characteristics of Female-headed Households by Labor Force Status of Head 15 Table 111-6: Distribution of Individuals 15-64 by Labor Force and Gender of Head 15 Table 111-7: Total Number of Household Members for each Labor Force Status, by Gender of Head 16 Table IV-1: Comparison Between Male-headed Households and Female-headed Households: Per Capita and per Adult Equivalent (Adjusted) Consumption, According to various Household Categories 19 Table IV-2: Determinants of Households' Welfare as Defined by Per Capita Consumption 23 Table IV-3: Poverty and Female Headship 25 Table IV4: Probit Models of Probability of Being Poor 26 Table IV-5: Elasticity of Poverty Measures by Gender of Household Head 27 Table IV-6: Female Headship as a Targeting Tool 29 Table IV-7: Poverty Outcomes with Transfers 29 Table V-1: Non-Employment Income and Estimated Female Labor Force Participation for Women 15-64 32 Table V-2: Non-Employment Income Received by Male and Female-headed Households 32 Table V-3: Consumption Shares by Gender of Household Head 34 Table VA4: Food Item Shares and Female Headship 35 Table VI-1: Children's Access to Health Care 38 Table VI-2: Enrollment Rates by Gender of Child and Household Head 39 Table VI-3: Estimated Probilities of Enrollment of 13-19 Year Olds Based on Probit Regression 40 Table VI-4: Children's Access to Nutrition Programs 41 Table VII-1: Health Status of Children Age 04 43 Table VII-2: Estimated Probability of Children's Diarrhea 44 Table VII-3: Gender of Head and Nutritional Status of Children Age 0-4 46 Table VII4: Gender Dummies in Nutritional Status of Children Age 0-4 Regressions 47 Table Vll-5: Percentage of Repeaters Among Children Enrolled in Schools 48 Table V11-6: Percentage of Children with Full Attendance During the Previous Week, Among Children Enrolled in School 49 Table VII-7: Enrollment in High Schools by Gender of Child and Household Head 50 List of Figures Figure E-I: Identifying Female-headed Households 7 Figure 111-2: Income, Age and Education in Female-headed Households with Partners 9 Figure IV-1: Distribution of Household Consumption per Capita and per Adult Equivalent 20 Figure IV-2: Distribution of per Capita Consumption, by Region 20 Figure IV-3: Distribution of per Capita Consumption, Male vs. Female-headed Households 21 xi Annex I: Calculation of Per Capita Consumption 57 I: The Estimation of the Distribution of Consumption Expendiures 58 m: Technical Annex on Poverty Measures 59 IV: Full Regression Results 61 I. Introduction Female-headed households are of possible policy concern because, from what we know about the broader issues of gender and welfare, we expect female-headed households to be less well off than other households. First, with a woman as the main income earner and women's less favorable labor market outcomes, we would expect female-headed households to be poorer, and perhaps to be more vulnerable to recession, including those induced by macroeconomic adjustment policies. Second, because they support the family through income-generating activities, female heads of household may face tighter time constraints on non-market activities that are important to their children's welfare. A relatively new strand of the gender and welfare literature counters this reasoning from female- headship to poverty to poor child welfare outcomes. It indicates that women use resources differently from men, so that resources in the hands of women will improve children's welfare outcomes more than resources controlled by men. Since female heads of household presumably have fuill control over their resources, their virtuous efficiency may produce higher child welfare outcomes than when a man heads the household. The strength of the forces, and the ultimate outcome in the welfare of female-headed households is an empirical question. This paper is designed to be a systematic empirical analysis of the welfare of female-headed households. We have chosen Jamaica as the country to study because female-headship is prevalent and of policy concern in Jamaica, and because we have access to good household survey data for Jamaica The first section of the paper sets the conte for the empirical work in the paper. It briefly sketches the literature on the factors and hypotheses to be investigated. Section II describes the data set used for the empirical examination of female-headship in Jamaica. Section m asks the questions 'What is female headship? Is it important in terms of numbers?" Section IV examines the link between female headship and poverty. It asks the questions "Are female-headed households poorer than male-headed households? By how much?" Section V asks 'Do female-headed households spend their resources differently than male-headed households?" Section VI examines the access of chfldren from female- headed households to social services. Section VHI com pares the welfare outcomes of children from female-headed households with those of their counterparts in male-headed households. Together, these sections ask Do female-headed households warrant extra government attention?" Section VIII concludes. The three hypotheses that guide the design of this study (ower income, more severe time constraints and differential resource use) have their empirical counterparts in most of the literature on women in developing countries. We will briefly sketch this literature and, where possible, include previous studies on Jamaica on these issues. The Budget Constraint The notion tat female-headed households wil be poorer than male-headed households is based on the evidence that women's income and wealth are less th those of men. The sibtation is neatly summarized in Commonwealth Secretriat (1989, pg. 2). "...women account for half of the world's pion, perjbrm wthw ds of the hours worked (though are recorded as working only one-tird of thse hours), receive one-tt of the world 's income, and have onehundredh of the world's propery registered in their name. 2 With such disadvantaged heads upon whom to rely for sustenance, female-headed households are at risk of poverty. The reasons that women's earnings are lower than men's are roughly divisible into three classes - - lesser human capital, discrimination, and less physical or financial capital. To take the last of these first, women have less access to financial capital and land ownership than men. The small literature on access to credit shows that women may have less access to credit than men (see Berger and Buvinic [19891, for a review). The reasons for differential access are consonant with the rest of the female-headed household literature. Women's tight time constraints may make it difficult for them to go through the applications procedures. Women's lower levels of literacy make it more difficult to apply for credit or to know where to apply. he de ju and de facto differences in property, tenure and title rights may limit their collateral. Cultural constaints may limit their ability to join networks that can help overcome the barriers to credit, their ability to travel, to act independently, etc. Finally, discrimination in the review of credit applications may be a factor. Human capital is built both by formal schooling or training and by job-related experience. Women's lesser human capital will tend to produce earnings lower than men's even in the absence of discrimination. The female-male gap in formal schooling is catalogued in World Bank (1990). In 1987, in the world, girls' enrollment rates as a percent of boys' rates were 84 percent at the primary level and 79 percent at the secondary level. This varied somewhat by region and very greatly by country. Chad, Guinea, and Rwanda all had ratios of under 50 percent As a region Sub-Saharan Africa was the worst, with averages rates of 77 percent at the primary level and 50 percent at the secondary level. Latin America and the Caribbean showed the highest participation of girls relative to boys of all regions, at 96 percent for the pdrimary level and 110 percent at the secondary level. Across the board, girls' participation relative to boys' increased markedly since 1970. Thus the extent to which women face a handicap in human capital is falling. Women's participation in labor market activities is relatively low. In the world as a whole, women's participation in the labor force is only 56 percent of that for men (UNDP, 1990). This of course varies by region. In East and Southeast Asia, the ratio is 71 percent, in the Middle East and North Africa it is only 24 percent. In Latin America and the Caribbean it is 36 percent. These aggregate figures do not take into account that women more frequently withdraw from the labor market and/or work part time, which will fiurther reduce the accumnlation of human capital through work experience. In Jamaica, however, lesser human capital among women is not the problem that it is in some countries. Primary school enrollment is universal for both boys and girls. At the secondary Ievel, girls' enrollment rates are higher than boys (see Table VI-2). For adults who have completed their education, women also fare slightly better than men (see Table m-3). Jamaican women's participation in the labor force is 80 percent of that of men (STATIN, 1990). It is also interesting to note that women are well represented in some of the high-paying occupations. For example, 8.2 percent of working women are in the professional, technical and administrative group while only 4.9 of working men are so classified. Similarly 15.2 percent of worling women are in executive, managerial and related occupations while only 7.6 percent of working men are so classified. Even when women's human capital is equivalent to men's, gnder discrimination in the labor market may lower their earnings markedly. Psacharopoulos gLal.'s (ongoing) comparative study 3 quanifies this phenomenon. Case studies for several Latin American countries have used employment suvey data to decompose male-female earnings differentials into the part due to different human capital characteristics and hours worked, and the part which is due to differential rewards to the same human capital, that is, discrimination. Studies for Argentina, Bolivia, Brazil, Colombia, Ecuador and Venezuela show that differences in education, experience and hours worked account for only about one third of male-female earnigs differentials. The other two thirds unexplained difference is about twice that found in industrialized countries. Buvinic (1990)surveys within-sector and within-job comparisons in male- female earnings and further documents the phenomenon of discrimination. For Jamaica, Scott (1990) decomposes the differences in men's and women's earnings controlling for education, experience, and hours worked, and correcting for selectivity bias by modeling women's decision to participate in the labor force based on location and household and personal characteristics. Working women earn only 58 percent of what men earn even though women work only six percent fewer hours per week than men. The average level of education among working women is higher than that of men by onehalf of a year. Scott concludes: wWomen 's higher [hwnan capitall endowment is offset by the vey strong effect of differen valuations of male and female labor in the marketplace. The unexplained [discriminatory] portion of the differenid is so strong that it not only explains the whole wage gap bu negates the effect of women hving higher endowments. In short, wage differentials in Jamaica are nota function of different levels of hwnan capital between men and women but are, unTead, due to the pricing mechanis m "(pg. 14). Whether the evidence of male/female earnings differentials implies that female-headed households will be worse off than male-headed households depends upon household structure and the sharing of resources within the family. If the female-headed household is a nuclear family from which the male principal breadwinner who formerly shared all his income altruistically, is now removed and sends no remittances or child support, then clearly the female-headed household will be worse off. The concept is, however, somewhat complicated. Female-headed households may have other adult members, who can contribute to the household income. Furthermore, the household may receive remittances either from the absent man or from other relatives. In addition, even if a man is present, he may not share all his income with the family and, of course, he is a consumer. The net effect on the household's welfare of his absence, as a contributor and consumer, given the various coping mechanisms the household may adopt, is not a priori clear. The Tihne Constraint If the female head of household is pressed for money, she is also likely to be pressed for time. Women, in general, spend longer hours on the combination of income generating activities (be they market or home-based) and domestic chores which contribute to household welfare. IDB (1990) summarizes the issues: "According to 1984 stuies made by IL) n different countries of Latin America, the following phenmna have been observed: (a) women i the workforce put in two fidl work shifts - one at home and one on the job; (2) the increase in family income generated by their remnerated work may have allowed the women in some sectors to contract ouside help for some of their tough household tasks, but tiWs help has not significany lightened women's household work load; and (3) in the households in which the regional economic crisis has signified an increase in the household work load, the male family members hae not increased their particlpation in hous,rhold chores accordingly . 