A POLICY RESEARCH WORKING PAPER 1275 Female-Headed Households, Fema!headed households - a growing segment of Poverty, and the Welfare 8razilian households - are a r w_ . . v s . ~~~~~~~~~~~heterogenous group)that .of Children in Urban Brazil heteerthat defies stereotyprkg. While most are not poor, a large Ricardo Barros number of the poor, Louise Fox l es lly ch ldren, live in. Rosane Mendonca these households. The most effective antipoverty interention.is to focus on raising the Ifntme of the. woing women in these househQlds. Special effort is .:sn needed to encourage hidIren in these households -to stay In school. - The World Bank Policy Research Departnent Office of the Director March 1994 | POLICY RESEARCH WORKING PAPER 1275 Summary findings Barros, Fox, and Mendonca analyze the characteristics Poor children ten(. to live in households headed by and behavior of households headed by women in urban women. Brazil and identify some of the consequences for child These housekolds are poor not because there are more welfare on the growth of these households. Among their children or fewer adults but because women earn less findings: than men. Women heading households do not earn less * Households headed by women are a heterogeneous than other women - on the contrary. However, if group, which varies strongly by region - as does the female heads of households earned as much as male extent of poverty among them. Such households are heads of households, the average income in households more common in the northeast and increase with headed by women would be above that for other urbanization. households and fewer single mothers would be poor. - Households headed by women are not, on average, The best interventions to eliminate poverty in this a "vulnerable group' in Brazil, as some are quite well off. group are those that focus on: The subset of such households that are very poor is quite * Ending wage discrimination. vulnerable. Households headed by women tend to be * Ending occupational segregation. poorer in the northeast, especially around Recife, than in Interventions that focus on raising skill levels and Porto Alegre in the south, where there is virtually no educational attainment for the whole workforce, gap. including women, would also help alleviate absolute * Less than half the households headed by women poverty, although not necessarily relative income contain dependent children, and only a third are headed differeiices. 'Workfare" or public employment policies by the stereotypical 'single mother." When there are would not help this group since most already participate children in households headed by women, especially in the labor force. households headed by single mothers, the income gap is Programs targeted to this group would not be greater than in other households. partir :arly progressive, given the heterogeneity and As a proportion of households in Brazil, households income spread among these households. But the results headed by women and containing children represent do suggest the need for special interventions for children only 3.4 percent of urban households, but this group in households headed by women, given those children's tends to be poor, which is worrisome for child outcomes. tendency to stay out of school. This paper - a product of the Office of the Director, Policy Research Department - is part of a larger effort in the department to analyze behavioral characteristics of vulnerable groups. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Kay Binkley, room E-1243, extension 81143 (46 pages). March 1994. The Policy Research Working Paper Series disseminates the findings of work in progress to encorwage the exchange of ideas about development issues. An objective ofthe series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be used and cited accordingly. The findings, interpretations, and conclusions are the authors' own and shobd not be attributed to the World Bank, its Executive Board of Directors, or any of its member covntries. Produced by the Policy Research Dissemination Center FEMALE-HEADED HOUSEHOLDS, POVERTY, AND THE WELFARF OF CHILDREN IN URBAN BRAZIL Ricardo Barros Louise Fox Rosane Mendonca This study was supported by World Bank Research Funds. INTRODUCTION Households headed by women have become an important phenomenon worldwide in the last half of the twentieth century (ICRW, 1988). Brazil is no exception, having experienced a steady increase in female-headed households as a share of total households over the last three decades, especially in large urban areas. In 1987, one-fifth of all households in metropolitan areas in Brazil were headed by women. As in other countries, female-headed households in Brazil are more likely to be in poverty at any point in timne than are male-headed households.' Little is known about the long term consequences of this trend or the implications for anti-poverty programs in Brazil. If the research of Wood, (1989) is any indicator, the consequences are worrisome. Using data from the 1960 and 19980 censuses, he found that as a result of higher poverty rates and higher levels of other risk factors associated with reduced survival probabilities, children born into female-headed households are more likely to die in childhood. While debate continues over the definition of the female-headed household, the causes of the observed increase worldwide, and the normative imnplications of this increase for women and society, the fact remains that addressing the question of poverty alleviation in this generation and the next requires understanding the economic behavior of female-headed householas. In both developed and developing countries, studies have shown that female-headed households tend to have different demographic, sociological and microeconomic characteristics from male-headed households, and that these differences have major implications for the design of policy and programmatic interventions (ICRW, 1988; Kossoudji and Mueller, 1983). For example, while children in poverty are always at risk, one of the major findings of studies in the United States is that, even controlling for income, children in female-headed households seem to have a lower rate of socioeconomic attainment than children in two-parent families (Garfinkel and McLanahan, 1986; McLanahan and Bumpass, 1988). This research indicates that the same anti-poverty policy interventions may See Goldberg and Cremen, (1990) for a review of this issue in OFCD and East Euromean countries. 2 have different success rates depending on household composition, and indicates that special measures may be needed for female-headed households. The purpose of this study is to analyze the characteristics and behavior of female- headed households in urban Brazil and to identify some of the consequences of the growth of these households for child welfare. We do this by identifying the types of female-headed households that are found in Brazil, which ones are more likely than others to be poor, why they are more likely to be poor, and the consequences of poverty and female headship for children in these households. One problem with analyzing these questions in other developing countries has been the representativeness of the data (ICRW, 1988). Headship is often loosely defined at the interview stage, and therefore the meaning of this variable in survey data is sometimes ambiguous, and can represent interviewer bias rather than an actual headship situation from an economic point of view. For this reason, we begin by analyzing whether the economic role of women in households designated as "female-headed" in our data set is actually different from the woman's role in those designated "male-headed." Finding that the gender of the reported head is indeed usually the gender of the person most responsible economically for the welfare of the household (in the case of both male and female heads), we then ask what types of female-headed households are found in Brazil, and why these households are more likely to be in poverty. Do female-headed households have a lower number of earners (or a lower participation rate among the potential earners), earners with less income-earning power (human capital) or more mouths to feed? Are the answers to these questions fundamentally different when discussing male-headed households in poverty? We find, first, that generalizations about female-headed households are difficult to make in Brazil, as this group is quite heterogeneous, including rich and poor households. However, there is one common theme among Brazilian female-headed households -- those with cl-ldren (less than one-half of the total) have a much higher probability of being poor. Of those female-headed households that are poor, the main reason for this lower income is not because of a lower number of earners per capita, but because of the lower earning power of these earners. As women tend to earn less on average than do men in Brazil, a household 3 lacking male-earned income simply has a much higher probability of being poor. In the final section we show that Brazilian female-headed households appear to be replicating the trend observed in the United States. Even after controlling for income, children in these households are less likely to attend school and are more likely to work. While work experience per se is not a bad thing, as the Brazilian economy seeks to compete in the high- tech end of manufacturing and services, workers without sufficient education will suffer increasingly limited opportunities. Thus, the growth of female-headed households raises questions about the intergenerational transfer of poverty. The data used in this study are from the Brazilian annual hodsehold sample survey data, PNAD2, from various years. This survey is performed annually (except in the census years). On the whole, these data are reasonably good, but they contain a few of features that limited our analysis. First, although the questionnaire contains a number of items regarding household income, most of these are designed to measure steady income flows. The survey does a poor job of recording agricultural income. For this reason, we concentrated our analysis on urban households (covering roughly three-fourths of Brazil's population). Second, the survey contains both standard questions (asked each year) and special questions designed to explore a particular topic in-depth. For example, while sex of head is asked each year, only in 1984 were detailed questions on marital status asked. For this reason, the bulk of the data presented here are from 1984. Finally, Brazil suffered high and highly variable levels of inflation in the 1980s. The increase in inflation has made regional price differentials, already known to be a problem in the 1970s, even greater, vastly complicating the task of performing intertemporal and interregional comparisons of household income.3 Given these data weaknesses, we proceeded as follows. In Part 1, the descriptive section, we look broadly at the characteristics of female-headed households in all urban 2 Pesguisa National por Amostra de Domicilios. conducted by the Brazilian National Institute for Geography and Economics. 