2171. ' 4 When female-headship implies that the woman must increase her income-generating activities, she naturally has less time for welfare-producing domestic activities. Children of female heads of households may have to substitute for her domestic labor, or complement her eamings, thereby reducing the time available for their schooling. The general health and nutritional levels of the family may decline as the activities which safeguard them are supplanted by more immediately urgent income-generating activities. Tlis would result in lower welfare outcomes for children now, and reduce their earnings potential in the future thereby transmitting poverty to the new generation. As we shall see in Section m, Jamaican female heads of household do work more in the market than other women, and hence face this time pressure. Different Preferences Even if female-headed households are doubly constrained by low income and too little time to carry out all their tasks, the priority that they give to their children's welfare may help safeguard welfare. Dwyer and Bruce (1988) review a number of studies in three veins of this literature. One set of studies in seekdng to explain child welfare outcomes, notes that women's income, more than men's is an important detminant. A second set of studies shows that more of women's incomes than men's goes to everyday subsistence and nutrition. A third set of work shows that women devote a higher and more constant share of income to family as opposed to personal needs. In Jamaica, Horton and MIller [no date] investigate gender differences in expenditure patterns using a sample of 145 households. The sample was not drawn randomly, but rather so that it would include primarily low income households, and from families who were willing to make the large time co.mmitment required to fill out daily questionnaires over 8 months in exchange for a mondlly cash remuneration equal to about 25 percent of sample mean expendiure. The households in the sample were indeed poorer than the national mean, had larger household size, and were more often headed by women than expected. Horton and Miller first analyze broad expenditure shares, but find no differences by gender. They then disaggregate within the food share. They look at how the item's share in the food budget changes with total consumption and compute an index of dietary quality based on eight micronutrients. The authors find that while female-headed households do not consume more calories per capita, they do consume foods that rank higher on the index of quality and have more nutrients. Thus, Horton and Miller's work shows some weak support for the idea that female-headed households more than male- headed households use their resources in ways that will benefit child welfare. We will therefore replicate some of their work with our much larger and more representative data set in Section V. Summary The received wisdom from the literature on women and welfare predicts that female-headed households will be relatively poor and time constrained. Their children may or may not be deprived depending on whether female-headed households' use of resources is sufficiendy child-focused to offset the time and income constraints that may be tighter in female-headed households. It is from this background that we proceed to a systematic empirical analysis of female-headship, poverty, and child welfare in Jamaica. The next section describes the data used for the anaIysis. 5 II: The Data The data used for this paper come from the November 1989 round of the Survey of Living Conditions (SLC) and the linked October 1989 Labour Force Survey (LFS). The SLC is an integrated bousehold survey instituted by the Jamaican government to provide a basis for the analysis of poverty and social sector policies and programs.' The LFS contains the household roster and a brief set of information on labor force information. The SLC has modules on health, education, nutritional status, housing, distance to social services, and participation in government social programs. Detailed information on remittances and households expenditures, including the value of gifts and home-produced food items, is collected for use as the general welfare measure for the family. A general description of the Jamaican Survey of Living Conditions is found in Grosh (1991). The LFS has a nationally representative sample of 6000 households. The November 1989 SLC revisited two thirds of the LFS sample Qi.e. about 4000 households) about a month after the first interview. The two data sets were then merged for the analysis presented here. The merge was slightly iinperfec Therefiore, for issues supported solely by the SLC, the full SLC data set was used. Where Labor Force data were needed, the smaller subset of successful matches between the two surveys is used. The merge was completed for 95 percent of individuals and 93.5 percent of household heads. No apparent selectivity bias is implied; the percent of female-headed households in the matched and unmatched sets is the same. The SLC is based on the World Bank's Living Standard Measurement Study surveys (see Glewwe, 1990). In the course of adapting the survey to the constraints of a single interview, some features of a standard LSMS were omitted. Those most relevant to this paper are detail on the time use of household members in home production and maintenance activities, and more detailed labor force and income information. 6 m. Female-headship in Jamaica: Who, How Many, Where? How Prevalent are Fenale-headed Households? Female-headed households comprise 42 percent of all households in Jamaica. This very high rate of prevalence is important to the understanding of the phenomenon in Jamaica. With female-headship affecting nearly half of society, any disadvantage suffered by such households will have repercussions for the welfare of society as a whole, not just for a small, fringe group. Hence, the study of female- headship is important. On the other hand, because of female-headship's very commonness, it may be that it has ceased to denote a small, homogenous segment of society with common problems amenable to common solutions. If, in their prevalence, female-headed households have become very heterogeneous, or hard to differentiate from male-headed households, it may no longer be that the gender of the head is an important feature of a household. With this in mind, let us look at what female-headship means in Jamaica. Why Are Females Heads of Households? Overall, 42 percent of households in Jamaica are headed by women. Of these, three quarters (or 31 percent of all households) are headed by women who are in the oldest generation present in the household and who do not have a spouse or partner in the household. These conform to the most common percepton of female-headship. In nearly all of the other quarter of female-headed households, the woman who heads the household belongs to the oldest generation in the household but does have a spouse or partner present. In only three percent of all households are there members of a generation older than that of the head. Even within that group, the portion of female heads with and without spouses/partners is similar (see Figure EI-1). As noted, one quarter of female heads of household have a partner present in the household. In contrast, 59 percent of male heads of household have partners present. The definition of partner here is very broad, including legal and common-law marriage as well as 'visiting unions" that do not imply permanency.2 2Ihe questionnaire first asked for union status, and then asked specifically if the partner was in the household. Thus some non-resident partners, especially those in visiting relations, are excluded here. 7 Figr 111-1 Idiyhig FundabeadKd Households- (23X)~ ~ ~ ~ 2 896 _ d Hiea isml (542) mo~54Hed is sinlt (n Z9 Ned - not ve _fth a spouse I6 Needisf e Had is te need belongl to Le t the oldest gner-31) _ prn titn of the houcehold. ). ! I el~~~~~~~~~leadIs role Nlee Livw witth a 16661 I X44X)' _ ll~~~~~ead is fewte alt ~~~~~~~~~~~~~390_ 22 3861 Ra is imei 54 Read is singte Cro rtrper pr esen) Only pawc pf t is faimei 99 the head LS in 3Z _the househotd _36 NOW td liv wf th * 3 r ed sote . __ _ Q~~~~~~~~spouse J Pater |ead'ls parents Car U' -" ' le s fewle pwnt-in-lo) am _-' auto, ofsd the household. ,,Redis let | X~~~~~~~~~~~ 11kad is single Cno '-I 105 CX .Hprnr pi esent H . B8oth parets of||{-|4edis felet | the head lfve in I11 -the househotid H4r ' |gHed lves mfth a| '- Uspou" / partner %edi l The wtI nubers of haLmbotds in the spte is the ffrst nubeor. lhe percentate of ell households iS the 'Mr fn parendlthes. 8 The common notion of female-headship suppose. that if the woman is in union, the male will be considered the head of household. Yet the reverse happens in a quarter of the Jamaican female-headed households. To explore this, we compared the partners' labor force status, income, age and educational level in those households where females with partners were declared to be the head (see Figure IH-2). In these households, the woman had the higher income3 in 39 percent of cases. In 12 percent of cases she had a lower income but was older.' In 9 percent of cases she had a lower income, and was younger, but had a higher education.' In 40 percent of cases, in spite of earning less, being younger, and being less well educated, the female member of the union was declared as head. Thus, in ninety percent of cases of female-headship there is a readily apparent reason why the woman might be declared as head (either she has no partner, or has a higher income, is older, or better educated than he). In the remaining ten percent of cases of iemale heads (Gust 4% of all households), the reason is less clear. The ownership of the dwelling, or the length or strength of relationship with the partner probably influences the assignment of headship. What Do We Mean by Headship? The measure of headship used is the declaration of the household member who was the respondent on the Labor Force Survey'. Because of limitations in the data set, we cannot systematically examine whether the declared head of household contributes more dollars or more hours of work to the family welfare than the non-head of household,7 much less what sort of influence they have on decision-making. Rosenhouse (1989) shows that using a definition of headship based on hours of market labor provided almost doubles the percentage of Peruvian households classified as female-headed. Fox and Paes do Barros (1990) show that in Brazil there is virtually no difference. They also show that use of an income criterion rather than reported headship raises female headship only slightly. 'By at least 10 percent. 'By at least two years. 5By at least two years. 61n order to determine whether the assignment of headship was dependent upon the gender of the respondent we compared the gender of the respondent and the declared head for households with partnered heads. Male respondents reported a male head in 94% of cases, whereas female respondents reported male heads in only 67% of cases. 'Although we did make use of individual income in describing the status of female heads vis-a-vis their resident partners, the Labor Force Survey's income questions do not contain enough detail to support thorough work on relative incomes, especially for the self-employed. Furthermore, hours of work in labor force activities are collected only fo: broad ranges of hours. Without exact values and without information on household chores, a redefinition of headship based on labor contribution in not possible. Figure 111-2: Income, Age and Education In Female-Headed Households with Partners 66 (13,) 95 he has equal or H ( e c ldr higher Incoso _ 119 _*esrrd 109 O2X*eI l X)_ both work _(X 0S5X) 5 (14%1 Inam 29 32 ha7 smor _ ~~~~~he is ole _ 43 hha eout or (8)3 hi8e r irncomc _ 1 h nre edul neither works _(X (19X) 31 she has higher (ax) Incoews8 assoe with l10ie he has equal or 1 25%o) | Partrars l ~12S he works ON fo < x 400 and _P - 0.0000 - - M~~~~~~Ajusud 12 - 0.4926 Head Self-EBnlcyd Other .00 .65 .18 .38 Head Prote/f m ln/C ledSae 32 8.95 .14 .35 H"d Odier Sector .02 .65 .13 .34 24 Poverty Measures and Female-headed Households In the discussion of poverty, we use two poverty lines - J$1669 per capita household consumption, and JS3005 per capita household consumption. These correspond to the poorest 10 and 30 percent of the population respectively. The 30 percent relative poverty line is approximately the same as the absolute poverty line derived for Jamaica in Gordon (1989). We use the Foster-Greer-Thorbecke family of poverty indices (see Annex 3). As the value assigned to the parameter a increases, the sensitivity to income inequality of the Foster-Greer-Tihorbecke index increases. When a=O, the FGT index reduces to the headcount index-that is, the proportion of persons who fall below the poverty line. When a= 1, the FGT index reduces to the poverty gap - that is, the average amount by which a poor person's income falls short of the poverty line. When a=2, the FGT index is sensitive to the distribution of welfare among the poor, with the highest weight on the poorest of the poor. (See Annex m for details.) Using the lower poverty line and per capita consumption, we find 9.04 per cent of people living in male-headed households to be poor, compared with 11.11 per cent in female-headed households (Table IV-3). The difference is statistically significant (t=4.28). The other two measures do not differ significaly by headship. When we use adjusted consumption as the welfare indicator none of the differences are significant for the 10 percent poverty line. For the more generous poverty line (J$3005), however, all comparisons show a higher incidence of poverty among female-headed households. Thus, though the results show a mixed picture, they are by and large consistent with the notion that female- headship is correlated with poverty. Next we will use a multivariate technique to determine whether headshipper se has an independent influence Gn the probability of being poor or whether other household -haracteristics, correlated with headship, are responsible for our findings. Table 1V-3: Poverty and Feunae Headslp 10% Poverty Line 30% Poverty Line Male Female t value of Male Female t value of Welfare Measure Head Head difference Head Head differ Mean per Capita Consumption $7,012 $5,564 14.43 $7,012 $5,564 14.43 Mean P.C. Consumption of the poor $1,230 $1,277 4.90 $1,959 $1,992 6.04 FaT a = 0 (Head Count) 9.04 11.11 4.28 28.08 32.35 5.79 FGP a - I (Poverty cap) 2.38 2.61 1.59 9.77 10.91 3.64 Fara - 2 0.S97 0.911 0.21 4.57_[ 5.08 2.80 1,pw , . , , , N E .................... > ..... . Mean Adjusted Per Capita Consumption $9,369 8,113 10.27 $9,369 $8,113 10.27 Mean Adjusted P.C. Consumption of die poor $1,931 $1,969 1.34 $3,034 $3,043 2.43 FGT a - 0 (Head Count) 9.79 10.26 0.97 29.2 31.0 2.47 FOT a - I (Poverty cap) 2.S9 2.57 0.16 9.78 10.3 1.81 FGT a - 2 0.95 0.93 0.30 4.51 4.69 1.02 26 The Probability of Being Poor: Mufivarlate Analys Using a probit equation, we estimate the probability of falling within the group of poor using the two poverty lines (10 and 30 percent poorest) and the two measures of consumption (per capita consumption and adjusted per capita consumption). The results indicate that, just as for consumption, female-headship has an independent impact, increasing the probability of being poor (see Table IV4). However, the quantative effect is minuscule: the results translate into a 1.4 percentage point higher probability of being poor for female-headed households, holding all other variables constant at the sample mean.'° Furthermore, the effect holds only for the lower poverty line and per capita expenditures as the measure of consumption. At the higher poverty line or using adjusted per capita expenditures, female headship does not raise the probability that a household will be poor. In these estimations, we controlled for the area, rurality, family structure and health status of the household, and the union status, age, education, and sector or work of the head of household. Ihe impact of these variables is very similar to the ones presented in Table IV-2 regarding the deter_min of welfare. Full estimation results are presented in Annex 4. Table IV-4: Probit Models of Probabifity of Being Poor ------ -Emd Hbip Dummy -Dcjx=d6*:V~~~~~~~~~~~~~iajcao -.- - - .--- - - -.-. . 10% Poverty Lhis-, Capita xpend 26 1.37 -.02- 10%. ;overty Lino-A . S:.Cd '." .81. ::,':,_'42. ; ,, _ . _. !. 30s Pbvoty i.r .2-9 -e Apeni f =.: -u rcruuo reni............SS=: ,S-- 30% :.-0-k- - - .0 :-- - ~~~131 .19 ------. ---: . - . . - - - - Wris is for dte 10 percent poverq line nd per capita consumption, the case for which the result is more pronounced (Table IV-1). Table IV-S: E bdky of Poverty Mmure by Gender of Household Head FEMALE-HEADED HOUSEHOLDS MALE-HEADED HOUSEHOLDS Value of 81nck wr 8Sowit Value of Lki* wn B=icit pove Meawigty t PavOty mean wit Poveq Mealut Measure Consumpdon bnequalkly Measrt Consumplion /aquk POTa -O 11.1 -2.7 6.2 9.0 3.1 10.0 PUTa - I 2.6 -3.3 10.9 2.4 -2.3 13.2 POT a - 2 0.9 -3.7 15.4 0.4 -3.3 19.0 POT a - 0 32.4 -1.3 1.1 28.1 -1.1 1.7 PGT a - 1 10.9 -2.0 3.5 9.8 -1.9 4.8 POT a - 2 5.1 -2.3 5.7 4.6 -2.3 7.7 Para -o 10.3 -2.8 ; .9 9.8 -2.8 7.1 PGTa - I 2.6 -3.0 9.3 2.6 -2.8 | 10.7 Par@a - 2 0.9 -3.S 13.6 0.9 -3.S | 16.1 FGT a - ° 31.0 | 1.4 | 1.1 29.2 |-1.4 | 1.4| FC:T a - ' 10.3 -2.0 3.3_ 9.8 _ _0 4.1 PGT& 1 2 4.7 -2.4 3.4 4.5 _ -2.3 6.6 28 Finally we estimated how changes in the level and distribution of consumption in Jamsica would change the poverty index (following Kakwani [1990aJ). The elasticity of poverty with respect to the mean consumption level shows how poverty would change if the whole welfare distribution were shifted up or down by a uniform amount, e.g. if the mean consumption rose but the Gini coefficient stayed the same. For Jamaica, a change in mean consumption level is not systematically different for female- and male-headed households (see Table IV-5). That is, in periods of growth, poverty among female-headed households will decrease in proportion to that of male-headed households. Conversely, during periods of falling consumption levels such as occurred with structural adjustment in the mid-1980s, female-headed households are apparently not more vulnerable."1 The elasticity of the poverty measure with respect to the distribution of welfare -hows how poverty would change if the total resources in the economy were constant but their distribution among households changed, e.g. if the mean consumption were constanc, but the Gini coefficient fell. For Jamaica, the change in poverty caused by a change in inequality is less in femalk-headed households than in male headed-households (see Table IV-5). That is, they are somewhat protected against further skewing of the income distribution, but would gain proportionately less if overall inequality were reduced. Female-headship as a Targeting Indicator Governments concerned with poverty will try to target the benefits of some of their social programs specifically to the poor. Because it is difficult and expensive to assess the households' welfare levels accurately, many programs are designed to reach persons or households with characteristics that are known to correlate strongly with poverty. In spite of the difficulties in defining female-headship, it is sometimes used as a criterion for eligibility of such programs (e.g. the United States' Aid to Families with Dependent Children and the Honduran Single Mothers' School Coupon). It is therefore interesting to assess how well the gender of the head of household will do as a proxy for welfare in targeting programs. When using a proxy for poverty, one is concerned with two types of errors: errors of exclusion, that is the failure to reach some of the poor, and errors of inclusion, the leakage of some of the benefits to the non-poor. Here we have tested the use of two proxy indicators - residence in a female-headed household and residence in rural areas - for both the 10 and 30 percent poverty lines based on per capita consumption. The results show that in Jamaica, the gender of the head of household is not a useful proxy for income in targeting. If residence in a female-headed household were used as the targeting criterion with the poverty line set to capture either the poorest 10 or the poorest 30 percent of t!" populatiun, just half of the poor would be reached. In contrast, if residence in a rural area were used as the targeting criterion, only 13.1 percent of those below the lower poverty line would be exluNded, and only 20.3 percent of those below the higher line (see Table PV-). In all cases the errors of inclusion remain high; even for rural targeting, 56.8 per cent of the benefit would accrue to the non-poor will benefit using the lIlt should be emphasized that in this conclusion, the growth or depression are assumed not to change the distribution (or structure) of the economy, which may run counter to fact. If so, the change in poverty will need to consider both the elasticity with respect to the mean and the elasticity with respect to the distribution. 29 higher poverty line. Note that the results for targeting on female headship are very close to what one would expect from targeting a random sample of the population. For a 10% poverty line, nearly 90% of benefits given to female-headed households accrue to the non-poor. For the 30% povertq line, two- thirds of the benefits given to female-headed households accrue to the non-poor. Table IV-6: Famale Bleadshp a a Tureing Tool Proxy Indicator for Poverty Female Headship Resident in Rurl Area Targeting Outcome 10% 30% 10% 30% Poverty Poverty Poverty Poverty Line Line Line Line Errors of Exclusion 49.6% 51.3% 13.1% 20.3% Errors of Inlusion 88.1% 67.6% 84.3% 56.S% The relative merits of using female-headship or rural residence as a targeting criteria are illustrated by setting up a simple hypothetical government program. Suppose that the government has a fixed budget to help the 3860 households in our sample, say JS 811,000 (about four tines the amount received by these households from the food stamp program). If the program gives the benefits equally to all persons in female-headed households, each person will receive JS 115. Alternatively the program could dispense l$ 96 to all persons in rural households. The latter will reduce poverty more than the former for more combinations of a and poverty lines (see Table IV-7). Table IV-7: Poverty Outcomes wnth Transrers Befor After Trasfer to individuals in: Program FHH Rurd 10% Poverty Line FGT a = 0 9.98 9.29 8.91 a= 1 2.48 2.17 2.03 a= 2 0.90 0.76 0.68 :30%Poverty Linc - FGTa = 0 30.01 29.12 29.98 a = 1 10.28 9.74 9.54 a= 2 4.80 4.45 4.28 Note: Program budget is J$ 811,000, distributed to aU individuals in feade-headed households (3$115 per persn) or to dal individuals in rurl households (JS 96 per person). 30 Conduion The simplest compaison of mean per capita consumption levels between male- and female-headed households shows female-headed households to be markedly worse off than male-headed bouseholds. Adjusting ependitures for household composition greatly lessens the force of the conclusion. Introducing controls for other household charteristics, and for the human capital characteristics of the head shows that the gender of the household head itself has a still smaller impact on household consumption levels. When focusing on poverty itself, the differences between male- and female-headed households diminish still further. Differences by gender of the head of household in both poverty measures and the probability of being poor are sensitive to the poverty line set, the welfare measure used, and the poverty measure used. While most of our results indicate that the probability of being poor is somewhat higher for female headed-households, the difference with male-headed households is small. Indeed, targeting social programs to female-headed households will not be a successful way of reaching the poor; better proxies for poverty are available. If the main cause for concern for female-headed households is the expectation that female headship is highly correlated with poverty, then this concer can be put to rest in Jamaica. 31 V: Differentia Resource Use So far we have explored whether the economic resources available to female-headed households differ from those in male-headed households. In this section we will concentrate on how the households use the resources available to them. There are two types of resources: time and money.'2 TIme Use In the stereotypical male-headed household, the male head works for income and the female partner spends time on household activities that produce welfare for the family, and especially for the children. The stereotypical female head of household must manage to both provide income and carry out welfare- enhancing household activities. With only one person filfilling both roles, the time constraint is likely to imply fewer hours devoted to one or the other of the important activities. We already saw that female heads are considerably less likely to be employed than their male counterparts (50 percent versus 79 percent, see Table M-4). Female heads of households' income-generating act vities are thus lessened. If there is also a difference in labor force-participation among women of woLing age (15-64) between women who are heads and who are partners of heads, then female-headed households would be constrained in their welfare generating household non-market activities more than male-headed households. We estimated a stndard labor force participation equation (see, for instance, Deaton and Muellbauer, 1980, Chapter 11) that shows how participation is determined by age, education, numbers of workers in the household, health status, region, non-earned income and headship. The results of the last two variables are of particular interest Female headship does have an independent effect on labor participation, but the impact is small. The estimaton results" icply that for a woman with average characteristics, but without any non-earmed income, the probability of being in the labor force is 64.5 percent if she is the head of the household versus 57.9 percent if she is the spouse of the head. We would expect that the need for female heads to work would be lessened by higher remittances received by female heads of household. This is borne out by the data. Non-earned income does reduce labor force participation. (see Table V-i). Moreover, households with a female head receive more in the form of non-earned income (see Table V-2) - JS1567 per household as opposed to J$1134. We will look at the implications for chfldren's welfare of this relatively small but statistically significant difference in time use between female heads and other women in Section VI, but first we will look at the use of the second resource: money. In particular, we will analyze to what extent male- and female-headed households show different consumption patterns. %n the remainder of this paper we will use a standard neo-classical model of household behavior as our analytical framework. This model is presented in Annex 4. "See Annex 4 for fill estimation results. 32 Table V-1: Non-Employment Imn and Es_mad Fanme Labor Fore adcaipatioin for Women 1544" Female Femle Heds Spouscs No Oduer- iome 64.5 57.9 Domestic Remittaces - $150 63.5 56.9 Remittance from abroad = $650 Food Stamps = $60 Domestic Remittances = $300 62.6 56.0 Remittances fom abroad = $1300 Food Samps =120 _ Domestic Reamittces = $450 61.7 55.0 Remittan fom abroad = $1950 Food Samps = $180 Table V-1: Non-Employment Income Received by Male and Fane-headed owsebolds Female-baded Male-headed Hiousehlolds Households D1made Remittan J$212. JS136. Reminanie fiomn abroad 703. 582. Food Sftmps 61. 41. lstun 62. 92. Cbld Supp 398. 153. O.her 131. 130. Total J$lS67 J$1134 Consumption Pattens First, we looked at broad consumption categories of food, housing, daily expendiures, durable goods, non-consumption expenditures and other consumption exendiures by quintile and gender of the household head.' There is litde difference by gender of the household head (see Table V-3). The 'For full esfimation results see Annex 4. "'Daily expenditures' are frequently purchased items such as food and bevecages consumed away from home, coal, kerosene, wood, personal care items and tobacco products. Non-consumption expenditures are for insurance, taxes, loan payments, charble donations, transfers to relatives 33 similarities are especially marked in the poorest quintile, where consumption differences might have the most marked impact on child welfare. These results are confirmed by a regression of the foodshare on the logarithum of total expenditures, the squared value of this variable, the number of children, adult males and adult females in the household and regional dummy variables. Female headship does not have an independent effect on the total food share. (See Annex 4, Table A4-6 for complete estimation results). When we look at food expenditures in more detail, however, we find differences in the combinion of foods purchased. Expenditures on such higher quality food items as meat, vegetables, milk and other dairy products tend to be higher in female headed households, all other things equal. Expenditures on alcoholic beverages are lower. (See Table VA4, and Annex 4 for details) These results are in broad agreement with those of Horton and Mi'ler. the household, and expendiures on weddings and funerals. Other oDnsumption expenses cover a whole gamut of items for the household, clothing, tasport, inment, etc. '6 The food share includes the value of explicit purchases plus food produced at home or received as gifts. Table V-3: Consumptlon Share by Gender of Household Head CONSUMPTION QUINILUS USIG PEt CAPrIA CONSUMPTION ALL aENDER OF HEAD | ENDER OF HEAD GENDEIR OF HEAD GENDER OF HEAD GENDEt OF HEAD GENDER OF HEAD I_________ FEMALE MAEI FEMALE MALE FMALS E . FEMALE MALE FEMALE MALI FEMALE FOOD 4S 47.0 45.2 429 424 311 319 . 