3 See Fox (1990) for a discussion of the methodological problems associated with trying to measure differences in living standards in Brazil. 4 areas. In Part 2, a more in-depth analysis of the determinants of poverty and the consequences for children is presented. This analysis is confined to three metropolitan areas, and poverty is defined according to relative income in each area. No interregional comparison of households by income level is attempted. While some generality is lost with this approach, we are much more confident with respect to our results for these three cases. With no aggregation of data among the areas, differences in our independent variables are not confounded by interregional price differences. I: ANATOMY OF FEMALE-HEADED HOUSEHOLDS IN URBAN AREAS While analysis of the female-headed household is a relatively recent phenomenon, as a family type, these households have a surprisingly long history in Brazil. According to historical records, almost 30 percent of households in the city of Sao Paulo in 1765 were headed by women (single women, married women whose husbands were absent, or widows). By 1802 this share had risen to 44 percent, before falling to 39 percent by 1836 (Kuznesof, 1985). Other researchers have also documented the importance of female-headed households in towns during the nineteenth century, especially in the northeast (Diaz and Stewart, 1989). Most historians attribute the high number of female-headed households in Brazil to the need for adult male labor migration in an economy that was characterized primarily by a plantation agriculture. Diaz and Stewart (1989), noting that almost 90 percent of female heads in a village in Bahia were black or mulatto (freed slave or free households), and that the adult male-to-female ratio was about even, point to the role of occupational segregation by race and gender in the sugar economy in creating these households. In the nineteenth century, as in the twentieth, female-headedness was associated with poverty. In the latter half of the twentieth century, the share of female-headed households and families in the population has been increasing steadily after hitting a low in 1960 (Table 1). In 1960 female heads constituted only 10.7 percent of all household heads and 14 percent of family heads in urban areas. By 1986 these shares had increased by 70 percent with respect 5 to all households and 50 percent with respect to urban families.' Starting with 1978, female-headed households are classified according to whether they have children. Over this period, female-headed households increased their share of total households by 18 percent. Not surprisingly, the fastest growth took place in the 10 largest cities in Brazil.5 Female- headed households without children increased faster in every region, especially in the metropolitan areas. The large cities with the fastest growth in female-headed households were Rio de Janeiro in the southeast, and Fortaleza in the northeast. What is Meant by a Female-Headed Household? Initially, household surveys sought to define the head of household in order to have a reference point in each household and to avoid double counting. Over time, headship has come to imply normative assumptions about authority and income-earning responsibility. 'fhese acsumptions do appear to have an empirical base, as estimates of earnings functions using household survey data find a positive effect of headship after controlling for human capital variables and gender.6 Recently, however, the designation of headship in national household surveys (the basis for much of the analysis of female-headed households) has been questioned. It is alleged that household surveys reflect the norms of the household membet questioned, or of the interviewer, rather than the relative economic importance within the household in designating the head, and that this bias has resulted in an underreporting of 4 The census data used by both authors provide data on households and families. Goldani (1989) analyzed urban famnilies, while Merrick and Schmink (1983) analyzed households. Not shown in Table 1, which combines both sets of data, is the ratio of families to households. However, it is known that the ratio of families to households declined from 1960 to 1980 (Fox, 1982), and it also appears that the ratio of female- headed families to female-headed households also declined. This is probably a reflection of two trends in Brazil: (a) declining number of multi-fanily households; and (b) the declining share of women employed as domestics, a common employment for female heads of families in multi-family households. I These cities are actually an agglomeration of cities known as metropolitan areas, and include the major city and surrounding suburbs. In this paper, we use the term city' to refer to these agglomerations. 6 See, for example, Birdsall and Fox (1985) for Brazil. 6 female-headed households by up to 50 percent7 (ICRW, 1988). Before continuing with the analysis, it is important to verify that our "operationalization" of the concept of female headship -- the designation in the household survey - performs reasonably well in separating households according to the sex of the person who bears the most responsibility for the economic welfare of the household. To evaluate the meaning of headship in the survey data, we compared the classification of households on two alternate definitions of headship. The first definition of household head is the person who has the highest income in the household. The second measure of headship has been proposed by Rosenhouse (1989). and considers as head the person wr.o provides the most "efforts on behalf of and commitment to the household" (p. 12). Effort is measured by the number of market-oriented hours worked, and has the advantage over the income definition of correcting for the problem of estimating the income of the self-employed. As incomes or hours worked could be very close, we also allow for joint headship where the value of the classification variable for the two highest-scoring members differs by less than 10 percent. If the survey classification is useful, in most households the same gender will be designated as head under all classifications. The top section of Table 2 shows the distribution of female- and male-headed households in three metropolitan areas if they are reclassified according to the income criterion.8 The last row in the table shows that overall, this criterion turns up slightly more "pure" female-headed households than the survey classification (20 percent, compared with 18.7 percent in the survey), and that in the case of 6.4 percent of the households, any classification would be arbitrary. The nature of the difference between the head reported in 7 For example, in Peru, Rosenhouse (1989) found that in the collection of the Living Standards Measurement Survey data, the oldest male member of the household was listed as the head despite the fact that in one-third of the households women were the main providers. In other cases, household survey interviewers are instructed to list the oldest male present as the head, resulting in teenage children being listed as the head of household. 8 Note that a household will be classified as having joint headship if both potential heads are the same sex. 7 the survey and the head we ide-itified is interesting. If the income criterion is applied, 17 percent of those classified in the survey as female-headed would move to the male-headed classification, but only 7 percent of those classified as male-headed would be classified as female-headed. This shows that, in Brazil, the presence of an income-earning male does not a priori exclude a household from Leing designated female-headed in survey data. The middle section of Table 2 shows the distribution of households in metropolitan areas according to the "hours worked" criterion. Overall, this criterion turns up roughly the same number of female-headed households as the survey classification. However, the number of households classified as jointly-headed rises to almost one-fourth of the sample. About the same number of reported female-headed households move to male-headed under the hours-worked criterion as unider the income criterion, but more reported male-headed households move to the female-headed column under the hours-worked than under the income criterion. Putting the two criteria together in the bottom section of Table 2, we see that of the 20 percent of households classified as female-headed according to the income criterion, less than two-thirds would classified as female headed if both criteria were used. Twenty-five percent would be classified as jointly-headed, the rest as male-headed. Of the 74 percent classified as male-headed under the income criterion, almost three-fourths would be classified as male-headed under the hours-worked criterion. If we consider that jointly-headed households have a head of the same gender as the reported head, then the concordance among the criteria is quite strong. Thus, the reported and the hours-worked criteria misclassify, relative to each other, 9 percent of all households and the income and the working criteria misclassify, relative to each other, only 7 percent of all households. Who are the alternate heads in these households? We investigated this question for those households where the income criterion turned up a different gender of head than the one reported in the survey (Table 3). In the case of reported male-headed households, the alternate head is most likely to be a spouse, especially if the reported head is himself a labor- 8 force participant.9 However, in the case of reported female-headed households, the alternate head is most like;y to be a son, suggesting that age is an important factor