340.4 RWr _ '1i. 11.9 IJJ II 133 12.4 14 14 1 35.3 17.5 14.1 13.9 DALY EWPSND 15.1 15.2 15.9 16.S 17.4 17.5 19.1; 17.3 193 16.4 18.0 16.7 SEMI.DURABLE CONSUMOIN 213 21.6 21.6 23.1 21.7 23.9 f-U;2, 24.0 25.7 26.2 23.2 24.0 NON- CONSUMFION ECPENDrrURES 0.7 07 13 1.5 2.3 6 1. 2.5 8.0 5.5 4.2 2.6 DURABL GOODS .... IA.4 1.1 1.7 1.6 2.3 2.1 ±.7 : 3.2 3.4 33 2.6 2.4 TOTAL 100 100 100 100 100 100 100 100 100 10 100 100 35 Tibia V4: Food Ite Saar. md Fae.e Heumip FEMALE EADW I | COOElFICIE!NT |I TSrAMS'lC Ois an FM. .23 |12 Plour -.14 -1.19 Brea& -.42 -1.56 Sur -.01 0.03 Rice .02 0.08 Pouhy -.32 -1.04 Soup, .29 3.89 Yama -.62 1.36 Conmewa .04 0.48 Milk .96 2.13 Fish -.46 -1.29 Otr ceak -0.50 1.57 Fnut -.23 0.97 Vegtablae 1.13 4.20 Aoholic Bveage -4.12 -11.71 Condiments .40 2.69 Non-Aboholic Bevam -.18 |0.85 Other}Fbood .71 2.02 OtherDairy Producs 1.13 3.93 Meat 1.64 3.31 Total Food Show .01 0.96 36 We also lookred at the share of expenditues for so called child goods (shoes, clotiing and lodthig materia for children, baby food and education expenditure) for households with children. The mean JOI of household expediture do wot vary significantly by gender of household head - from 1$ 1637 in female-headed households to J$ 1684 in male-headed households (t=0.58). When we estimated die share equations for children's goods, female-headship did matter. After controlling for the households' expenditure level, region and household structure, the share of children's goods in total consumption is higher for female-headed households though the difference is tiny - 4.7% on average rather than 4.3% in male-headed households (see Annex 4 for full esimation results). Finally, we looked at expenditure for domestic help which we expeted to be especially high for female-headed households. The results came as a surprise: 18 percent of male-headed households reported such expenditures versus only 12 percent of female-headed households. Average expenditure levels were JS485 and J$294, respectively. Of course, part of this result stems from the fact that more male headed households are in the upper tail of the income distribution. But it is only among the poorest 10 percent of the populaton that femaIe-headed households are more likely to spend on domestic help tha other households. Among the poor, however, the numbers are very low: 0.37 percent of male headed households versus 2.49 percent of female heads. In sum, we do find evidence that female heads use their time differently than other women and that consumption patterns differ between male- and female-headed households. The difrerences, however, are so small that they are unlikely tD have a discernable impact on child welfare. This issue will be taken up in the next section. 37 VI: Children's Access to Social Services Tbis section explores the children's access to social services by looking for differences by gender of household and gender of the child. If statistically significant differences are found in average access to a service, then multivariate techniques are presented to learn what causes the difference"7. Health Services Preventive Care. There are no significant differences in children's access to preventive care by gender of the household head, whether all children are considered together, or the analysis is performed separately for girls or boys. This is true using as the definition of preventive care the percentage of children for whom clinic visits for preventive care were reported in the six months preceding the survey, coverage for individual vaccines, or complete immunizaon coverage. In only one of the eighteen cells tested was a significant difference found, hardly a systematic result (see Table VI-1). Curative Care. For those children who were reported to be ill in the four weeks prior to the survey period, just over half of them received medical care (see Table VI-1). Again, there were no significant differences in access by gender of the head of household or by gender of the child. 171n fact, multivaiate techniques were used throughout, but are not reported here to simplify the presentation. 38 Tabic VI-1: Children's Access to Health Care Children of Both Sexes Boys Girls Male Female Male | Female Male Female Head Head Head j Head Head J Head % receiving t stat _receiving t stat % receiving t stat Preventive 20.2 | 18.5 0.84 19.5 18.1 0.51 20.9 J 18.9 0.68 VaccinationsL i =_= > /3 OPV 92.4 91.4 0.54 92.3 91.2 0.52 92.4 91.7 0.35 >3 DPT 92.7 91.8 0.61 92.3 91.3 0.50 93.1 92.3 0.39 BCG 97.6 97.1 0.51 97.0 98.7 1.53 98.2 95.5 1.96 Measles 90.7 90.6 0.07 90.5 90.5 0.00 90.9 90.6 0.11 Fun 83.8 84.1 0.16 84.3 83.4 0.32 83.2 84.9 0.57 Immunizat.on Consultation 54.9 53.0 0.39 51.0 53.0 0.29 58.7 53.1 0.82 IWhen Child (Last 4 1/ Vaccination coverage is figured for children 1-4 yeaws old. Use of preventive care is for childnC 0-5 years old. * Significant at 5% level. Education Enrollment in Secondary Education. hzzause enrollment of primary age children in Jamaica is nearly universal, we concentrate on the enrollment of secondary age chiUdren. Just over half of chUdren from age 13-19 are enrolled in school. Children mostly stay in school to age 14, but after that, enrollment rates drop sharply with age. This is partly the result of drop-outs, and partly the result of a multi-track school system. Only the highest quality schools are intended to hold students until they have completed 13 years of schooling. The lowest track of secondary ends after only nine years of schooling. As apparent from Table VI-2, girls of secondary age have higher enrollment rates than boys. Compared to the gender of the child, the difference by gender of the household head are smaller. We explore these differences furither using probit models to predict the probability of enrollment for various household characteristics. 39 Table VJ-2: Enrobuent Rat. by Gender or Child and Household Head _____________ ~~Boys Gizrl-s _ _ _ Male Female t Male Female t Age of Child Ed Head Statistic Head Ead Statistic 7-12 Yeus 98.8 98.8 0.10 98.0 99.1 1.51 13-19 Yeas 52.8 52.0 0.30 60.1 57.2 l.00 Controlling for other factors, the gender of the head of household and of the student are significant determinants of enrollment (in both the substantive and the statistical sense). Girls are more likely to be in school than boys, regardless of the gender of the head of household. Boys appear to be better off in female-headed households, but girls' enrollment rates are higher in male-headed households. (see Table VI-3). Although this may be of some concem, girls in female-headed households are still noticeably more likely to be in school than boys. Keeping boys in school, regardless of gender of the household head, is of higher policy concern than keeping children in female-headed households in school. The other factors controlled for in isolating the effect of the sex of the child and household head were the age of the child, the family welfare level and the education of the parents. All are important determinans of the probability of eaollment. An additional year of education for the father matters slightly more than an additional year of education for the mother in raising the chance that the child will be enrolled. This is, again, a result somewhat the reverse of that expected. Nutrifion Prograns Both of the large nutrition programs in Jamaica are tied to use of other sevices. School feeding programs obviously reach only those in school. lhe food stamps program is tied to the public health care system. The access to nutrition programs may, therefore, follow the same pattern as found for the use of health and education services. 40 Table VI-3: Estimated Probabilities of Enrolhnent of 13-19 Year Olds Based on Probi± Regression' Boys Girls male Femade m ale Female Head Head Head Head Age 13 98.4 98.9 99.4 99.3 14 94.8 96.1 97.9 97.3 1S 75.8 79.7 86.3 84.1 16 59.8 64.3 73.6 70.4 17 24.0 28.2 37.8 34.2 18 6.7 8.6 13.5 11.5 19 4.0 5.2 8.7 7.3 Per Capita Consump.ion 1669 41.6 46.8 57.3 53.6 3005 46.1 51.3 61.7 58.0 4000 49.3 54.5 64.7 61.1 6000 55.6 60.7 70.4 67.0 8000 61.4 66.3 75.4 72.3 12000 71.4 75.7 83.2 80.7 School Feeding. Jamaica has two main school-feeding programs. The Nutribun program distributes daily to schools centrally prepared portions of milk and a bun or cake. The program is concentrated in urban areas or schools with good road transport to allow the daily delivery of the products. The "traditional' program consists of subsidies and commodities distributed to schools which prepare cooked meals on the school grounds. The allowances are usually less than necessary to deliver a good meal every school day. Although the Nutnbun program also has some problems achieving its daily delivery goals, on average it feeds children more regularly. Some schools also have snacks, breakasts, or other forms of food available under a variety of community and state-supported programs. Overall, slightly over half of the enrolled children aged 7-19 receive some kind of school feeding (see Table VIA4). There is no difference in access to some kind of school feeding by gender of the head of household, or of the child. Children in female-headed households are more likely to be in the "'For full estimation results see Annex 4. 41 Nutribun program than children in male-headed households, who are more likely to be in the cooked meal program.' 'is is most likely due to the fact that as shown in Table M-1, female-headship is higher in urban areas than rural areas, where Nutnbuns are the prevalent form of scbool feeding, whereas cooked meals are served in mral schools where female-headed households are less concentrated. Table VI-4: Children's Access to Nutrition Programs Both Sexes BS B Girls Program Male Female t Male Female t Male Female t Head Head Statistic Head Head Statistic Head Head Statistic Schiol Feeding - Children 7-19 Eurolled in School . = _ _ Nutribums 26.5 31.3 3.06 27.5 29.8 1.08 25.5 32.7 3.26 Cooked Meal 21.0 17.2 2.82 21.2 17.3 2.09 20.8 17.2 1.89 Odher 5.S 4.3 2.09 5.5 4.3 1.15 6.1 4.2 1.80 None 46.5 47.2 0.38 45.7 48.4 1.16 47.5 46.0 0.63 lF6od-Stamps - Children 0-4 14.0 12.5 0.94 12.7 13.1 10.17 15.4 11.8 1.50 ood Stams. The food stamps program has two parts, a means-tested portion for poor households, and a maternal-child portion. The maternal-child portion is linked to public primary health centers. Any child under age five is eligible for food stamps upon presentation of the child's birth certificate at a pnmary health center when a food stamp officer is present to conduct the registration, which is done on a limited number of days per month. The food stamps are then collected once every two monfs at the health center. Participation in the program therefore requires traveling to the clinic and standing in line periodically. Where time constraints are particularly severe, this may discourage participation. At the time of the November 1989 SLC, the child's allotment of food stamps was J$20 per month. Overall, about 13 percent of children age 0-5 received food stamps in November 1989 (see Table VI-4). There were no significant differences in receipt by gender of the child, or by gender of the head of household. Summary We have considered children's access to preventive and curative health services, primary and secondary education, school feeding programs, and food stamps. No differences in access were found for health care. For secondary school enrollments, we found that female-headship marginally increased the chances of boys, but did produce lower enrollments for girls. Overall, however, girls rates exceed those of boys. For nutrition program-, the gender of the head of household produced no difference in access either to 9These conclusions hold for aU children age 7-19 and for enrolled children 7-12 as well. 42 food stamps or school feeding, though it was correlated with the type of school feeding program received, probably indirectly through location. We have found little evidence thatJamaican children in female-headed households fare any worse in terms of access to social services than children in male-headed households. This is perhaps not surprising since economic welfare is usually a strong deteminan of social service use, and female- headship and povety are not strongly linked in Jamaica. Any barriers to access that might be imposed by the time oDnstraint faced by female-headed households seem to be minor or, at least, to have been surmounted. We have so far focused on the social service-provided inputs to producing good child welfare outcomes. In the next section we will look at the actual welfare outcomes themselves. 43 VU: Child Welfare Outcomes One of the most compelling reasons to be concerned over the issue of female-headship is the possibility that, if they face tighter income and time constraints, these may imply poor child welfare outcome. This would be undesirable not only for the present welfare loss, but because the early child welfare outcomes are also strong influences on the later human capital upon which the children will have to rely to earn their livelihood later in life. In this section, we focus on child welfare outcomes. Health Children get sick more often than adults. Their susceptibility to illness is increased if they are malnourished or if their families are unaware of, too busy, or too poor io practice basic health practices that help prevent and treat common intestinal and respiratory upsets. Diarrhea arnong Under Fives. Diarrhea is one of the most common childhood maladies. Preventive measures such as infant breastfeeding, access to potable water or boiling impure water, cleaning vegetables adequately, and good personal hygiene are effective in decreasing diarrhea's occurrence. Proper feeding practices and rehydration can shorten its duration and minimize its impact on overall health status when children do have diarrhea. The incidence of diarrhea is therefore a good indicator of how effectively the household produces health. Overall, about 6 percent of children under five years of age had had an episode of diarrhea in the two weeks prior to the survey. From the means, it would seem that children in female-headed households have diarrhea more often than children in male-headed households. Considering both sexes of children together, about 8 percent of children in female-headed households had diarrhea in the two weeks prior to the survey, while the figure was only 5 percent for children in male-headed households (see Table VII-1). Given the alarming nature of this result, we examine it further using a probit regression to predict the probability that a child will have had a recent episode of diarrhea based on a number of factors. Table VlI-1: Health Status of Children Age of 0-4 Both Sexes Boys Girls Repoting Male Female t male Female : Male Female t Head Head Stat Head Head Stat Head Head Stat Diarrhea in 4.82 7.68 2.35 4.61 7.65 1.81 5.04 7.71 1.50 last 2 weeks I_I_I Ilness in 26.9 25.5 0.61 25.2 24.9 0.08 28.8 26.2 0.79 last 4 wes . _ _ 44 Table Vl-2: EstiFated PlobabiIily of Children's Diarrhea* Boys Girls Male FemAle Male Female Age in Months Head Head Head Head 0-11 5.1% 8.7% 5.6% 8.1% 12-23 7.0 11.7 7.8 10.9 24-35 2.4 4.6 2.8 4.2 36-47 3.1 5.7 3.5 5.3 48-59 0.8 1.6 0.9 1.5 *For fiull estimation results see Annex 4. In the multivariate probit, neither the gender of the child, nor the gender of the head of the household is a significant determinant of the probability that the child will have diarrhea. It would therefore seem that the difference in mean probabilities of diarrhea found using only bivariate techniques is explained by the correlation with headship and the age of the children. After controlling for age of the child, neither the per capita consumpdon variables, nor parental education matter. Family structure has some influence, though the pattern is not clear. A lower number of adults in the household is weakly correlated with diarrhea. Presumably with fewer adult hours in which to provide child care and to "produce' health, the child may be more susceptible. Although the number of young children in the household is not a significant determinant, more children age 6-10 is positively associated with the incidence of diarrhea. (see Annex 4). The most important determinant of the probability of diarrhea is the age of the child. The probability rises from zero to two years, when it peaks, which is consistent with results found in many countries, as weaning has taken place for almost all children by this stage with its concomitant exposure to food and water borne organisms to which the child has not built up resistance (see Table VH-2). Incidence of Illness among Children 04. Next we consider all reported illness among children 0-4 years of age reported in the month prior to the survey. About a quarter of children were reported as having been ill (see Table VII-1). Looking at the mean incidence, there are no significant differences in the incidence of all illness among children in female-headed households vis-a-vis those in male-headed households. Nutrition Nutritional status is a very sensitive and important indicator of child welfare. Malnourished children are more susceptible to illness and are less able to respond to stimuli and learn in either formal or informal 45 settings. If their growth is reduced, they will have less muscle-mass with which to perforn work as adults (McGuire and Austin, 1987). Here we consider three measures of nutritional status. Weight for age is a general summary measure. Low weight for height, or wasting, is an indicator of acute malnutrition. Low height for age, or stunting, is a cumulative indicator of chronic or repeated episodes of malnutrition which disturb a child's growth pattern. The reference norms used here are from the United States' National Center for Health Statistics. Each child's height and weight have been converted to its z score - the number of standard deviatons of the reference norm specific for age and sex. We present both the scores as an indicator of satisfactory development, and the percentage of children who fall below the cut-off points and are considered malnourished.' Of the thirty-six combinations of definitions of nutritional status, gender of child, and gender of the head of the household, thirty-four show no significant differences by gender of headship (see Table VU-3). For the other two - the z score for height for age for both sexes of children and moderate wasting for girls - children in female-headed households appear to fare less well than children in male- headed households. In order to explore these results further, we performed regressions on the z score for weight for age, height for age, and weight for height. The female-headship dummy was not significantly different from zero in any case. Aside from the gender of child and head combinations, we controlled for per capita consumpdon, age of the child in years, area, household age structure, parents' education, and union status of the head of household. In general, the results were quite inaccurate, with right hand side variables predicting less than four percent of the variation in nutritional status, presumably because we were missing the important factors of the parents' heights and weights (see Table VII-4). 2 Children are moderately wasted if their weight for height is from 70 to 80 percent of the reference norm, and severely wasted if their weight for height is below 70 percent of the norm. Children are moderately stunted if their height for age is from 85 to 90 percent of the reference norm, and severely stunted if their height for age is below 85 percent of the reference norm. By World Health Organization standards, children are classified as moderately malnourished if their weight for age is between 60 and 80 percent of the reference norm, and severely malnourished if their weight is below 60 of the reference norm. By the Gomez standards, children are mildly malnourished if their weight for age is between 75 and 90 percent of the reference norm, moderately malnourished at 60 to 75 percent of the norm, and severely malnourished below 60 of the norm. 46 Table VII-3: Gender of Head and Nutritional Status of Children Age 0-4 Both Sexes Boys Girls Male Female t Male | Female | t Male Fedale t Hand EHead | Stat Hed Head Sta.t Head Head stt Weight fbr Age - - - - - - - zScore -0.14 -0.29 1.8 -0.11 -0.28 1.4 -0.18 -0.3 1.2 mild malnut-Gomez 25.2 25.8 0.29 25.9 26.7 0.24 24.3 24.9 0.17 mod. mlnut-Gomez 1.8 3.2 1.81 1.7 2.4 0.77 1.8 3.9 1.73 sev. malnut- Gomez 1.0 0.5 1.14 1.2 0.7 0.68 0.8 0.3 1.00 mod. malnut-WHO 5.8 7.0 1.00 6.5 6.6 0.07 5.0 7.3 1.36 sv.malut-WHO 1.0 0.5 1.14 1.2 0.7 0.68 0.8 0.3 1.00 Hgh .-.for A- - .-.. .. ,. zScore -0.05 1-0.251 2.1* 1 .01 .28 1.9 I -0.13 -0.22 .9 modeat estlmting 1.4 1.8 T063 1.4 2.0 0.57 1.3 1.6 0.30 sevee stuting 1.1 1.5 0.68 1.2 1.5 0.34 1.1 1.6 0.64 Weifor Hei '.,-'g.-.h-,-','t z Score -0.14 |-0.13 -0.13 -0.19 -0.16 -0.37 |-0.08 | .1 |.1S modewasting 1.11 1.8 1.07 1.71 1.0 10.88 1Q5 2.6 2.33* severe wasting 1.4 1.0 0.67 1.7 1.2 0.55 j1.1 .8 0.38 47 Table VHI-4: Gender Dnmies ii Nutritional Stabs of Children 04 Regrons DUMMY VARIABLE REGRESSION WEIGHT FOR WEIGHT FOR HEIGHT FOR AGE HEIGHT AGE Boy-Male Headed Household Omitted Omitted Omitted Boy-Female Headed Household -.09 .09 .27 (.81) (.81) (-1.88) Girl-Male Headed Household -.10 -.10 -.15 ________________________ (-.99) (-99) (1.18) Girl-Female Headed Household -.11 -.11 .24 (-.93) (1.10) (-1.70) Adj. R2 .08 .03 -.01 Note The z score for die child's height/weight is the dependent variable. Al regressions have the same regressors: a constant per capita consumption; dummies for the 4 combinations of gender of child and head; dummies for age in years; dummies for area number of members of the household 0-5, 6-10, 11-18, and adults; the education in years of the mother and father, and dummies for heads union satus t statistcs are given in parethesis. For full regression resuts see Annex 4. 48 Education In lieu of a measure of cognitive achievement which is not yet available2' we use three measures of education outcome - repetition, daily attendance and enrollment in the most exclusive of the tracks of secondary school. Repetition. Repetition is a sign of poor educational performance. It is costly to the state to have to provide instruction for repeaters, and costly to the family to have to support a student through repeated years in school. Table VII-5 shows the prevalence of repetition among secondary students in Jamaica. The table suggests that prevalence is more linked to rurality and age than to the gender of student or of household head. Only among teenage boys it there a significant difference by gender of the head of household. Boys of age 13-19 in female-headed households repeat less than those in male-headed households. Probit analysis controlling for area, gender of student and head, age of student, per capita consumption, parental education and presence in the household, access to transport, and type of school, confirm the notions. The influence of per capita consumption and of rurality is strong (see Annex 4). For boys, those in female-headed households repeat less often than do those in male-headed households. For girls, gender of the household head is insignificant. Table Vl-S: Percentage of Repeaters among Children Enrolled in Schoosl SEX OF THE PERSON _____ BOY GIRL l Male Female t stat Male Female t stat Head Head Head Head All Children 6.0 5.3 0.67 5.8 5.6 0.26 Kino M.A. 4.0 4.6 0.33 6.3 3.6 1.28 Odher Towns 8.6 4.1 1.59 7.1 6.6 0.18 Rural reas 6.2 6.1 0.06 5.3 6.4 0.68 Age of te Stdentiyes _ - _ _ 7-12 Years 5.1 [ 6.3 0.87 5.5 6.0 0.35 13 -19 Years 7.5 3.6 T 2.2-i 6.4 T 4-9 0.86 2Note that cognitive achievement tests for children enrolled in school were part of the November 1990 Survey of Living Conditions, for which data are currendy being processed. 49 Table VII-6: Percentage of Children with Full Attendance During the Previous Week, Among Children Enrolled In School Sex of the Student Boys Girls Male Female Male Female Head Head Head Head % % % % All Children 86.4 87.1 88.2 87.9 Area- Kingston M.A. 93.2 91.8 95.5 94.3 Other Towns 87.6 89.1 86.3 90.1 Rural Areas 83.4 83.8 86.3 829 Age of the Student, in Years -- 7-12 years 85.4 86.6 87.9 86.4 13-19 Years 88.0 88.1 88.8 90.3 Nce: Fuflaimeadace is defined as not having mised a school day duig the week preceding the mamy DailyAttendance. Low daily attendance is common in Jamaica. It is an education policy concern because it significantly reduces the educational contact time of the absent children. Furthermore, having to accommodate the pervasive absenteeism can reduce the speed at which the teachers can advance through the year's curriculum for the whole class, thereby lowering the overall quality of education of even those students who do attend regularly. Overall, for the sample of enrolled students in primary and secondary schools, about 85 percent of students reported having attended all official schools days in the week preceding the survey (See Table VII-6). In predicting which children are likely to be absent, neither the sex of the child nor of the head of household matters. This is suggested by the bivariate analysis, and is confirmed by a probit regression controlling for age of the student, per capita consumption, area, distance to public transport, parental education and presence in the household, and the availability of school lunches. Enrollment in High Schools. The secondary school system is divided into several tracks. Three are general curriculum, the others vocational. The three general tracks differ significantly in quality. Placement in the tracks is determined by performance on the Common Entrance Exam which most students take at the end of the sixth grade. All of the schools are state schools, so that the tuition is not a constraint to entrance into the better tracks. The constraints are the student's test score and the geographic access (The higher track schools are not accessible from some runad areas). so Table VII-7: Enrollment in High Sdhools by Gender of Child and Household Head -_____ Boys _ Girls | Male Female t Male Female t Head Head Stat Head Head Stat Among Children 13-19 10.9 14.9 2.13 21.5 18.0 1.55 Among Secondary Students 22.8 30.0 2.00 38.6 34.1 1.17 Girls are enrolled in 'high schools', the most exclusive track of the Jamaica secondary system, markedly more than boys (see Table VII-7). Differentiating by the gender of the head of household shows that among boys, those in female-headed households are enrolled in high school significantly more often than those in male-headed households. Among girls, the differences by gender of household head are not significant. Probit analysis controlling for gender of the student and head, age of the student, per capita consumption, parental education and presence in the household, and access to transport confirms the result (see Annex 4). summar Using a variety of measures, we have explored children's health status, nuritonal status, and education performance. Again, there is little support for the hypothesis that children in female-headed households are disadvantaged. These children are faring well in the present. The current welfare outcomes are also good indicators of the formation of human capital on which they will rely in their adult, productive lives. They should fare well in the futre as well. The result that the welfare outcomes of children in female-headed households are not below those of children in male-headed households is consistent with what we have learned in previous sections of the paper - that poverty and headship are not strongly linked, that female-headed households use their resources in ways that in some cases are more child-oriented than male-headed households, and that their children have equal access to social services. 