used by households to define their head. Given that women tend to outlive men, we suspect that a number of the reported female-headed households are actually extended-family households, with an important income contribution from an adult male relative, usually a son. To verify this assumption, we looked at the number of households in which at least one member is older than the reported head and is not the head's spouse. In 94 percent of reported male-headed households and 92 percent of repGrted female-headed households, the head is the oldest member. Overall, in over 90 percent of households, the household survey data designate as head of the household an adult who is the same gender as at least one of the adults most responsible for the family's economic welfare. For that reason, we continued to use the reported head in our analysis. If a net gender bias exists in the headship designation, it favors female headship, as female-headed households were more likely to be reclassified than male-headed households. This bias toward female headship designation appears to be a result of an age bias, as women live longer and therefore are more likely to be named head. Nonetheless, this analysis suggests that some caution is warranted in using the headship concept too narrowly in a situation like that found in Brazil, where most households depend on more than one income-earning member. Characteristics of Female-Headed Households Female-headed households in urban Brazil are a heterogeneous group (Tables 4 and 5). Given the analysis above, it is not surprising that female heads are on average about seven years older than male heads, and a substantial portion are widows.'0 Household type differs by region. Female-headed households are most likely to be found in the northeast, 9 This was confirmed in a special tabulation, not shown. 10 Note that marital status information is available only for women under the age of 56. 9 the poorest region, and least .ikely to be found in the south. Female-headed households in the more developed southeastern and southern regions are also less likely to have children. Nonetheless, given the tremendous regional differences in Brazil in income, economic structure, and culture, it is striking that the differences in prevalence between regions are not stronger. Only one-third of urban female-headed households correspond to the Moynihan stereotype of a single mother and her children.1I Over one half of female-headed households have no minors living in the household. Of those without minors, 40 percent are single, while 44 percent have an adult child living in the household. Where there are minor children in the h, ,sehold, another adult is also iikely to be present. On the whole, female- headed households tend to have slightly fewer adults than the average for Brazil, and the adults are more likely to be women. Among all female-headed households, the largest group comprises single women. Divorce' is the most usual route to female-headedness for those with minors in their household (39 percent), but 30 percent of female heads with under-age children never married, and 28 percent are widowed (perhaps providing for their grandchildren). The heads of the Moynihan-type households are among the youngest in the group, and the most likely to have been divorced. The economic characteristics of female-headed households are different from the average in Brazil on a number of dimensions (Table 5). Although the per capita income of urban female-headed households is 89 percent of that of all urban households, the average income of female heads is 50 percent of the average in urban Brazil. This discrepancy " Moynihan (1965) was one of the first to bring to national attention in the Un.ted States the problems of poor single mothers with children among African-American households, sparking an intense debate regarding public policy ard female headship. Among other things, Moynihan was accused of racism and of stereotyping women in his pioneering work (see Wilson and Neckerman, 1986, for a discussion). Nonetheless, his work is credited with raising the issue of household structure and poverty, and is still widely discussed in the United States today. 12 In the Brazilian survey, the classification "divorced' includes couples who have separated but have not legally terminated the marriage. 10 suggests that these households depend heavily on other household members' income for support. If the average dependency ratio was the same as in male-headed households, the per capita income of these households would be poor indeed. Female heads appear to have less income-earning capacity than their male counterparts. Female heads have, on average, less education than their male counterparts, which is not surprising given that (1) female heads are older, and older cohorts tend to have less education; and (2) within older cohorts, women tend to have less education than men (Lam, et al., 1992). Note that the female head's older age on average does not necessarily mean more work experience, given the need to take time out from paid employment in order to have children. Female heads are much less likely to be economically active. This partially explains why these households tend to get a lower share of income from the household head. In sum, female-headed households are well identified in household survey data in Brazil. They are, however a heterogeneous group, containing rich and poor, widows and divorc6es, and including a number of male earners. Despite their heterogeneity, in the aggregate this household type does possess some characteristics significantly different from male-headed households. One of those is the tendency for these households to be overrepresented among the poor. Previous studies, including Fox (1982), Merrick and Schmink (1983), and Pastore, et al. (1983), analyzing data from the 1960s, 1970s, and 1980s, and using various definitions of poverty, all found that female-headed households had a 30-50 percent greater chance of being in poverty than did male-headed households. We now turn to an analysis of the causes and consequences of this association using our survey data. We focus especially on female-headed households with children, as this aspect in particular has been underanalyzed in previous studies. 11 Urban Poverty in Brazil Using a poverty line of about $US300 per capita per annum in 1985 (roughly 13 percent of Brazil's per capita GDP), about one-fourth of Brazil's population is in poverty."3 Slightly under half of those living in poverty reside in urban areas. Poverty in Brazil is correlated with demographic factors as well as the income-earning characteristics of the household head. Regionally, 47 percent of urban poverty (and 55 percent of total poverty) is in the northeastern region, where the per capita income is about 60 percent of the country as a whole. Poor urban households tend to have fewer earners, more children, and a higher dependency ratio than the average for all urban households. Heads of poor urban households are as likely to be economically active as the average for urban Brazil, but their education level is, on average, 40 percent lower, their average hours worked and unemployment rates are higher, and thus the average income of the head and the household is lower. Twenty-two percent of poor urban households are headed by women, indicating that these households have a roughly 20 percent greater chance of being in poverty than the average Brazilian household. Earlier, we saw that female-headed households exhibit different economic characteristics from other households. Are female-headed households who are poor also different from other poor households? To answer this question, we compare the characteristics of poor urban households with those of poor urban female-headed households, and poor female-headed households with children (Table 6). We disaggr'egated this analysis by region in order not to confound the analysis with the strong regional effect on poverty of the northeast, the area with the highest incidence of female-headedness as well. In all areas, female-headed households have a somewhat higher incidence of poverty. What is striking about Table 6, however, is the poverty rate for female-headed households 13 The analysis of the correlates of poverty is taken from Fox (1990). The poverty line used and the analytical issues associated with this line are described on pages 2-6 of that paper. 12 with children and no other adult. In the northeast, nearly two-thirds of these households are In poverty, while in each of the southern regions, over one-third are in poverty. Even among the poor, female-headed households are still smaller, with a smaller number of children. Clues to the high poverty rate among female-headed households with children and no adults can be found by comparing their household characteristics. While these households have roughly the same number of children as do other poor households, they have a lower number of earners, and consequently a higher dependency ratio. With respect to income- earning characteristics, poor female heads with children actually have slightly higher education levels than all female heads on average. Yet their average income is lower than other poor households (although not lower than all female-headed households on average). One factor influencing the lower earnings is undoubtedly the tendency for female heads to work fewer hours. It is possible that for a female head of household in Brazil, the responsibilities of household maintenance, parenting, and so forth, are not always compatible with full-time employment. Easing this trade-off will need to be an important part of an anti-poverty strategy for these households. We have established that even after controlling for regional differences, female- headed households are more likely to be poor than other households, despite some favorable factors, including a lower dependency ratio. Female-headed households with children are especially susceptible to poverty, a troubling result. In the next stage of the analysis, we look more in-depth at why that is so, and how it affects the children in these households. II: THE DETERMINANTS OF POVERTY AMONG FEMALE-HEADED HOUSEHOLDS IN METROPOLITAN BRAZIL AND THE IMPLICATIONS FOR CHILD WELFARE For the next stage of the analysis, two innovations are adopted: (1) we confine our analysis to three metropolitan areas; and (2) we modify slightly our definition of household income. The