51 VII: Discussion Review. We started with the premise that female-headed households were doubly burdened by tighter income and time constraints than male-headed households. These constraints could lower the access to social services important to child welfare, and lower child welfare outcomes. A countervailing influence could be a strong preference on the part of decision makers in female-headed households to use their resources in ways beneficial to children. We did a comprehensive empirical exploration of these hypotheses. Poverty and female-headship appear to be linked, but not strongly enough to make headship a useful targeting indicator. Female heads of households work more in the workplace, but the difference in participation rates is only 6 or 7 percentage points. The foodshare in female-headed households is not larger than in male-headed households. Significant differences do exist among expenditures; female headed households, by and large, seem to buy a more nutritious food basket. The results on children's access to social services and child welfare outcomes show very few significant differences between the two types of households. Although in a few cases the gender of headship is statisticaily significant, in some of these cases the influence is favorable and in others unfavorable, so there is not a systematic case that headship affect child welfare outcomes. The results in These results may seem surprising, or out of step with the main body of female-headship literature. They are not, however, unique. In Ghana, for example, for households with children under age 15 and using per capita adult equivalence measures of consumption, female-headed households have higher mean welfare levels than male-headed households (Lloyd and Brandon, 1991). In Peru, the same welfare measures show no differences in welfare levels between households that declare female heads and those that declare male heads (Rosenhouse, 1989). In Cote d' Ivoire, female headed households have, on average, higher welfare levels than their male-headed counterparts (Glewwe, 1989a). The sensitivity of the link between poverty and female-headship to definitions and methods shown for several Asian countries in Visaria (1980). Moreover, to show that female-headship is not a hardship to household welfare in Jamaica is not to dismiss concern for female-headship in other countries. Jamaica, after all, has a long history of women's access to education, and to the early passage of equitable family and labor law (World Bank, 1989). Furthermore, while feminist may find that social customs constrain Jamaican women's actions in ways that they find undesirable, it must be recognized that the social inequalities are several orders of magnitude less than in societies where customs such as purdah or non-Islamic variants prevail. The Jamaican story can be interpreted to show that the actions that are usually advocated to protect women's welfare and that have been more fully implemented in Jamaica than in many countries have, indeed, resulted in better welfare outcomes for female-headed households than found in many other countries. The Limitations. While we have been, perhaps, more than usually comprehensive in our treatment of monetary welfare for the household and health, mntrition and education outcomes for children, there are some issues we have not been able to address, largely for want of data. First, we have not looked at the welfare of the women themselves. Is producing these adequate child welfare outcomes straing them in unacceptable ways? Is their leisure, health, or happiness suffering? We camot say. Women's welfare is, of course, an important outcome in its own right, but may also be important in the 52 sustainability of children's welfare outcomes. Second, we have not been able to address the dynamics of household formation. What causes a household to be female-headed? If women only form independent households when they can guarantee themselves and their children a minimum level of welfare, there may be potential female-headed households sheltering within large extended families or with males to whom the women are tied more by necessity than choice. These households are presumably worse off than independent households, but we don't measure it. Third, how do female heads of households cope? We have shown that the differences in market work, remittances and expenditure patterns are small. The answer may lie in differential patterns of time use for home production, corwmunity participation and leisure, though data for such analyses are scarce. Part of the explanation may be the fairly well developed social sector infrastructure in Jamaica that provides relatively easy access to basic services such as primary education, primary health care, immunization and nutrition services. Policy Lmplications. Our investigation of poverty, welfare and female-headship in Jamaica has two strong policy implications. First, programs concerned with poverty should concentrate on poverty and not be distracted by female-headship. Female-headship is not a usefil targeting proxy for poverty in Jamaica. Similarly, programs concerned with child welfare need not be especially concerned for children in female-headed households. The programs should focus on other issues of access such as ensuring service in rural areas, to children of vulnerable ages, and to the poor. The second policy implication is for those concerned with women's issues in Jamaica. The work presented here suggests that issues specifically of female-headship are not the first priority. Rather the focus should be on issues that potentially affect all women. By ensuring equitable treatment of women in labor and capital markets and in family law, female heads of household will of course benefit, but so will women who live in households headed by men. Methodologcal Implications. Perhaps the strongest lesson from our work is not for those interested in social policy in Jamaica, but for those concerned with female headship in other countries. Much of the gender and welfare literature uses relatively simple statistical analysis. Indeed, much of it is based not on empirical analysis of full household data sets, but on strings of reasoning built on summary staistics from secondary data sources.' These techniques are not sufficient, and may produce misleading results. Our analysis of poverty and female-headship is a good example of the importance of full-scale analysis of primary data. Looking only at mean consumption levels led to the conclusion that female- headed households were very much poorer than male-headed households. Examination of the whole welfare distribution showed that among the poor, female-headed households were not so overrepresented. Multivariate analysis also contributed to the understanding of the determinants of poverty. Finally, a targeting analysis showed that despite the impression created by differences in mean welfare, female- headship is not a viable targeting indicator. Further work in the field of female-headship should seek high standards of statistical rigor and data quality. This implies, of course, the need for household survey data that is recent, publicly 2The bodies of literature fostered by the Population Council and by the International Center for Research on Women are two important exceptions to this criticism. 53 available, and sufficiently comprehensive in scope to cover the range of issues relevant to poverty, child welfare and female-headship. A second methodological implication has to do with the usefulness of the concept of beadship. It is a very blunt instrument to use in chipping away at the intricacies of gender, intra-household bargaining, and welfare. This is particularly true when, as in this paper, headship is used as reported in a survey rather than as a more rigorously defined concept based in indicators of the relative wbargainingw position of the various members of the household. To what extent the results are influenced by the fact that female headed households in Jamaica consdtute a large heterogenous group, rather than being a small fringe group suffering from stigma and open discimination, as is the case in many other societies, is also an unanswered question. Our results should be interpreted with those limitations in mind. 54 Bibliography Berger, Marquerite and Mayra Bivinic. (1989) Women's Ventures: Assistance to the Infbrmal Seor in Latin America. Kwmarica Press. Connecticut. Bolles, A. Lynn. (1986) 'Economic Crisis and Female-Headed Households in Urban Jamaica in Nash and Safa, eds. Women and Change in Latin America Bergin and Garvey Publishers. Massachusetts. Buvinic, Mayra. (1990) 'Women and Poverty in Latin America and the Caribbean: A Primer for Policy Makers", processed. IDB, Washington Commonwealth Secretariat. (1990) Engendering Adjustment for the 1990s Commonwealth Secrariat. London. Davies, Omar and Patricia Anderson. (1987) "The Impact of the Recession and Adjustment Policies on Poor Urban Women in Jamaica" processed. Institute of Social and Economic Research, University of the West Indies, Kingston. Deaton, Angus and John Muellbauer. (1980) Economics and Consumer Behavior, Cambridge University Press. Deaton, Angus and John Muellbauer. (1986) "On Measuring Child Costs: With Applications to Poor Countries", Journal of Political Economy. Vol. 94, No. 4, pp. 720-44. Deaton, Angus. (1989) "Rice Prices and Income Distribution in lhailand: A Non-Parametric Analysis", Economic Journal, Vol. 395, pp. 1-37. Dwyer, Daisy and Judith Bruce. (1988) A Home Divided: Women and Income in the Third World Stanford University Press. Stanford. Folbre, Nancy. (undated) "Mothers on Their Own: Policy Issues for Developing Countries" processed. International Center for Research on Women. Washington. Folbre, Nancy. (1991) 'Women on Their Own: Global Patterns of Female Headship" Gallin and Ferguson, eds. The Women and International Development Annual, Vol. 2, Westview Press. Boulder. Foster, James, Joel Greer and Erik Thorbecks (1984), "A Class of Decomposable Povery Measures.- Econometrical Vol. 52, pp. 761-765. Glewwe, Paul. (1990) "Improving Data on Poverty in the Third World: The World Bank's Living Standards Measurement Study" Policy, Research and External Affairs Worldng Paper 416, The World Banl. Washington. Glewwe, Paul. (1987a) "The Distnbution of Welfare in Cote d'Ivoire 1985" Living Standards Measurement Study Working Paper No. 29. The World Bank. Washington. 55 Glewwe, Paul. (1987b) IThe Distribution of Welfare in Peru in 1985-86" Living Standards Measurement Study Working Paper No. 42. The World Bank. Washington. Gordon, Derek. (1989) *ldentifying the Poor: Developing a Poverty Line for Jamaica" Jamaican Poverty Line Project Working Paper No. 3, Planning Institute of Jamaica. Kingston, Jamaica. Grosh, Margaret E. (1991) The Household Survey as a tool for Policy Change: Lessons from the Jamaican Survey of Living Conditions' Living Standards Measurement Study Working Paper No. 80, The World Bank. Washington. Horton, Susan and Barbara Diane Miller. (undated) "The Effect of Gender of Household Head on Food Expenditure: Evidence from Low Income Households in Jamaica" processed. Departnent of Economics, University of Toronto. Toronto. International Development Bank. (1990) 'Working Women in Latin America" Economic and Social Progress in Latin America 1990. pp. 205-256. International Development Bank. Washington. Kakwani, Nanak. (1990a) "Poverty and Economic Growth: With Application to Cote d'Ivoire" Living Standards Measurement Study Working Paper No. 63. The World Bank. Washington. Kakwani, Nanak (1990b) *Te-'-ig for Significance of Poverty Differences: With Application to Cote d'Ivoire- Living Stan.Jrds Measurement Study Worldng Paper No. 62. The World Bank. Washington. Lloyd, Cynthia and Stacy Brandon. (1991) "Women's Role in the Maintenance of Households: Poverty and Gender Inequality in Ghana", processed, Population Council, New York Massiah, J. (1989) "Women's Lives and Livelihoods: A View from the Commonwealth Caribbean" World Develonment Vol. 17, No. 7, pp. 965-978. Pergamon Press. Oxford. McGuire, Judith S. and James Austin. (198) 'Beyond Survival: Children's Growth for National Development" Assignment Children. 1987. No. 2, UNICEF, New York. Moser, Caroline. (1989) "Gender Planning in the Third World: Meeting Practical and Strategic Gender Needs' World Develment vol. 17, No. 11, pp. 1799-1825. Pergamon Press. Oxford. Paes do Barros, Ricardo and Louise Fox. (1990) "Female Headed Households, Poverty, and the Welfare of Children in Urban Brazil" processed. The World Bank. Washington. Rosenhouse, Sandra. (1989) "Identiing the Poor: Is 'Headship' a Useful Concept?" Living Standards Measurement Study Working Paper No. 58. The World Bank. Washington. Psachapoulos, George and Zafiros Tzannatos. (1991) Female Employment and Pay in Latin America: ARegional Study processed. The World dBank. Washington. Scott, Kinnon. (1990) "Female Labor Force Participation and Eamings: The Case of Jamaica" in Psacharopoulos and Tzannatos, eds. Female Enloyment and Pay in Latin America: A Reional Stuy processed. The World Bank. Washingtn. 56 Silverman, B. W. (1986) Density Estimation for Statistics and Data Analyis. Chapman and Hall. London Statistical Institute of Jamaica. (1990) Consumer Price Indices: New Series Januara 1989-October 1990 Statistical Istitute of Jamaica. Kingston. Statistical Institute of Jamaica. (1987) T'e Labour Force: April 1987 Statistical Institute of Jamaica. Kingston. Statistical Institute of Jamaica and the Planning Institute of Jamaica. (1989) The Survey of Living Condiions: Final Reort Statistical Institute of Jamaica. Kingston. United Nationals Development Programme. ("90) Human Development Report 1990 Oxford University Press. Oxford. Visaria, Pravin (1980) "Poverty and Living Standards in Asia' Living Standards Measurement Study Working Paper No. 2., World Bank, World Bank. (1990) World Develogrment Report 1990. Oxford University Press. Oxford. World Bank. (July 1989) 'Jamaica: Country Assessment of Women's Role in Development' processed. The World Bank. Washington. 57 Annex I: Calculation of Fler Capita Consumption The welfare measure used in this paper is a comprehensive measure including food consumption (46 items purchased, 16 items received as gift or home produced); daily expenditures (fuel, tobacco, foods consumed away from home); consumption expenditures (41 items covering household expenses for clothes, household furnishings and supplies, transport, etc.); non-consumption expenditures (insurance, taxes, weddings, donations, outgoing transfers); home repair costs; rent (actual or imputed); and the use value of durable goods. Different recall periods are used for different items. The values from them were annualized. For some of these items, two recall periods are used on the questionnaire. In all cases the long period was used in the annualization process. For example, for food items, the expenditure in the last week and in the last four weeks is asked. The value for the last four weeks was multiplied by 13 to get the annual expenditure for that item. The imputed rent calculation is based on the regression of renters' rent payment on charactistics of their dwellings (housing materials, size, availability of services, etc.). Then for households that own their dwelling, the dwelling's characteristics are valued from the results of the renters' regression. This synthetic value of imputed rent is used for owner-occupied dwellings. For renters, the actual rent paid is used. For each of the fifteen durable goods, the average depreciation rate is figured from the information on age, value at purchase, and current value. This depreciation rate is then multiplied by the current reported value to derive the use value for the current year. Finally the household's total consomption is divided by the number of household members. 58 Annex II: Ibe Estimation of the Dlstribution of Consumption Expenditures The non-parametric estimation of the density function of per capita consumption was compiled with the kernel method. The estimated probability of consumption level x is given by AX) (h (1) hni., h where: n is the number of observations, h is a smoothing parameter (or 'bandwidth'); K is a kernel function which satisfies the condition f K(x)dx=1 (2) The estimation process consists of constructing a series of "bumps' centered at each observation. The width of each is determined by the smoothing parameter h. The sum of the bumps gives the esimator (for more detail, and illustrating pictures, see Silverman, 11986]). In this paper, we used a Gaussian functional form of the kernel: K(x)= -C 2 (3) Silverman (1986) shows that the value of h which minimizes the mean integrated square error (MISE) of the estimator for a Gaussian kernel is: ch q,, = 1.06 'I, where a is the sample standard deviation. The non-parametric estimation was computed with the software Gauss. The computer program is a modified version of the program used by Deaton (1989). In contrast to Deaton our program computes weighted density; that is each household observation is weighted by the household size. Consequently, the density function is related to the distribution of individuals rather than to the distribution of households. This feature is particularly important since we compare subgroups (female househelds vs. male households) where the mean household size is significantly different. We used a total of 100 points (value of x in formula [1]). The graphs are truncated for JS15,000 since we are interested mainly in the lower part of the distribution. The value of the density is multiplied by 10b. 59 Annex m: Tedhnical Annex oil Poverty Measures The poverty measures used in this paper are the Foster, Greer and Thorbecke (FGT) family of additively separable poverty measures. Their formula is: P I (z-x) n:( WhereN: Total number of households in the sample z : poverty line 7in J$ per year) x;: welfare variable (per capita consumption) n1: number of individuals in household i n: total number of individuals Ij: dummy variable equal to 1 if xc ChI- 0.000 Hud Self-Employed Agriculture .38 3.47 23 .42 Ls Likelthood - -582.44 Hed Self-Employed Oher .07 .54 18 .38 Hed PtafesdA/MmlnlCtdcal/sale -.86 -2.32 14 .35 Hed Other Setor .05 .28 13 .34 Note: Te probabiity of beig poor retft to individuals. Th eathmto Is based on household oberviton. Tabhs A4-2: Probi: RaeSsaulon Resuks DEPENDENT VARIABLEt PROBABILITY OF BRING IN POOREST 10% OF ADJUTD PER CAPITA CONSUMTON DISTRIDUTMONI m .0, gd. d.w. J6. UNIVERSI; ALL HOUSEHOLDS DUMMY VARIABLES CONTINUOUS VARIABLES Ml Head o|lual AgeofI ud -.00 .05 48.3 17.2 FemaleHead .08 .81 .42 .49 Ageof HudSquared/100 .01 1.00 26.5 17.93 One Member-Male -.40 -2.49 .14 .34 Yeare of Educaton - Hud -.04 -1.61 73 230 One Menmber-Pemak. -.06 _ .34 OS .22 Yeo of Education Squared/100 .16 .00 93.2 53.90 HeAd In L ChiP- 0.000 Head Self-Emiloyed Agriculture .27 2.85 .23 .42 Log ULkelihood - -763.42 Head Self.Eayloyed Othr -.09 -.74 .11 .36 Head Protoaa/Admin/Ciedcal/Siae -*65 -2.62 .14 .35 Head OMar Setor .03 .17 .13 .34 Noto: Te proahtUky of btiK pow rffen to indhivdu Tie eMtmadati Is baued on houseold obeAvatdo. Table A4-3: Probit Regreaalon Results DEPENDENT VARIABLEs PROBABILITY OF BEING IN POOREST 30% OF PER CAPITA CONSUMPTION DISTRIBUTION; mean t1, std. dev. .41. UNiVERSE: ALL HOUSEHOLDS DUMbfY VARALES __ CONTINiUOUS VARIABLES Me Head onsirvi = = Age of Head -.02 140 48130 17.20 Female Head .11 1.63 .42 .49 Ago of Head Squared/100 .02 2.03 26.51 17.93 Or. Membet-Mae -.73 -S.35 .14 .34 Yean of Education -Had -.06 .2.99 7.30 2.30 One Member-Female -.39 -2.31 0S .22 Years of Educaton Squared/lOG .03 33 98.20 53.90 Hud In Legal MeMASge onited .34 .47 Yean of Educaion- Max. In HH -.07 4.72 1.31 5.90 Head In Common-law Margle .30 -3,60 .20 .40 Number of Children 0-5 .30 10.10 .56 .91 Head to Vitin or Cul Union .23 2.12 15 .36 NumberofCbii cd6-I0 .22 6.75 54 .14 Head Wowed or Divocced .09 .90 .11 .31 Number ofChildren 1-1a .22 3.55 .77 1.12 Head Never Marred 15 1.62 .20 .40 Nunber of Petona wh Poor Heah .23 3.65 IS A41 ing_oo -.54 -5.45 .21 .45 Conain -.55 -1.64 Other Town 0cafted .1. .39 SUIMMAY STATISTICS Rural Anas -50 6.13 .53 J0 Number of Observationa - 3536 Chi, (23) - 1011.39 Hesd Not Wong omc- - 32 .47 Pob > ChP- 0.000 Heed Sklf-sployed Ag4culure .26 3.31 .23 .42 Lag Ukelihood m -1314.39 Head Self.EmployedOdher .14 1.59 13 33 Head ProfeAu/Mmin/Ckiecal/Sels -.43 -3.23 .14 35 Head Otbr Secto .13 .1.12 .13 34 Noteh Te probabilty of being poor reen to indicldualh. U. esimation Is based on bousehold ebuevalon. Table A44: Pmbit iReRaulon Result DEPENDENT VAIUABLE. PROBABILITY OF BEING IN POOREST 30% OF ADJUSMED PER CAPITA CONSUMPIlON DISTRIBUTION; m. .24, std. dw. .43. UNIVER3S: ALL HOUSEHOLDS DUMMY VARIABLES . CONntiNUOUS VARIABLES _ Mae Hud omaued ASg of Head -.02 -1.69 4830 17.20 14na )luHd .09 131 .42 .49 A of Hud Squar.d/l00 .03 2.80 2651 17.93 One MAenbu-Mae -.44 -4.00 .14 .34 Yenotf Educadoan-Head -.07 -335 7.20 230 One MemberFemale -.11 -.30 .05 .22 Yau ofEducadooSquamd/100 -.12 .1.30 9.20 53.90 H"d Ia laaal Manla oatted .34 .47 Yearsl ofducaa - Maz. ia HH -.05 -336 lIJ 5.90 Head I Commae- w Manhae 30 3.35 .20. .40 Number of aliidreo 05 .11 3.13 .56 .91 Had to V ldg oa Caam Ur .02 .13 A1 .36 NuwiberefChIldn 6410 .04 1.36 _4 .U4 Ho" Wiowe orDinved .05 .55 .11 31 NumberofChLIdreaL-I 1 .16 6.62 .77 1.12 Hud Never Madod .21 2.49 .20 .40 Nuw'e otfetoawia Poaet Hefh .17 2.13 .15 Al1 Klgi -.47 -4.96 .28 .45 C_om -.21 .69 aObeg ToWe o mlid _I .39 StUMAURY SrATISTICS Runl Au .45 6.12 .53 .50 Nuaierof Obueevdoa - 3536 Rad Ngo Woiag* omued .32 .47 PR(b > t- 12.45 Head Slf.Employed Aghekuku .32 4.32 .23 .42 LIg lkellbad - -1538.97 Head 81-EImployed Wu .05 .63 .16 .38 Hed Pu(bIMAdIn/ChlaUlake -.46 -3.70 .14 .35 Head Ow Sectcr -.07 -.72 .13 .34 Note: The probiltly of being poor rtof to ldIviduals. The ealtloa Is based on housebold obse,vloba. . Table A4-S: Probit Rogresion Result |DIlP1NDLNr VARIABLE: PROBABILITY OF PARTICIPATING IN THE LABIOR FORCEI mea .60, dtd. dev. .49. UNWERSE: FEMALE HEADS OF HOUSEHOLDS AND FEMALE SiPOUSElS, AGE 15-64 :~ t1 .- o CS t a M ean S d, y. l ... .Std. . ... ...-'''. Male Head omitted . . Domestic Remittances (OOO'S OF J$) -.02 -.42 .13 6.58 Female Head .l7 2.59 .S2 .50 iaternatlonal Remitnces (000'S OF IS) -.04 -2.13 .56 19.02 Kingston .01 .13 .31 .46 Food Stamps (O0O'S OF JS) .02 .11 .05 1.47 Other Towns omitted .18 .39 Inters (OOO'S OF J$) .15 1.66 .07 8.28 Rural .12 1.S6 .SI .50 Child Support (OOO'S OF JS) .02 .78 .30 12.05 Number of Children 0-5 -.01 -.29 .36 .81 Other Uneaend Income (OOO'S OF J$) -.01 1.33 .08 8.66 Number of Children 6-10 .03 .86 .70 .89 Age .1S 7.92 40.60 12.20 Number of Children 11 18 -.00 -.13 .99 1.19 Age Squared/100 -.18 -8.11 17.98 10.39 0% Number of Aduls -.04 -1.00 2.49 1.31 Number of Employed Persons In Household -.05 -1.04 .70 1.06 Yea of Education-Head .01 .67 7.56 2.29 Own Health Poor -.32 -6.78 .41 .71 Yea of Education Squared/100 .14 1.66 Number of Persons with Poor Hath -.55 -4.09 .10 .35 =___=___=_ Consant -2.73 -6.48 = SUMMRY STATISTICS Number of Obsrvations - 2189 Chi1 (20) - 203.32 Prob > Chit M 0.000 L,, LIkelibood - 1372.12 Table 4-6: Pood Sbae Regreaaloue to la Feale Tell Toala Number Number Number Dummy Dummy HeadAbip EzpeaCilureA Expeadiure of Adull of Aduh of for for Rutal "atoo Comm Dummy (000's of Ji) (OO0 of 3J) emale Mal chidreti Klagac Ameas Adj. Rs mue Oils ad Fas 13.50 0.23 -5.05 0.00 0.05 0.52 0.03 0.03 0.02 0.09 5.74 (.44) (1.51) (1.3) (0.04) (1.14) (2.62) (1.15) (0.65) (0-.1) ll"ar L".99 0.14 -0.94 .0.01 0.13 0.17 0.2s 0.13 0.1 0.19 26 .-35) (.1.12) (1.47) (.0.31) (4.95) (4.74) (14.12) (1.3) (2.33) Breads 41.73 40.42 .6.02 0.24 0.01 .0.01 0.13 .0.03 0.17 0.05 6.56 (5.90) (0.156) (.4.14) (3.27) (0.13) (4.12) (4.51) (4.12) (0.9) utaw 53.77 0.01 46.21 0.26 0.06 0.10 0.26 0.16 0.54 0.1u 4.21 (9.09) (0.03) (-.09) (5.33) (1.69) (2.14) (10.67) (1.26) (43I) Rke 7.65 0.02 40.46 .0.09 0.19 0.14 0.21 0.61 a.2m 0.07 4.5 (1.52) (0.06) (4.44) (.1.61) (3.19) (2.5) (7.57) (3.93) (.132) Poukay 46.46 .0.32 5.66 0.13 0.21 0.22 0.34 0.55 025 0.17 4.74 (5.72) (-1.04) ..4) (1.54) (2-22) (2.30) (7.36) (am) (1.1) go" 0.03 0.29 0.32 .0.02 .0.01 0.06 .0.01 -O.2 0.06 0.03 0.94 (5.16) (3.93) (4.16) (2.62) (4.4) (.1.17) (0.3) (3.-9) (0.02) Yaom 2.1 .0.63 1.64 .0.15 .0.30 0.30 0.16 *1.12 130 0.11 4.9 (.24) (1.86) (.90) (.1.60) (2.90) (3.03) 0.22) (4.11) (7.13) Corsea I4.50 0.04 -2.96 0.12 0.01 0.06 0.4 0.11 0.05 0.03 07 (6.26) (0.48) (.5.53) (4.76) (I. 1) (2.44) (2.7) (1.55) (0.71) Mlk .43.63 0.96 10.36 0.46 .0.03 0.00 -0.27 0.05 4.10 0.06 I1.97 M. (2.15) (4.27) (.3.72) (4.21) (4.-04) (4.03) (4.14) (4-30) Fisb 443 0.46 -7.15 .0.37 0.03 .0.07 4.21 *1.07 .0.56 0.01 3.36 (4.71) (.1.29) (.3.72) (3.76) (0.24) (4.71) (4.4) (-3.72) (.2.17) OGe ceeas 30.21 .0.50 -5.34 0.29 -0.04 .0.13 00 -.49 1.21 0.02 5J (3.66) (1.57) (3.15) (3.31) (4.43) (.1.44) (4.00) (-5.92) (5.17) Falia *-10.55 0.23 2.17 -0.07 40.16 .0.12 .0.27 0.17 -0.03 0.03 5.53 (.1.7) (0.97) (1.70) (.1.03) (.2.21) (l.71) (-7.77) (0.39) (41.5) Vgdablea 16.34 1.13 *2.67 0.16 4.25 0.03 -0.17 0.00 -0.75 0.03 5.62 (2.32) (4.2) (11.4) (2.21) (.2.99) (0.42) (.4.33) (.0.02) (.3.91) Aeabolio Beverages .20.90 -4.12 4.74 .0.21 0.32 .0,62 4.37 0.5 -0.42 0.07 2.61 (-2.27) (11.71) (2.50) (.11) (3.97) (.5.96) (-7.21) (0.52) (1.60) Coodlmeau -13.09 0.40 2.56 40.09 40.09 -0.03 -0.10 0.16 .0.19 0.06 2.74 (-3.33) (3.69) (3.13) (.2.26) (.2.01) (.1.39) (-4.64) (-1.34) (.1.80) Noo-AlcoboleBDeveragea -13.02 0.23 2.33 40.07 .0.06 4.14 -0.19 40.09 0.02 0.03 2.09 (-2.29) (4.84) (3.00) (.1.23) (4-.83) (.2.16) (.5.19) (4.50) (0.12) Oaser Foods -33.77 0.71 3.03 .0.36 4.23 .0.06 0.47 1.57 0.05 0.06 6.17 (-4.32) (2.02) (4.27) (-3.76) (-2,13) (4.59) (9.16) (5.57) (0.25) 5 tiber Dairy Producu -1630 1.13 3.60 -0.12 4.32 4.13 4.10 0.53 4.40 0.07 7.24 (.2.19) (3.93) (2.32) (-1.47) (-3.70) (-1.59) (-2.43) (2-30) (-1.90) Mea -11.10 1.64 1.44 0.07 4.22 0.17 -0.22 -1.15 0.33 0.06 10.28 (4.s5) 3.31) (0.54) (0.53) (-1.43) (1516) 0.3.06) (-2.37) (0.90) ToWal Food Shbae 0.01 0.01 0.16 4.01 0.01 0.01 0.02 0.05 0.04 .29 40.04 (0.04) (0.96) (3.91) (4.16) (3.62) (4.22) (13.49) (0.39) (6.2) Mes .47 9.35 97.65 .94 1.29 1.35 .29 .54 Sid. Dev, .49 .32 15.97 1.41 1.47 2.02 .45 .50 Nole: sm i l psteAheAls Me I Values. Table A4-7: OLS Regression Results DEPENDENT VARIABLE: SHARE OF CHILDREN GOODS IN TOTAL CONSUMPTION; mean 4.95; std. dev. 4.27. UNIVERSE: ALL HOUSEHOLDS Vasiable coefficient T v*luea MeaI Std. D:v. Variable CoefficienT t valus Mean SSd. D DUMMY VARIABLES CONTINUOUS VARIABLES Male Head omitted - In (Per Capita Expenditure) 5.97 2.22 10.05 .72 Female Head .59 1.82 1.48 .50 In (Per Capita Expenditure)1 -.31 .11 101.47 14.47 Kingston 1.29 .25 .29 .45 Number of Adult Women -.3S .08 1.16 1.57 Other Towns omitted - Number of Adult Men -.16 .08 1.39 1.66 Rural -.03 .23 .55 .50 Number of Children 0-5 .99 .08 .87 1.00 Number of Children 6-10 .73 .09 .82 .92 ________ _ Number of Children 11 18 .60 .07 1.17 1.20 __________ ________ Constant -25.18 11.09 __________ ________ ______ SUMMARY STATISTICS ______________ _____=___=_ Number of Observations = 2465 F (10,2455) - 33.08 Prob > P .000 Adjusted R2 = .115 I~~~ . = - ! ~~~~~~~~~~~~~~~~~Table A4-8: Probit Regreuon Results0 DEPNDET VRIBLE PRBAITCOF BEING ENROLLED; mean j S,td.D~. d offcv. JO.a ea 8dDe. UNIVERSE: CHILDREN AGE 13-19 DUMMIY VARIABLES CONTINUOUS VARIABLES 3oy-Ma1e Headed Household omited Per Capita Expenditure .09 7.7S 5532 4730 Boy-Femal Headed Houhold .13 1.28 .24 .43 Per Capiu Expenditure - Squared x 10' .99 -4.58 5297S 199104 Old-Female Heeded Household .30 2.90 .24 .43 Father - Yeam of Edcadon .04 4.1! 5.75 3_t2 Oli-Maeb Heuded Housold .40 4.17 .24 .43 Mother - Yen of Edueetion .03 2.24 6.66 3.28 Ap 13 onIed .12 .33 Miles to Bus Stop __03_-2_17 1_2S 2 _74 Age 14 -.52 -2.31 .s .36 Constant 1.24 5.40 Age 15 -1.45 -7.00 14 34 Age 16 -1 91_ _9__ 1 Age 17 -2.85 -13.94 .15 .36 Age 18 -3.64 -16,86 .14 .3S Age 19 -3.90 -17.57 .14 .35 __ Kingston omittod - .28 .45 __ OtherTowns -18 -170 .18 .38 _ Rural -.05 -.61 5S .50 Mother in Household 22 2 7S .65 .48 SUMMMY STATISTICS Father in Household 04 _ 47 .39 .49 Number of Observations -2488 ChO (18) = 1618.32 - - - _ Prob > Chi = 0.00 _______ Log Likelihood = -900.49 ==- - - _ Table A4-9: Pobit Regresslon Results DEPENDENT VARABLE: PROBABILITY OF HAVING DIARRHEA IN THE TWO WEEK PRECEEDING THE SURVEY; mean .063, dId. dew. .U2. UNIVERSE: CHLDREN UNDER AG ES Vad~b1. ~ ., CoefteE~ t vekaeu MeM ~taL bev. VaA. ..ff .i ....ui en id Dv DUMMY VARIAILE CONTNUOUS VARIABLES Boy - Mu Headed Houehold oniluted . _ Yer of Education o Mother -.00 -.06 8.43 2.36 Boy - Pnalb Hcded Household .29 1.69 .26 .44 Yeo of EducAtion - Father -.02 -157 6.96 3.30 Oid - Female Headed Houshold .24 1.46 .24 .43 Pcr Capits Expeaditure .07 .99 | aw Male Headed Houshold .OS .33 .24 .43 Per Capia Expeaditum- Squared -.01 -1.53 Age 0-1 monha omitted Nunber of Children 0.5 .04 .74 2.06 1.15 Age I yaer .77 3.57 .19 .39 Numberof Childrn 6-10 .15 2.94 .94 1.04 A Zge2year _ _ .94 4.48 .20 .40 Number of Chidn 1-18 -.0') -.98 1.10 1.29 Age3yeaf .44 I.96 .20 .40 Number of Adults -.08 -2.09 2.99 1.67 Age4year .JS 2.52 .20 .40 Co asa -1.75 -4.39 K.goa omited ' .26 .44 OthrTown_ -.21 -1.25 .17 .37 _ = | Rund _ -.33 -2.34 .57 .50 Head In Legal Marriage omied . .38 .48 Head In Common bw Marriage -.30 -2.19 .30 .46 _ Head In visltlng or cral union -.54 -2.58 .11 .31 SUMMARY STATISTICS eH"d Widowed or Divorced -.53 -2.21 .09 .28 Number of Observadoas - 1530 ChI2 (11) - 70.26 HeadNever Marded -.16 -.90 .13 .33 Pwb > ChP - 0.00 Log ULkelihood - -323.59 I~~~ * m == _- , Table A4-10: 01. Regrssion Results DEPENDENT VARIABLE: Z SCORE FOR WEIGI!T FOR AGE; mean -.22, std. dev. 1.51. UNIVERSE: CHILDREN UNDER AGE FIVE DUMMY VARIABLE CONTINUOUS VARIABLES Boy - Male Headed Household omitted Per Capita Consumption .04 3.40 4708 3898 Boy - Female Headed Household _.09 -.81 .26 .44 Number of Childrn 0-5 -.04 -1.14 2.06 1.15 Girl - Female Headed HousehoW -.11 -.93 .24 .43 Number of Children 6-10 -.02 .0.61 .94 1.04 Girl - Male Headed Household -.10 -.99 .24 .43 Number of Children 11-18 -.05 -1.65 1.10 1.29 Age 0-12 Months omined - Number