metropolitan areas chosen for in-depth analysis are Recife, Sao Paulo, and 13 Porto Alegre. Each is distinct in economy and culture. Sao Paulo is the largest metropolitan area in the country, the one with the highest average per capita income, and also the most industrialized area. Recife and Porto Alegre are similar in size, both being considerably smaller than Sao Paulo. Recife is the largest metropolitan area in northeast, and has a high incidence of both poverty and female-headed households. Porto Alegre is the largest metropolitan area in southern Brazil; it has a relatively homogeneous population and an average level of income close to the level for Sao Paulo. Overall income levels vary a great deal across the three regions our cities represent, as does the distribution of households according to their per capita adult income, and, consequently, the income brackets used to group households. Table 7 shows this difference for our three cities. Brazilian income distribution is one of the most unequal in the world; in 1987 the lowest 50 percent of households (grouped by per capita income) received only 12 percent of total household income, while the top 5 percent received 24 percent (Fox, 1990). As we are primarily interested in the poor, we cut off our analysis of income groups at the bottom 50 percent. Virtually no difference exists between Sao Paulo and Porto Alegre with respect to mean income in these percentiles. The mean income for households in the same percentiles for Recife, however, tends to be co.nsiderably lower; close to one-half of those for Sao Paulo and Porto Alegre. Although it is likely to be true that households in the same income class are poorer in Recife than in Sao Paulo and Porto Alegre, there are no guarantees that this is in fact true owing to the problem of regional price differences. Female-headed households in the three metropolitan areas mirror the similarities and differences in all urban areas noted above (Table 8). Recife, in the northeast, has more female-headed households per capita, and these households are much more likely to contain minors. Among households with children, the average number of children is slightly higher in Recife, as is the dependency ratio. Female heads of households in Recife are more likely to be widowed and less likely to be economically active, although they are only marginally older on average. While average incomes by fractiles of the income distribution are about 14 the same in both Porto Alegre and Sao Paulo, heads of female-headed households tend to have a slightly higher income in Porto Alegre. Measuring Income in Female-Headed Households In this part of the analysis, we rank households according to a slight modification of the traditional concept of household per capita income. Specifically, for the rest of the paper, household income is defined as the income from all sources of all adult members (over age 18) of the household divided by the total number of members in the household. We do this because we seek to assess how outcomes for children (such as their labor-force participation and school attendance) depend on the rank of their household in the income distribution. Hence, it is useful to be able to measure the household resources without the contribution of non-adult members, in order to prevent households from upgrading their ranking by using their children in the labor market. 14 To what extent does the exclusion of the income of non-adult members really have an impact on the ranking of households? In other words, to what extent do poor households in fact upgrade their rank by allocating children and teens (under age 18) to the labor force? To answer this question, we performed a simple experiment using data from the Sao Paulo metropolitan area, classifying households by their per capital adult income and by their per capita total income (Table 9). Ninety-four percent of all households are classified in the same relative income group whether we rank them by per capita adult income or per capita total income, and only 3 percent move up to a higher group. The proportion of households that remain in the same class, however, varies considerably from income class to income class. Poorer households rely much more heavily on the income of minors. Among those 14 Ideally, we would like to rank households based on the resources that they would have in the case that all children were attending school and out of the labor force. The income measure we use -- the adult total income per capita -- would equal this ideal measure whenever the labor supply of the household adult members was not influenced by the way in which the non-adult members allocate their time. We consider our measure preferable to one that would confound the analysis of children's labor-force participation because children's income is included. For a further discussion of this issue, see Barros and Mendonca (1990). 15 classified as being in the poorest group based on the ranking by per capita adult income, only 88.3 percent are also in this group according to the rank by per capital total income, whereas among those classified as being in the top group according to the per capita adult income, almost 99 percent are also classified as non-poor according to the per capita total income. Counting only the income of adults, we confirm once again in Table 10 that female- headed households are overrepresented at the lower end of the income distribution. This table shows the share of three types of female-headed households in each income class. The top half of the table shows the absolute incidence of female-headed households in each income group, while in the bottom half each cell is weighted by the share of their type of household in the population in each city. When the number ia. the bottom half of Table 10 is greater than 1.0, the household type is overrepresented in the income group. We also note a second trend -- the lower the relative income, the higher the likelihood that the household will be female-headed. Furthermore, while female-headed households are clearly overrepresented in the lower one-fourth of the income distribution in each city, the degree of overrepresentation is much greater among female-headed households with minors and greatest among the Moynihan-type households. 15 Why Are Female-Headed Households Poorer? Female-headed households are relatively poor in Brazil. Table 11 shows, for each city, the gap between the average per capita income of all households and the average per capita income of female-headed households (in minimum salaries). In the last column, the percentage difference in the mean incomes between female-headed households and others is IS This type of analysis is imprecise because of the problem of boundness. For example, if we invert the ratios in the bottom half of Table 10 (that is, take the inverse of each cell), the size of the difference between one cell and the other will change. To check for this problem, we perforned a logit analysis on the first row in Table 10, the lowest income group. This analysis confirmed the results in Table 10. Female-headed households with children are indeed overrepresented compared with all female-headed households, and the Moynihan-type household is three times as overrepresented as are all female-headed households. For more details, see Barros, et al., (1993). 16 calculated. For female-headed households with children, the average income gap is about the same in each city, while for all female-headed households, the gap is much greater in Recife, and there is virtually no gap in Porto Alegre. Once again, the heterogeneity of female- headed households is observed. These results raise two questions. First, is there a common explanation across cities for the noticeable income gap between female-headed households with children and other households? Second, why are female-headed households relatively poorer in Recife when compared with other households in the same area? Household income per capita can be disaggregated into the following factors: * the number of earners (the proportion of adults with positive income); -- the average income of the earneis (the earning capacity); and * the ratio of the number of children to the number of adults (the dependency ratio).16 A class of households may be overrepresented among the poor if (1) few adults have positive income -- available earnings capacity is used less intensively, either because of a lower income per hour or a lower number of hours worked; (2) the income of the earners is low -- a smaller earnings capacity; or (3) the dependency ratio is high. In order to measure the importance of each factor in contributing to the relative income gap of female-headed households in each city, we performed a counterfactual simulation. For each of the household composition variables listed above (e.g., earnings 16 Algebraically, this can be stated as follows: y y, (p) / (I + d) where: y = household income per capita y,= average income amonig adults with positive income p = proportion of adults who have income d = ratio of adults to non-adults 17 capacity, intensity of earnings capacity use, and the dependency ratio), we asked the following question: What would the income gap between female-headed households and other households be if the average in female-headed households on one variable was the same as the average in the other households?"7 In other words, what is the contribution of a lower value, on average, of this variable to the income gap between household types? The relative difference in means among household types and the percentage point reduction in the income gap between female-headed households and other households resulting from this simulation is shown in Table 12. Across metropolitan areas, the highest reduction in the average poverty gap comes from the earnings capacity sinulation. This is not surprising, as the greatest difference in means among groups appears on this variable. If earners in female-headed households had the average income of earners in other households, their average per capita income would be above the average for other households in all three areas. If earners in female-headed households with children had the same earnings capacity as other households, the average income gap would fall by two-thirds. As adults in female- headed households tend to have high participation rates, changing the utilization of earnings capacity to that of the average household actually increases the income gap. Indeed, an income gap appears between female-headed households and other households in Porto Alegre, where it did not exist before. Changing the dependency ratio to the average among households reduces the income gap for female-headed households with children, as these tend to have a higher dependency ratio, but not for all female-headed households, who as a group have a slightly lower dependency ratio. The difference in average earnings is, therefore, the common reason for lower income across female-headed households. Given the importance of this variable in accounting for 17 The simulation procedure is described extensively in Barros, et al., (1993). The basic idea was to change the mean of the variable but maintain the distribution in the class of households (female-headed household, or F, and others, or 0). Thus, for each simuiation, the value of the variable x in each household was increased by a scalar (s), such that: s - x /x, ; where x1 and x, are the means of variable x for households in classes F, 0. 