of Adults 05 1_87 2_99 1_67 Age I year 1.11 9.30 .19 39 Yea of Education 00 43 286 ____________________ ______ ~~~~~~1Mother____ Age 2 year .13 1.07 .20 40 Years of Education - .00 .0 6.96 3.80 Father Ago 3 year .09 .74 .20 .40 Constant -.62 -2.78 Age 4 year .02 .13 .20 .40 Kingston omitted .26 44 Other Towns .05 .46 .17 .37 Rural .07 .76 .57 .50 _ Head in Legal Marriage omitted .38 .48 Head in Common-law Marriage -.01 -0.07 .30 .46 Head in visiting or casual union -.11 -.75 .11 .31 SUMMAY STATISTICS Head Widowed or Divorced .02 .13 .09 .29 Number of Observations - 1549 Head Never Married -.13 -0.94 .13 .33 F(20,1528) = 7.75 - ~~~Prob .Z F = 0.00 Adjusted RI - .08 , ~ n=., . Table A4-11: OLS Regresion Raults DEPENDENT VARIABLE: Z SCORE FOR HEIGHT FOR AGE; mean -.17, std. dev. 1.81. UNIVERSE: CHILDREN UNDER AGE FIVE Variable : ' 6 :'.,; . ' '1Coffici Ia v. i i t ean: Sd.y. DUMMY VARIABLES _ CONTIUOUS VARIABLES Boy - Male Headed Household omitted = _ Per Capita Consumption .02 1.76 4708 3898 Boy - Female Headed Household -.27 -1.88 .26 .44 Number of Children 0-5 .01 A6 2.06 1.15 Girl - Female Headed Household -.24 -1.70 .24 .43 Number of Children 6-10 -.05 -.97 .94 1.04 Girl - Male Headed Household -.IS -1.18 .24 .43 Number of Children 11-18 -.06 -I.S4 1.10 1.29 Age 0-12 Months omitted - Number of Aduls .04 1.19 2.99 1.67 Age 1 .46 3.09 .19 39 Year of Education - .02 .92 8.43 2.86 | _______________________________ ____________ M oth err____________ ______Mothe Age 2 -.13 -.89 .20 40 YewaofhEducation- -.00 -.23 6.96 3.80 Age 3 .11 .77 .20 .40 Constant -.26 -.93 . Agp 4 .06 .41 20 .40 _ Kingston omitted .26 .44 _ Other Towns .03 .23 .17 .37 _ I~~~~~~ . -.I Rural .16 -1.28 S7 .50 Head in Legal Marriage omitted .38 .48 SUMMARY STATISTICS Head in Common-law Marriage .01 .09 .30 .46 Number of Observations ____________ _a "- , 1528) - 2.17 Head in Visiting or Casual Union -.14 -.81 .11 .31 Prob > P = 0.00 t _______________________ --Adjusted R' -.01 Head in Widowed or Divorces .19 1.05 .09 .28 Head Never Married -.09 -.52 .13 .33 Table A4-12: OLS Rcgreion Results DEPENDENT VARIABLE: Z SCORE FOR WEIGHT FOR HEIGHT; mean -.13, std. dev. 1.28. UNIVERSE: CHILDREN UNDER AGE FIVE y .... 1..-11:0s * ............. . %>t l-- DUMMY VARIABLES CONTINUOUS VARIABLES Boy - Malo Headed Household omitted - - - Per Capita Consumption .02 1.99 4708 3898 Boy - Female Headed Household .04 .37 .26 .44 Number of Children 0-5 -.04 -1.26 2.06 1.15 Girl - Femca Haded .11 1.10 .24 .43 Number of Children 6-10 .03 .78 .84 1.04 Household Girl - Male Headed Household .07 .81 .24 .43 Number of Children 11-18 .01 -.23 1.10 1.29 Age 0-12 Months omitted Number of Adults .03 1.S3 2.99 1.67 Age 1 .71 6.87 .19 .39 Years of Education - -.00 -.30 8.43 2.86 -_______________ _ -_-_----___M other Age 2 .29 2.85 .20 .40 Years of Education - .01 -.61 6.96 3.80 > _________________________ ~~~~~ ~~~ ~~~Father __ _ __ _ Age 3 12 1.17 .20 .40 Constant -.72 -3.69 - - Age 4 .09 .86 .20 .40 Kingston omitted .26 .44 Other Towns .10 .98 .17 .37 = Rural .22 2.S4 .57 .50 - Head in Legal Marriage omitted .38 .48 SUMMAY STATISTICS Head in Common-law Marriage .02 .30 .30 .46 Number of Observations = 1543 Head in visiting or casual union .10 .77 .11 .31 F (20, 1S52) 3.66 ________________ ______ a - ~~~~Prob > F = 0.00 Head widowed or divorced .00 .02 .09 .28 Adjusted R2 - -.03 Head Never Married .03 .23 .13 .33 Table A4-13: Probit Regrssion Results [DEPENDENT VARIABLE: PROBALITY OF REPETITION IN SCHOOL; mean .058, ntd. dev. .233. UNIVERSE: CHILDREN AGE 13-19 X Y Wt'l"'S'o''fi'cn".t'' Y" ale M' SW' DIv.. .. ra ec ' au M':1 Si' ' Dt 1. DUMMY VARIABLES _ _ CONTINUOUS VARIABLES Boy-Male Headed Household omitted - .28 Per Capita Expenditure .05 2.84 5702 5011 Boy-Female Headed Household -.44 -2.30 .23 .42 Per Capita Expenditure - Squared -.00 -1.27 - - Girl-Female Headed Household -.24 -1.36 .24 .43 Father - Years of Education .00 .15 6.11 3.73 Girl-Male Headed Household -.05 -.35 .25 .44 Mother - Years of Education -.02 -.99 7.05 3.15 Ago 13 omited .22 .41 Miles to Bus itop .01 .71 1.27 2.60 Age 14 -.21 -1.15 .27 4S Constant -1.79 -4.60 Age 15 .09 .52 .20 .40 _ Age 16 -.10 -.SI .20 .40 Age 17 .34 1.57 .08 .27 Age 18 1.00 3.37 .02 .15 Age 19 .00 .00 Kingston omitted 2S 4AS Other Towns .43 2.26 .17 .38 Rural .38 2.26 .55 .50 Private School omitted SUMMARY STATISTICS - - Number of Observations - 1310 Public School -.05 -.18 .96 .30 ChP (18) = 41.8S Mother in Household .02 -.14 69 46 Log Likelihood = .26920 Father in Household -.21 -1.42 .41 .49 * AlU of the 14 children of age 19 have a value zero for the depedent variable. They have been dropped from this regression. Table A4-14: Prbbi Rasso Rh | DEPENDENT VARIABLE: PROBALITY OF FULL ATTENACE IN RFENE SCHOOL WEEK mte 89, st. de. Jil. UNIVEPRSB: CHILDREN IN SCHooI, AG8 13 19 ~ .SA,Ey, 'hi 4A~ 6 Cofiiot V*U IeM8. DUMMY VARIABLECS CONTINUOUS VARLABLES IoyMAIb HNeded HH omitted . .28 . Psi Capita Expenditur -.01 -.24 5702 5011 Doy-Feomb Headed HH -.06 -.42 .23 .42 Per Capitu ExpenditireiSqured .00 .76 . - OidIFemalo Headed HH .02 .15 .24 .43 Fther - Ymn of Education .02 1,35 6.11 3.73 aid-Male Haded HH .04 .32 .2S .44 Mother - Yeaz of Education .01 .M 7.05 3.15 AP 13 omited .22 .41 MUes to Dusop -.05 .3.03 1.27 2.60 Ag 14 *.08 -.61 .27 .45 Cooam 1.03 3.07 APis .20 1.30 .20 .40 = Ap 16 -.02 -.17 .20 .40 ____ _ Ap 17 -.0S 2-.41 0 .27 _ A_s_S .OS .18 .02 .15- AP 19 _ .32 .62 .00 .00 _ Kingstonandted =2 U 45 II OtherTowns -.14 -.89 .17 i3 ___ _ _= Rum) .30 -2.11 .55 .50 Fe School Lunch -.61 .1.50 SUMMARY STATLIICS Paid School Lunch .36 2.34 Nun*r of Obsewaom - 1280 Mother la Houehoid .03 .31 .69 .46 al (18) - 41.86 - ~~~~~~~~~~Ptob >. CM' a-0.00 Father in Household .14 .1.18 .41 ,49 Lg oL od - 43132 Private School omited PubUi School .22 .94 .96 .30 _~ - __ Table A4-15: Probi Regresion Results DEPENDENT VARIABL.:; PROBALITY OF BEING IN HIGH SCHOOL; mean .30, rtd. dev. .459. UNIVERSE: CHILDREN AGE 13-19 - ENROLLED IN SECONDARY SCHOOLS lhl~~~Of~1 I auaMai st,d, ,Vrlb1 t.ol DUMMY VARIABLES _ CONTINUOUS VARIABLES Boy-Male Headed Household omi#d .28 Per Capita Expendture .10 9.19 5702 5011 Boy-Female Headed Houswehold .24 2.19 .23 .42 Per Caph Expenditure - Squared/1000 -.97 -4.30 . . Girl-Peale Headed Household .31 2.87 .24 .43 Father - Year of Education .01 1.45 6.11 3.73 Girl-Maie Headed Household .45 4.71 .25 .44 Mother Ye of Education .01 1.11 7.05 3.15 Ag 13 omitted .22 .41 Miles to Bus dop -.04 -2.27 1.27 2.60 Age 14 -.07 -.72 .27 .45 Constdan -1.54 -9.42 . Agp 15 .12 1.23 .20 .40 Age 16 .15 1.55 .20 .40 co Age 17 .33 2.28 .08 .27 Age18 _.90 3.54 .02 .15 _ _ Age 19 .06 0.17 .00 .00 Kingston omitted . .2s .45 . Other Towns .26 2.56 .17 .38 Rurl -.14 -1.53 .55 .50 = = Mother in Household -.02 -0.28 .69 .46 SUMMAY STATISTICS Father In Houwhold .06 .68 .41 .49 Number of Observations = 1681 Chis (18) - 381.7 -_ a Prob > Chp' =0.00 Log Lilelihood - -900.82 Table A4-16: Probk Regression Resuls DEPENDENT VARIABLB: PROBALITY OF BEING IN HIGH SCHOOL; mean .16, std. dew. 367. | UNIVERSB: CHILDREN AGE 13-19 -- , 3,~~~~~~~~~~~~~~K~i; W i DUMMIY VARIABLES CONTINUOUS VARIABLES Boy-Mak Headed HH omitted Per Capita Expendiure .09 8.65 5532 4730 Boy-Female Headed HH .23 2.18 .24 .43 Per Capita zxpnd. - SquaredIooo -.88 -4.39 52975 199104 Girl-Female Headed HH .39 3.67 .24 .43 Father - Yean of Education .02 2.07 5.75 3.82 Girl-Mate Headed HH .49 5.23 .24 .43 Mother- Yean of Education .01 1.10 6.66 3.28 Age 13 omitd - .12 .33 Miles to Bus stop -.03 -1.85 1.28 2.74 Age 14 -.08 -.70 .15 .36 Constant -1.55 -9.21 = Age 15 -.02 -.14 .14 .34 ___ Age 16 -.12 -1.09 .17 .37 Age 17 -.57 -4.62 .15 .36 Age 18 -1.03 -7.0S .14 .35 Age 19 -1.70 -8.36 .14 .3S Kingston omited - .28 4S Other Towns .01 .S .18 .38 Rural -.23 -2.68 SS .50 Mother in Household .10 1.27 .6S .48 SUMMARY STATISTICS Father In Household .02 .24 .39 .49 Number of Observations - 2488 ChP (18) - 381.7 -_ Pmb > ChP = 0.00 Log Likelihood - -906.23 I~~ . = -- - - Distributors of World Bank Publications ARCMDIA ldl NelielOlum KENA S0UHAPRICA.1O01MAA CmuomHlhSRSL 416iultSbast AMC 3UkSu.IUAJIJiLL Fh xlkw GalCarGuam.. Can QUaru _ u __puustmest adUd vIqms mldil.1b.WuXrOl4cI4 P.C. b3.. 45e 1n 3,Almn ENLAND NlAk P. son lb14 Abmdagn K*k-pp. Cap. Teown AUSTRALIA.FA NEW GUINUA, P.0Q3lo 128 KO3IA. RWUUCOF L3SOLOMOlIANDU6 SF4X1I1 HelduM 10 Paaa HoeX okCaapodm Fore_aewiuw VANUATU,ANDWISRUNSAMOA P.,QBm%11.KwWdhm lxAuom sws,im 0*. BDk&Jaanwb TRANC Sagd P.CL Box 4100 6415WMiham bod Wank Pubica tionsG b4Ihau 31 64awmumdl'm MALAYS 1 ob,a U_114 Piu d SPAIN AUSTRIA CURtANY Pa 112 17.3.1 Panad Sam MundP-Pnm _ I6bS GMdd WA C. UNO.Vw1sg 5WO Kuulal.myr caiiI.37 C_b=31 Ai hSS 2KW1 Mdii A.1011 WX EICO WnPOEC 2wuto" iJUDCAM PANECLAOESH HONC KONG MACACO Aauwo GftbI2-810 COnMld.Ct,311 MluoIa,duirlmDhwiupmuM Au. 2800 Lad. 1KW 1Askpu2. M.tD 00000dov.. A_9caSadqeyO!DAS 4&aWymdIAb.S Ham& Rds WIuaala.. NgrmiiA SKILANKA ADIDH ALD1V DhnRmalit/Ar 2ZdF De looal LabHon..0b DhiaZIM CmmluHenglag P.O. 32.. P.0.11246 7430ASH.Abum 100.UrOamemp.bmA. Bmh UWINDIA Grb_wmUm 156.NwAewdSwak llisdPubaluas PUtLa. NEWZALAND Coo2 CdIbp4111 151 mount End ECOONZLIL ,gMadm- G002 PilvaNsw5q i4 SWEDI' 761KDJLAku NawMaima Fw Kniua910 maet An| ad IbmFd~d.r 15iL Hedi. Ug .aapb ba 1Kx UULOWNi Nailedo Aa. IuIA5-012Sbhh )MR De LA.aoy fAmb.y-400 UniveyPe.mkdaid Av. du iA 20 T.Cfwmmildlogd Fh 2 mnuudu 1114 AfAll End Pdvab JSIqs W u .-A N.wDdhl-100 IbudaP. Q.doxD3 CANADA S-I 8 Solos LeDtfluurf_ 17kal imAwnm. NORWAY C.P.M2SftBcm.A.im.e Caktt.-70OW2 Nwninm hfa.C SWIIZELAND DbmdaulDQn1s BlokDepamc F s J4ESSU joydeva Hosed uP. So. 412SS bad L&,ilP Payee S01 R Umplu-SOO1 0.r10de0 _u CHM B-sla-5fi0e CH IMLamamem IDeWtclCSSA. PAXOSTAN Amw6mVxNd.3ftIU6 35-5112 Kadipia Mir- BookAgmm7 Fdh U _ S~_ Crss Rod E5Sh..Quid.eAzue Lhrabu.ftyet Hydw.bmd- go07 P. 0aN. .728 S kmd 0vou LahxeS4 Camepous13im ahbRnE PI _UDRF162dFo a3 I= ZL_~ PuIaialtrnado m N _Tblmarlm N_.6m Ni0w1Lmu, mu I.D0FoSDogia A_u'dmi.d-301009 EdIDulDinrU.SA TANZANA ago Apdeo3U4 OdoidUniviyPam P.U.1a Hous. LIM1 P..Q11.53 COLOMBIA I4As.kuag Msbtab Roed lidedcLkL Ld2SOw-I01 PiUWEdIS Dar,Sa m Ap1mbdoAwvo3G= li.uhmdEndokCmWW Dope 01 Clobalazwlrold SuI15C tyadw 10 THAILAND 40 SamaiNapinCeAluDITewl IpFUR01310t01 cT D'mvo Naa440I 01 AymIaAuG C- I;LV.dds 306Sio Road Cmdr. d'Ecdomettl.Diful.ai CamsEtw il Smaleak AhidnmCA) INDONBA U UbtmuoMma 04 P.sF Pt Uida. Undxd TRINIDAD &TONAGC6ANTZMJA Abuljoa0gPlPate la!neebudur1 POLAND SARSUDAIBARNADDIL P.O. BDo 1 Inag1Aso PlblbSuwlm DOMWCA. CRlXADA.GUYANA. CWYRUS jalit1 ULPIaba3l/37 JAMAICA. MONStAT. SL CmehdAppleditimmak 00477Wax,was .7TS4kN9VIS.S`LLVCA cmCAl IRLAND 3f. VDECN GT&ND1S 4Obb_9,111_mI C _SSq4U"~7 Feid.af -bieid~ Syulo9FUmStdt .. Uo.2ll00 45Hamourtd Isiournell epWafbts Hin. O DubNI2 ULO0aa3 cutpx 02416 W mTiamidad. Winthmu DENMrK ia _ __lifillitsilillw Y_ont Liub lLad PORTUCAL TURKEY DlC410DkdUbaEgC TiAwA,4l4S50 RuxDOoCwm7O74 NoimpSobNM.IS 1 'LiD m Ciposg DODJCAN UUUUC ITALY bul EdftmTa_ w.C.porA- LkosaCa m .abmiSOPA SAUDIOARAAQATAR 3mtumadIee ili.al 09 Vi.. Du3. DI CaMelsD I/ il,b*s.kSk9e UNITED KINCDOM Apw%"d.dCIff25U0Z-1 ca.dkpinazI P. B=30.1 S]aubhLW. Saab Dimbap 81128 Pk El~~~~~~~~~ydma IWI 7.0.5.. BGWPT,ARA5 WUUJCOF JAPAN SINCAPOU.LTAIWAN. EIrland AlAba.. EmawrUookSm MYANMAUtWI AlJCrlSa Hmp34Oa3Bukye,b 11 inaonaUPhbSmdm VENEZUELA Cal Tokyo Plinalm.Lf. Lbuadladb CeMddWblidb6Mg Aped.. 4. mflaq fPddb I D C 100A 5 g L3 LSMS Working Papers (continued) No. 59 Labor Market Performance as a DEterminant of Migration No. 60 The Relative Effectiveness of Prwate and Public Schools: Evidencefrmn Two Developing Countries No. 61 Large Sample Distribution of Seueral Inequality Measures: With Application to Cte d'lwire No. 62 Testing for Significance of Poverty Differences: With Application to Cote d'lwire No. 63 Poverty and Economic Growth: With Application to Cote d'lvoire No. 64 Education and Earnings ini Peru's Informal Nonfarm Family Enterprises No. 65 Formal and Informal Setor Wage Determination in Urban Low-Income Neighborhoods in Pakistan No. 66 Testing for Labor Market Duality: The Private Wage Sector in Cote d'luoire No.67 Does Education Pay in the Labor Market? The Labor Force Participation, Occupation, and Earnings of Peruvian Women No. 68 The Composition and Distribution of Income in C6te d'lwire No. 69 Prwce Elasticities from Survey Data: Extensions and Indonesian Results No. 70 Efficient Allocation of Transfers to the Poor The Problem of Unobserved Household Incno No. 71 Investigating the Detenninants of Housdeold Welfare in C6te d'lwire No. 72 The Selectivity of Fertility and the Determinants of Human Capital Investments: Parametric and Semiparametric Estimates No. 73 Shadow Wages and Peasant Family Labor Supply: An Econometric Application to the Peruvian Sierra No. 74 The Action of Human Resources and Poverty on One Another What We Have Yet to Learn No. 75 Te Distribution of Welfare in Gkana, 1987-88 No. 76 Schooling, SkiUls, and the Returns to Government Investment in Education: An Exploklion Using Data from Ghana No. 77 Workers' Benefits fron Bolivia's Emergency Social Fund No. 78 Dual Sdection Critenia mth Multiple Alternatives: Migration, Work Status, and Wages No. 79 Gender Differences in Household Resource Allocations No. 80 The Household Surveyasa ToolforPolicy0Change: Lessonsfromn the Jamaican Surveyof living Conditions No.81 PatternsofAgingin ThailandandC6ted'Ivoire No. 82 Does Undernutrition Respond to Incomes and Prices? Domninance Tests for Indonesa No. 83 Growth and Redistribution Components of Changes in Poverty Measure A Decompoition with Applications to Brzil and India in the 1980s No. 84 Measuring Income fron Family Enterprises with Household Surveys No. 85 Demand Analysis and Tax Reform in Pakistan No. 86 Poverty and Inequality during Unorthodox Adpstment: The Case of Peru, 1985-0 No.87 Family Prductivity, Labor Supply, and Wdfare in a Low-Income Country No. 88 Poverty Comparisons: A Guide to Concepts and Methods No. 89 Public Policy and Anthrpomnetric Outcmes in Cote d'Ivire No. 90 Mearing the Impact of Fatal Adult Illness in Sub-Sahan Africa: An Annotated Household Questionnaire No. 91 Estimating the Detmninants of Cognitie AcvemEnt in Low-Income Countri The Case of Ghan No. 92 Emnomic Aspects of Child Fostering in Coe d'lvoire No. 93 Invesfrnent in Human Caitl: Sdcooling Supply Constmints in Rural Ghana No. 94 Willingness to Payfor the Quality and Inknsity of Medical Carec Low-Income Househols in Ghana No. 95 Measurement of Returns to Adult Healtk Morbidity Effects on Wage Rates in Ce d'Ivoire and Ghana