18 the income gap, it is useful to further disaggregate the earnings capacity variable, comparing the average earnings of (1) males and females, and (2) heads and nonheads. This is done because, all other variables being equal, in Brazil, heads and men tend to earn more than nonheads and women (Birdsall and Fox, 1985). Following the previous approach, earnings capacity may be lower in a given class of households because: * males in these households have a smaller earnings capacity; * females in these households have a smaller earnings capacity; * more of the earners are female; * heads of these households have a smaller earnings capacity; 3 nonheads may have a smaller earnings capacity; or more of the earners are nonheads. The same simulation procedure as above was repeated for these six variables. The relative difference in means and the reduction in the poverty gap resulting from the simulation are shown in Table 13. The simulation results show that lower incomes among earners in female-headed households are not due to smaller earnings capacity among female earners, or primarily to a lack of male-earned incomes, but to a lack of earners with the earnings capacity of males in other households. The earnings of males in female-headed households (who tend to be secondary earners) is, on average, one-third to one-half of that of female earners in either type of household. Equating the male earnings capacity among female-headed households with that of other households would reduce 20 to 30 percentage points of the observed income gap, with a higher effect in female-headed households with children (where the difference in means is greater). This is because males in male-headed households are not secondary earners, so they have higher earnings on average. Although there is a large difference between female-headed households and other households in the percent of female earners, reducing this percent had virtually no effect on the income gap, as the males in female-headed households have an average income roughly equal to the females in these 19 households (which, in turn, is similar to those in other households). Therefore, equating this variable across households makes no difference to the income gap. The effect of the male-female earnings differential in Brazil shows up again in comparing the mean earnings of heads. While the average earnings of secondary earners is about the same in both types of households, the average earnings of female heads is about one-half that of male heads. If female heads had the earnings capacity of male heads, the income gap would disappear in all households. In sum, female-headed households tend to have a lower income than other households primarily because of the average earnings of the female head. As women who participate in the labor force tend to have, on average, the same education as men in a given age cohort, supply- or demand-side discrimination is clearly an important factor." Supply-side discriminatory factors include (1) fewer hours worked, and (2) the tendency of women to work in occupations that pay lower salaries (an effect of occupational segregation), or within occupations in lower positions.'9 Demand-side discrimination includes both the failure of employers to hire women for higher-paying jobs for which they are qualified, or the failure to pay them salaries that are equal to those of their male counterparts. The relative importance of these factors is not the subject of this work. However, the lesson from this analysis is that raising female heads' incomes, especially those of female heads with children, is an important element of a poverty-reduction strategy. 18 We use the word 'discrimination" to encompass a range of actions based on prejudices or stereotypes which generate outcomes systematically different for men than for women. The discriminator may be a man or a woman, including the female head herself (e.g., choice of occupation). 19 Birdsall and Fox (1985) found evidence both of occupational segregation and wage discrimination affecting women's salaries in Brazil. 20 Children In Female-Headed Households Concern about the growth of female-headed households is partly motivated by the potential consequences of this household structure for the welfare of children. Poor households in Brazil, as in other countries, are more likely to contain more children, and as a result, minors are generally overrepresented in the bottom third of the income distribution. However, even though female-headed households with children are overrepresented among poor households in our case studies, this does not imply that children living in female-headed households are over-represented among poor children. Given that female-headed households have a lower number of children on average, it is possible that they are not. Table 14 presents a distribution of children across income class for each metropolitan area, among all minors and those living in a female-headed household. Table 15 shows the relative incidence of minor children living in female-headed households, indicating the degree of overrepresentation. Not only are children much more likely to live in poor households, but poor children are quite likely to live in female-headed households. Indeed, minor children in female-headed households are more overrepresented among the poor than are female-headed households as a whole. Children in female-headed households tend to be older (Table 16). In all three cities, older children are overrepresented in all income groups. However, younger children in female-headed households are disproportionately concentrated in poor households, as the ratio between the average share in all income groups and the average share in the lowest income group is highest for the youngest childre-r.20 Are children in female-headed households worse off than other children in their income group? It is well known that poverty is hard on children (National Commission on Children, 1991; Wood, 1989). However, what we attempt to analyze is whether the female- headed household structure exerts a separable, independent effect on children, above and 20 This result was also verified by a logit analysis. See Barros, et al., (1993) for details. 21 beyond the poverty effect. This in done in steps. First, we evaluate the performance of children in female-headed households compared with all children with respect to childhood outcomes. Next, the portion of any performance difference due to the tendency to be in poverty is identified (e.g, the sensitivity of the outcome variables to household income), with the remainder of the difference attributed to an independent effect of female headship. Evidence of an independent effect would indicate that the consequences of poverty are more severe for children in female-headed households than for all children. The PNAD data set allows the analysis of two potential outcomes for children: school attendance and labor-force participation. We interpret a lower rate of school attendance for children as an indicator that children are worse off. The interpretation of the labor-force participation of children and their well-being is more complex. It is not necessarily true that holding children out of the labor market will benefit society. For example, consider the case in which some families in the society are very poor and children in these families can find jobs that (1) will permit them to continue attending school; and (2) will pay them wages that are high enough to alleviate their poverty. In this case, it seems that society will be better off permitting children to work. However, if children work at the expense of school attendance, when basic reading, writing, math, and problem-solving skills are being taught, labor-force participation is clearly disadvantageous in the long run. An increase in the labor-force participation of children that leads to reducdons in school attendance probably leads to a reduction in society's welfare. By contrasdng the school attendance of children who live in female-headed households and who work with the attendance rate of all children in the labor force, we can investigate the extent to which children in female-headed households have mon. difficulty than other children in resolving the conflict between studying and working. We analyze school attendance separately for children ages 7-9 and ages 10-14. Due to the nature of the PNAD questionnaire,2' we have to limit our study of labor-force participation to children aged 10 to 14 years. 21 Labor market-related questions are applied only to persons 10 oi. more years old. 22 With respect to all outcomes, children in female-headed households appear to be worse off in Recife, but the contrast is not quite so stark in the other two cities (Table 16). Among 7-9 year-olds in Sao Paulo and Porto Alegre, children living in female-headed households have school attendance rates very similar to that of all children in the same metropolitan area and age group. For both age groups in Recife and for 10-14 year-olds in Sao Paulo and Porto Alegre, children living in female-headed households have lower school attendance rates than all children in the same age group and metropolitan area. In Sao Paulo and Porto Alegre, only in female-headed households do school attendance rates decrease dramatically with age. In each metropolitan area, the labor-force participation rate of children 10 to 14 years old living in female-headed households is higher than that of all children in the same area. Surprisingly, labor-force participation rates are higher in the wealthier cities of Sao Paulo and Porto Alegre than in Recife. This may be the result of better labor-force conditions in the southern cities. The data in Table 16 do point to a conflict between work and school for children in female-headed households. Among children in the labor force, the proportion not currently attending school is higher among those living in female-headed households. The magnitude differs by metropolitan area. The greatest difference is found in Recife. In Sao Paulo and Porto Alegre, the difference is not significant. Sao Paulo is notable, as children in this city appear to be the best at combining school and work. Isolating the Poverty Effect A large body of literature supports the notion that the level of household resources is an important determinant of children's outcomes in Brazil.1 As children in female-htaded households are much poorer than children living in non female-headed households, it could be that differences between children in female-headed households and all children, as those reported in the previous section, are simply a consequence of the greater poverty among 2 See Calsing and Schmidt (1986), Comia (1984), and Barros and Mendonca (1990). 23 children in female-headed households. In other words, are variations in children's outcomes across households with different structures really due to differences in the structures of the household per se or are they entirely due to other differences across households, like household income, which are associated with differences in household structure? Children's outcomes in the three metropolitan areas are indeed sensitive to income (Table 17). School attendance increases and labor force participation decreases for all metropolitan areas in a monotonic fashion. The percent of working children not in school also declines with income. Thus, it is quite possible that most of the female-headed household effect observed above is actually a poverty effect. To isolate the poverty effect, we onc- again perform a counterfactual simulation. First, we estimate the outcome for children in female-headed households by income class. Then we give childrer. in female- headed households the same distribution of income as all children have (i.e., we raise the average income of female-headed households). We then estimate the new outcome. The difference between the new outcome (flh*) and the old outcome (flh) is the impact of poverty. The remainder is the pure female-headed household effect.23 The results of this simulation are shown in Table 18. The results vary substantially by metropolitan area. In Sao Paulo, which has the best outcomes for all households and the smallest difference between female-headed households and other households, most of the observed outcome difference is due to a female-headed household effect. Raising the income of female-hseaded households with children would eliminate one-third or less of the difference observed in outcomes. In Recife, lower incomes explain much more of the difference in child outcomes -- 70 percent of the difference in school attendance rates for children age 10- 14, and 80 percent of the labor-force participation rates. In both Recife and Sao Paulo, income plays a much more important role in explaining outcome differences in older children than in younger children, but in Porto Alegre, the reverse is true. In Porto Alegre, giving 23 With any decomposition, an index-number problem arises. Rather than present two sets of numbers, we present those showing the lowest female-headed household effect. For an algebraic exposition of the decomposition and simulation procedure, see Barros, et al., (1993). 24 female-headed households more income would actually result in children spending slightly less time in school. This result reflects the insensitivity of this outcome to household resources in this metropolitan area. CONCLUSION In this study we have provided an "x-ray" of female-headed households in Brazil. One of the first conclusions of the analysis is that female-headed households are a heterogeneous group, showing strong variation by region. Female headship is more common in the northeast, and increases with urbanization. Women become female heads through various routes. One route is the termination of marriage, either through divorce (most common) or widowhood, but the most common route for women in Brazil is to never marry. Less than half of female-headed households contain dependent children, and only about one- third are characterized by the stereotypical "single mother." Female-headed households are not, on average, a "vulnerable group" in Brazil, as some are quite well-off. Others are very poor, and this subset of female-headed households is quite vulnerable. The extent of the poverty among female headed-households varies substantially by region. In the northeast, the female-headed households are, on average, poorer, especially in the metropolitan area of Recife, compared with Porto Alegre in the south, where there is virtually no gap. Female-headed households with children do show a wide income gap when compared with other households, especially for the single-mother- household type. Although this type is small in terms of percent of households in Brazil (in urban areas, only 3.4 percent of households), the strong tendency of this group to be poor and the consequences with respect to child outcomes in these households of the interaction between poverty and household structure, is worrisome. This effect is even more worrisome when the strong tendency for poor children to be in female-headed households is considered. Female-headed households have lower incomes not because they have more children, or fewer adults, but because the head of the household, being female, earns less. This 25 female head does not earn less than other females -- on the contrary. If female heads earned as much as male heads, female-headed households would have an average income above that for other households, and fewer single-mother households would be in poverty. This analysis shows that the best interventions for the elimination of poverty in this group are those that focus on (1) ending wage discrimination; and (2) ending occupational segregation. Interventions that focus on raising skill levels and educational attaimnents of the whole work force, including women, should also be useful in alleviating absolute poverty, although not necessarily the relative income differences identified here. "Workfare" or public employment policies would not be of particular benefit to this group, since most are already participating in the labor force. Given the heterogeneity of female-headed households, and the income spread, programs targeted to this group would not be particular!y progressive (although they may be welfare-enhancing for other reasons). However, our results suggest the need for special interventions for children in female-headed households, given their tendency to stay out of school. School attendance is highly correlated with income, so it is not surprising that children in female-headed households have poorer attendance records when compared with all children. What is surprising is that even controlling for household income, children in female-headed households have poorer school attendance records than other children. This is especially true in the case of older children, who are more likely to in the labor force and out of school. This finding indicates that in female-headed households, the challenge of balancing the desire (perhaps need) to earn additional income with the desire for additional educational attainment is particularly problematic. One possible explanation for this result may be that female heads are "time poor" (Vickery, 1977). In other words, female heads simply do not have enough hours in the day to work, maintain a household, and make sure their children get an education. Another possible explanation is that, given the lower returns to education for women in Brazil, in raising their children, female heads value education less than labor force experience. Clearly, this issue requires further study. 26 An interesting result is the stronger association between poverty and female headship in the poorer area of Recife than in the richer metropolitan areas in the south. It is often assumed that poverty among female-headed households is one of the most difficult forms of poverty to eradicate. However, those who hold this belief would predict that female-headed households would be more overrepresented among poor households in the richer cities, not -he converse, as we found. Obviously, differences in other variables known to be important in explaining household composition could fully account for this result, including cultural differences, patterns of labor migration, and the like. This issue also suggests the need for further analysis, beyond the scope of this paper. 27 REF1RENCES Barros, Ricardo and Rosane Mendonca. 1991. "Taxa de participacao no mercado de trabalho e frequencia a escola de criancas e adolescentes no Brasil." Mimeo. Barros, Ricardo, Louise Fox, and Rosane Mendonca. 1993. "Poverty among female-headed households in Brazil." Mimeo. Birdsall, Nancy and Louise FQx. 1985. "Why males earn more: Location and training of Brazilian school-teachers." Economic Development and Cultural Change, 33(3):533- 555. Calsing, Elizeu F. and B.V. Schmidt. 1986. "Situacao socio-economico e a demografica dos menores no Brazil e a nordeste." in 0 menor e a Pobreza, Brasilia. Brazil: IPLAN/IPEA/UNICEF. Cornia, Giovanni Andrea. 1984. "A survey of cross-sectional and time-series literature on factors affecting child welfare." World Development, 12, (3): 187-202. Diaz, Arlene and Stewart, Jeff. 1989. "Occupational status and female-headed households in Santiago Maior do Iguape, Brazil 1835." Mimeo. Fox, Louise. 1982. "Income distribution in Brazil: Better numbers and new findings." Ph.D. diss., Vanderbit University. Fox, Louise. 1990. Poverty Alleviation in Brazil. 1970-87. Internal Discussion Paper, Latin America and the Caribbean Region Report No. 072, The World Bank. Garfinkel, Irwin and Sara S. McLanahan. 1986. Single Mothers and Their Children: A New American Dilemma. Washington, D.C.: The Urban Institute Press. Goldani, Ana Maria. 1989. "Women's transitions: The intersection of female life course, family and demographic transition in twentieth century Brazil," Ph.D. diss. Goldberg, Gertrude Schaffner and Eleanor Kremen. 1990. The Feminization of Poverty: only in America? New York: Praeger Press. International Center For Research on Women (ICRW). 1988. "Women-headed households: Is. -es for discussion." Joint ICRW/Population Council Working Paper. Kossoudji, Sherrie and Eva Mueller. 1983. "The economic and demographic status of female-headed households in rural Botswana." Economic Development and Cultural Change, 31:831-59, July. 28 Kuznesof, Elizabeth Anne. 1985. "The role of the female-headed household in Brazilian modernization: Sao Paulo 1765 to 1836," Journal of Social History. Lam, David, Guilherme Sedlacek, and Suzanne Duryea. 1992. "Increases in women's education and fertility decline in Brazil." University of Michigan Research Report No. 92-255, September, 1992. McLanahan, Sara and Larry Bumpass. 1988. "Intergenerational consequences of family disruption," American Journal of Sociology, Vol. 49, No. 1 (July 1988): 130-52. Merrick, Thomas W. and Marianne Schmink. 1983. "Households headed by women and urban poverty in Brazil." In Women and Povert, in the Third World, ed. Mayra Buvinic, Margaret A. Lycette and William P. McGreevey. Baltimore: Johns Hopkins University Press. Moynihan, Daniel P. 1965. The Negro Family: The Case for National Action. Washington D.C.: U.S. Department of Labor, Office of Family Planning and Research. National Commission on Children (U.S.). 1991. Beyond Rhetoric: a New American Agenda for Children and Families. Washington D.C.: U.S. Government Printing Office. Pastore, Jos6, Helio Zylberstajn and Carmen Silvia Pagotto. 1983. Mudanca Social e Pobreza no Brasil: 1970-1980 (O gue Ocorreu com a Familia Brasileira?). Sao Paulo, Brazil: Estudos Economicos, FIPE/PIONEIRA. Rosenhouse, Sandra. 1989. "Identifying the poor: Is "headship" a useful concept?," Living Standards Measurement Survey. Working Paper No. 58. World Bank. Washington. D.C. Vickery, Claire. 1977. "The time-poor: A new look at poverty." Journal of Human Resources 12(1):27-48. Wilson, William Julius and Kathryn Neckerman. 1986. "Poverty and family structure: The widening gap between evidence and public policy issues." In Fighting Poverty, ed. Sheldon Danziger and Daniel H. Weinberg. Cambridge: Harvard University Press. Wood, Charles H. 1989. "Wowen headed households and child mortality in Brazil, 1960- 1980." Deparanent of Sociology, University of Florida, Gainesville. 29 Table 1: Brazil - Indices of Growth of Female-Headed Families and Households, 1960-1987 Share of Female-Headed Families and Households (1960-1986) (percent) Year Urban Areas All Brazil (share of families) (share of households) 1950 - 12.1 1960 14.0 10.7 1970 15.7 13.0 1980 18.0 15.6 1986 20.6 18.4 Sources: Mrricik and Schmink (1983). Goldai (1989), and author's tabulations Growth of Female-Headed Households in Urban Areas by Location (1978.1986) (percent change in share) Location FlIH FHH wlth children All Brazil 18 14 Center/Northwest 13 0 Southeast 26 16 Northeast 5 0 South 21 16 Metropolitan Areas 33 14 Source: auhor's tabtlations Families Headed by Females by Metropolitan Area (1978-1987) (percent) Percentage change Metropolitan Area 1978 1981 1982 1983 1984 1985 .1986 1987 in share, 1978-1987 Northwest Forteleza 18.8 21 21.2 22.2 22.2 23.8 21.6 24.6 30.8 Recife 22.7 24.3 23.6 23.6 25.5 25.6 23.2 27.2 19.8 Salvador 20.8 21.7 22.2 22.5 23.8 25.3 25.4 26.2 26 Southeast Belo Horizonte 20.3 20 19.9 21.7 21.9 23.3 23.9 23.9 17.7 Rio de Janeiro 18.3 21.2 21 21.7 23.1 23 24.1 24.3 32.8 Sao Paulo 15.3 16.4 16.2 17.2 18.7 18.8 20.7 19.1 24.8 Scath Curitiba 14.5 17.3 15.1 15.5 18.9 18.4 18.1 18.1 24.8 Porto Alegre 16.9 19.1 4.75 19.4 20.3 20.5 21.3 21.2 25.4 Center/Northwest Belem 23.5 24.7 23.7 25.7 26.1 25.6 27.5 27.1 15.3 Source: author's tabulauoas 30 Table 2: Brazil - Distribution of Households According to Headship Criteria, Metropolitan Areas, 1984 (percent) Reported Actual Headship, Income Criterion Headship Male-headed Female-headed Joint-headed Total Male-headed 86.6 7.3 6.1 100 Female-headed 17.2 75.2 7.6 100 All 73.6 20.0 6.4 100 Reported Actual Headship, Hours Worked Criterion Headship Male-headed Female-headed Joint-headed Total Male-headed 67.6 9.1 23.3 100 Female-headed 17.4 58.1 24.5 100 All 58.1 18.3 23.6 100 Income Hours Worked Criterion Criterion Male-headed Female-headed Joint-headed Total Male-headed 73.2 6.2 20.6 100 Female-headed 12.2 63.3 24.5 100 Joint-headed 28.4 16.4 55.2 100 Source: author's tabulations 31 Table 3: Brazil - Distribution of Non-Reported Income Heads By Kinship to Reported Head, Metropolitan Areas, 1984 (percent) Kinship of Income Head to Reported Head Sex of Other Reported Head Spouse Offspring Relative Total Male 69 25 6 100 Female 3 74 22 100 Source: author's tabulations Note: Joint-headship cases are excluded. 32 Table 4: Types of Female-Headed Households (FHH) in Urban Areas (percent) Share of Total, All Urban Areas, 1984 Female-Headed Households 18.2 - without husbands 17.9 - with minors (under 18) 8.2 - with minor children 7.5 - with adult children 4.3 - with minor children, no other adult 3.4 Regional Distribution (1986) Metropolitan Northeast North/Center Southeast South Areas FHH as share of total population 21 18 19 17 20 Percent of FHH without adult male 61 61 63 64 65 Percent of FHH with children 52 55 37 41 40 Percent of FHH with children 33 33 26 24 25 without adult male Distribution by Marital Status (1984) With Minors Marital Status Al FHE With Minors No Adults Single 37.7 30.0 35.9 Married 2.6 2.7 1.8 Widow 25.9 28.2 19.1 Divorced or Separated 33.8 39.1 43.2 Source: author's tabulations 33 Table 5: Brazil - Characteristics of Urban Female-Headed Households Compared With All Urban Households, 1984 Characteristic FHH All Share of households (percent) 18.2 100.0 Average income per capita (minimum salaries) 1.6 1.8 Household Composition Household size 3.2 4.2 Number of minors 0.9 1.5 Number of adults 2.1 2.5 Share of household with another adult (percent) 56.0 38.0 Number of people with positive income 1.9 1.9 Share of households with children (percent) 45.0 64.0 Share of adults who are women (percent) 84.0 54.0 Dependency ratio (children per adult) 0.6 0.7 Characteristics of Head Age (years) 51.0 44.0 Education (years) 4.4 5.4 Mean income (minimum salaries) 2.2 4.4 Percent economically active 48.0 78.0 Share of household income earned by head (percent) 65.0 74.0 Source: author's tabulations 34 Table 6: Brazil Characteristics of Poor Urban Households by Region and Household Type, 1984 (percent) Characteristic Northeast Southeast South Incidence of Poverty All households 33.2 11.1 12.4 All female-headed households (FHH) 35.7 14.3 14.9 )FHH w/children, no adults 62.3 36.5 37.4 Characteristies of Poor Households Average Family Size All households 5.6 5.2 5 All FHH 4.5 4.1 3.9 FHH wlchildren, no adults 4 3.8 3.6 Average Number of Chbldren All households 2.8 2.6 2.5 All FHH 2.1 1.9 1.9 FHH w/children, no adults 2.6 2.5 2.3 Average Number of Eaers All households 1.2 1 1 All FHH 0.9 0.8 0.7 FHH wlchildren, no adults 0.6 0.6 0.6 Dependency Ratio All households 1.3 1.3 1.3 All FHH 1.4 1.4 1.5 FHH w/children, no adults 2.6 2.5 2.3 Characterics of Heads of Poor Households Mean Income (in minmum salaries) All households 0.9 0.87 .088 All FHH 0.56 0.56 0.63 FHH w/children, no adults 0.63 0.67 0.73 Mean Education All households 2.6 3.6 3.3 All FHH 2.4 3 3.1 FHH w/children, no adults 2.5 3.5 3.6 Mean Age All households 44.3 42.4 42.9 All FHH 48.3 46.7 46.9 FHH w/children, no adults 41.7 38.9 39.8 Mean Hours Worked Per Week All households 33 30.1 31.3 All FHH 18.7 17.8 17.3 FHH w/children, no adults 22.6 25.1 23.7 Soutc: au&oes wbdon 35 Table 7: Brazil - Per Capita Income of Household Groups, Three Metropolitan Areas, 1984 (multiples of minimum wage) Percentile of Recife SAo Paulo Porto Alegre Income Distribution Lowest 5ib 0.12 0.28 0.29 Lowest 10th 0.18 0.42 0.41 Lowest 25th 0.32 0.75 0.72 Lowest 50th 0.60 1.36 1.35 Source: author's tabulations Note: Income refers to the sum of the income from all sources of all adult members of the household. Adult members include the household head, his or her spouse, and all other members age 18 or more. 36 Table 8: BrazDl - Characteristics of Female-Headed Households (FMI) in Three Metropolitan Areas, 1984 (percent) Recife Sao Paulo Porto Alegre Share of Total Households All FHH 21.8 16.5 18.5 FHH w/ minors 11.6 6.2 6.7 - no adults 3.7 2.6 3.1 Share of Total Minors All 15.0 9.5 10.8 Age 0-6 12.9 7.0 8.8 Age 7-9 15.6 10.0 11.5 Age 1o-14 17.6 13.6 15.0 Per Capita Inacome (mi[mum salaries) All households 1.3 2.5 2.6 FHH 1.1 2.4 2.6 FHH w/minors 0.6 1.2 1.3 Characteristics of Female-leaded Households Characteristics of Heads Percent economically active All FHH 44.0 55.0 53.0 FHH w/minors 51.0 68.0 64.0 Average age All FHH 51.0 49.0 50.0 FHH w/minors 47.0 43.0 43.0 Average income (minimum sa!aries) All FHH 1.7 3.0 3.4 FHH w/minors 1.4 2.6 2.9 Marital Status Single 37.6 37.8 40.7 Married 0.9 2.3 1.7 Widowed 27.0 .20.7 22.0 Divorced 34.5 3V.2 35.6 Characteristics of Households Average number of children All FHH 1.2 0.7 0.7 FHH w/minors 2.2 1.9 1.8 Dependency ratio All FHH 0.7 0.5 0.5 FHH w/nlinors 1.3 1.2 1.2 Source: author's tabulations 37 Table 9: Brazil - Distributdon of Households According to Per Capita Adult Income and Per Capita Household Income, Sao Paulo Metropolitan Area, 1984 Adult Income Total Income Percentile Percentile Lowest 5% 5-10% 10-25% 25-50% Above 50% Lowest 5% 4.4 0.4 0.2 0.0 -- 5-10% 0.6 3.9 0.5 0.1 -- 10-25% -- 0.8 13.2 1.0 -- 25-50% -- -- 1.1 23.2 0.7 Above 50% -- -- -- 0.7 49.3 Percent Reclassified 10.0 24.0 12.0 7.0 1.0 Source: author's tabulations otes: Adult income refers to the sum of the income from all sources of all adult members (age 18 and over) of the household. Total income refers to the sum of the income from all sources of all members of the household. Table 10: Brazil - Absolute and Relative Incidence of Household Type by Per Capita Income Class, Three Metropolitan Areas Recife Sbo Paulo Pr tO Alegre Income All FHH FHH w/minors All FHH FHH w/minors All FHH FHH w/minors Class FHH wlminors No Adults FHH w/ninors No Adults FHH w/minors No Adults Absolute Incidence Lowest 5% 40.9 35.5 21.0 27.0 19.3 11.2 26.7 19.9 13.2 5-10% 32.2 27.7 9.9 19.7 16.7 9.7 21.6 15.9 8.4 10-25% 24.1 15.8 5.9 19.5 11.2 4.1 21.8 11.3 4.5 25-50% 21.8 12.1 2.7 14.0 5.3 1.6 16.3 6.0 1.7 above 50% 18.1 60. 1.2 15.6 2.8 1.1 17.5 3.5 1.9 All 21.8 11.6 3.7 16.5 6.2 2.6 18.5 6.7 3.1 w Relative Incidence Lowest 5% 1.9 3.1 5.7 1.6 3.1 4.3 1.4 3.0 4.3 5-10% 1.5 2.4 2.7 1.2 2.7 3.7 1.2 2.4 2.7 10-25% 1.1 1.4 1.6 1.2 1.8 1.6 1.2 1.7 1.5 25-50% 1.0 1.0 0.7 0.8 0.9 0.6 0.9 0.9 0.5 above 50% 0.8 0.5 0.3 0.9 0.5 0.4 0.9 0.5 0.6 All 1.0 1.0 1.0 1.0 1.0 1.0 Soume: auzows tabulations Note: Income refers to the sum of the income from all sources of all adult members (age 18 or more) of the household. Relative incidence is the share of the house type in the income group weighed by its share in the population. A number higher than one indicates over-representation. 39 Table 11: Brazil - Average Per Capita Income and Income Gap by Household Type and Metropolitan Area, 1984 Average Per Average Area and Capita Income Income Gap Household Type (Minimum Salaries) (percent) Recife All households 1.27 FHH*s 1.04 0.22 FHHs w/children 0.58 0.57 S3o Paulo All households 2.52 FHHs 2.39 0.06 FHHs w/ children 1.14 0.56 Porto Alegre All households 2.54 FHHs 2.56 0.01 FHHs w/ children 1.26 0.52 Source: author's tabulations Note: Income Gap is equal to the average income of the household type (Yp) subtracted from the average income of all the other houses (Yc) divided by Yc or (Yc - YF / Yc). A negative value indicates the average income of household type is above that of other households. Table 12: Brazil - Simulated Reductions in Income Gap by Household Type and Metrpolitan Area, 1984 (percent or minimum salaries) Reave Differences Simutated Income Gap Area and Eamings Capcity Dependen Income Earnin Capacity Depenqd Housebold Type Capacity Utilization Ratio Gap Capacity Utilization Ratio Recife FHH 0.61 1.18 0.97 0.22 -.40 0.36 0.23 FHH wlchildren 0.55 1.12 1.70 0.57 0.18 0.63 0.45 Sao Paulo FHH 0.64 1.21 0.87 0.06 -.57 0.26 0.07 FHH w/children 0.55 1.18 2.12 0.56 0.19 0.64 0.40 0 Porto Algr FHH 0.72 1.20 0.93 -0.01 -.40 -0.21 -0.02 FHH wlchildren 0.62 1.17 2.33 0.52 0.32 -0.07 0.21 Source: author's tabulations Notes: Relative difference is the average value for a given type of household divided by the average for all households. inimcm is equal to the average income of the household type (YF) subtracted from the average income of all the other houses (Yc) divided by Yc or (Yc - YF I YC). A negative value indicates the average income of household type is above that of other households. It is reported in minimum salaries. Simulated income eaD is the estimated average income of household type (Y,) if this group had the mean of household group (Yc) on the simulation variable. Table 13: Brazil - Relative Differences on Earnings Capacity and Simulated Reductions in Income Gap by Household Type and Metropolitan Area, 1984 (percent) Area and Male Female % Earnings Head Other Adult % Area Household Type Earmings Earnings Female Earnings Earnings Non-Head Relative Difference Recife FHH .52 .97 2.44 .49 .94 0.94 FHH w/children .45 .87 2.38 OA3 0.80 0.89 Sic Paulo FHH .48 1.11 2.59 0.52 0.95 0.96 FHH wlchildren 0.35 .98 2.59 0.46 0.73 0.96 Porto Alegre FHH 0.45 1.17 2.44 0.61 0.93 1.09 FHH wJchildren 0.39 1.02 2.39 0.52 0.79 1.04 Sinulated Income Gap Sc Recife FHH 0.01 0.17 0.19 40.89 0.19 0.22 FHH w/children 0.31 0.42 0.54 40.43 0.44 0.60 Sio Paulo FHH -0.19 0.23 0.03 -1.11 0.03 0.06 FHH wkchildren 0.24 0.56 0.59 -.47 0.42 0.56 Pott Alegre FHH -0.23 0.40 0.01 -1.34 40.03 40.01 FHH wkchildren 0.21 0.55 0.49 40.57 0.43 0.52 Source: authorfs tabulations Notes: Relative difference is the average value for a given type of household divided by the average for all households. Income gap is equal to the average income of the household type (YF) subtracted from the average income of all the other houses (Yc) divided by Yc or (Yc - Y, / YC). A negative value indicates the average income of household type is above that of other households. It is reported in minimum salaries. Simulated income Lay is the estimated average income of household type (YF) if this group had the mnean of household group (Yc) on dke simulation variable. 42 Table 14: Brazil - Distribution of Minor Chlldren and Share in Female-Headed Households (FHH) by Income Class and Metropolitan Area, 1984 (percent) Share of Minors Recite Sio Paulo Porto Alegre Income Class All minors % In FHU All minors % In FIlE All minors % In FEB Lowest 5% 9 37 10 19 11 22 5-10% 10 23 10 15 10 15 10-25% 23 14 23 13 22 12 25-50% 28 12 27 6 26 9 Above 50% 30 9 30 5 32 7 All 100 15 100 9 100 I 1 source: authors ubadons 43 Table 15: Brazil - Relative Incidence of Minor Children in Female-Headed Households (FHH) by Income Class and Metropolitan Area, 1984 (percent) Share of Minors Recife Sao Paulo Porto Alegre Income Class Al minors %It F All minors % In F All minors % In FEIH Lowest 5% 1.8 7.4 2.0 3.8 2.2 4.4 5-10% 2.0 4.6 2.0 3.0 2.0 3.0 10-25% 2.3 1.4 2.3 1.3 2.2 1.2 25-50% 1.1 0.5 1.1 0.2 1.0 0.4 Above 50% 0.6 0.2 0.6 0.1 0.6 0.1 Souse: authors abulatons 44 Table 16: Brazil: Outcomes For Children By Type of Household, Metropolitan Area, 1984 (percent of children) Recife SAo Paulo Porto Alegre Children's Age FFH AU T-st FHH All T-st FHH All T-st and Status Age 7-9 - not in school 23 14 3.6 8 8 0.0 12 10 1.1 Age 10-14 - not in school 22 14 5.1 12 8 2.9 18 12 3.1 - in labor force 10 7 2.9 13 8 3.1 16 9 3.9 Age 10-14 In the labor force - not in school 69 56 2.2 42 36 0.9 69 67 0.2 Source: author's tabulations Note: T-st refers to the T-statisics for testing whether the difference between FHHs and all households is zero. 45 Table 17: Brazil: Outcomes For Children By Income Class and Metropolitan Area, 1984 (percent of children) Income Class Recife Sbo Paulo Porto Alegre Children's Age and Status <5 25-50 >50 <5 25-50 >50 <5 25-50 >50 Age 7-9 - not in school 22 12 4 14 6 2 17 6 3 Age 10-14 - not m school 21 14 4 12 7 3 18 10 4 - in labor force 10 6 2 11 8 4 13 8 3 Age 10-14 in the labor force - not inschool 61 47 42 40 32 23 69 65 56 Source: author's tabulations Note: Income class refers to percentile in the household per capita income distribution. 46 Table 18: Brazil - Simulated Outcomes for Children In Female Headed Households, 1984 (number or percent) Recife Sbo Paulo Porto Alegre .tcome by lldren's Age Poverty Poverty Poverty Status All FIH FHl Effect All FHH FHH Effect All FiH FHl Effect Age 7-9 -not in school 14 23 19 0.5 8 8 7 - 10 .12 11 0.6 Age 10-14 -not in school 14 22 17 0.7 8 12 11 0.2 12 18 16 0.4 -in the labor force 7 10 7 0.8 8 12 11 0.3 9 16 14 0.2 hildren Aged 10-14 In the Labor Force -not in school 56 69 65 0.3 36 42 40 0.2 66 67 70 - Source: author's abulations otes: FHH' is the simulated outcome for children in female-headed households. Poverty effect is the proportion of the difference between households explained by the difference in means. For example, 0.5 implies tha 50 percent of the difference In means Is attributable to the lower income of PHHS. Policy Research Working Paper Series Contact Title Author Date for paper WPS1249 Competitiveness and Environmental Piritta Sorsa February 1994 P. Kokila Standards: Some Exploratory Results 33716 WPS1250 Explaining Miracles: Growth William Easterly February 1994 R. Mattin Regressions Meet the Gang of Four 39026 WPS1251 Excise Taxes John F. Due February 1994 C. Jones 37699 WPS1252 On the Dangers of Decentralization Remy Prud'homme February 1994 TWUT 31005 WPS1253 Can Competition Policy Control 301 ? J. Michael Finger February 1994 M. Patena K. C. Fung 37947 WPS1254 What Are OECD Trade Preferences Alexander J. Yeats February 1994 J. Jacobson Worth to Sub-Saharan Africa? 33710 WPS1255 Intrahousehold Resource Allocation: Lawrence Haddad February 1994 P.Cook An Overview John Hoddinott 33902 Harold Alderman WPS1256 World Fossil Fuel Subsidies and Bjom Larsen February 1994 C. Jones Global Carbon Emissions in a Model 37699 wRth Interfuel Substitution WPS1257 Old-Age Security in Transitional Louise Fox February 1994 E. Vincent Economies 82350 WPS1258 Decentralizing Infrastructure: Richard Bird February 1994 WDR For Good or for III? 31393 WPS1259 The Reform of Fiscal Systems in Robin Boadway February 1994 C. Jones Developing and Emerging Market Sandra Roberts 37754 Economies: A Federalism Perspective Anwar Shah WPS1260 When Is a Life Too Costly to Save? George L. Van Houtven February 1994 A. Maranon Evidence from U.S. Environmental Maureen L Cropper 39074 Regulations WPS1261 A Political-Economy Analysis of Arvind Panagariya March 1994 N. Artis Free Trade Areas and Customs Ronald Findlay 37947 Unions WPS1262 Flexibility in Sr Lanka's Labor Market Martin Rama March 1994 P. Cook 33902 WPS1263 The Effects of Barriers on Equity Stijn Claessens March 1994 F. Hatab Investment in Developing Countries Moon-Whoan Rhee 35835 Policy Research Working Paper Series Contact Title Author Date for paper WPS1264 A Rock and a Hard Place: The Two J. Michael Finger March 199'4 M. Patenia Faces of U.S. Trade Policy Toward 37947 Korea WPS1265 Parallel Exchange Rates in Miguel A. Kiguel March 1994 R. Luz Developing Countries: Lessons from Stephen A. O'Connell 34303 Eight Case Studies WPS1266 An Efficient Frontier for International Sudhakar Satyanarayan March 1994 D. Gustafson Portfolios with Commodity Assets Panos Varangis 33732 WPS1 267 The Tax Base in Transition: The Case Zeljko Bogetic March 1994 F. Smith of Bulgaria Arye L Hillman 36072 WPS1268 The Reform of Mechanisms for Eliana La Ferrara March 1994 N. Artis Foreign Exchange Allocation: Theory Gabriel Castillo 38010 and Lessons from Sub-Saharan John Nash Africa WPS1269 Union-Nonunion Wage Differentials Alexis Panagides March 1994 I. Conachy in the Developing World: A Case Harry Anthony Patrinos 33669 Study of Mexico WPS1270 How Land-Based Targeting Affects Martin Ravallion March 1994 P. Cook Rural Poverty Binayak Sen 33902 WPS1271 Measuring the Effect of External F. Desmond McCarthy March 1994 M. Divino Shocks and the Policy Response to J. Peter Neary 33739 Them: Empirical Methodology Applied Giovanni Zanalda to the Philippines WPS1272 The Value of Superfund Cleanups: Shreekant Gupta March 1994 A. Maranon Evidence from U.S. Environmental George Van Houtven 39074 Protection Agency Decisions Maureen L Cropper WPS1273 Desired Fertility and the Impact of Lant H. Pritchett March 1994 P. Cook Population Policies Lawrence H. Summers 33902 WPS1 274 The New Trade Theory and Its Asad Alam March 1994 A. Alam Relevance for Developing Countries 87380 WPS1275 Female-Headed Households, Ricardo Barros March 1994 K. Binkley Poverty, and the Welfare of Children Louise Fox 81143 in Urban Brazil