Poverty, AIDS and Children’s Schooling A Targeting Dilemma

2020-12-08 来源：榕意旅游网

Poverty, AIDS and Children’s Schooling:

A Targeting Dilemma

Martha Ainsworth

Operations Evaluation Department

World Bank

Mainsworth@worldbank.org

Deon Filmer

Development Research Group

World Bank

Dfilmer@worldbank.org

World Bank Policy Research Working Paper 2885, September 2002

The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the 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 cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Policy Research Working Papers are available online at http://econ.worldbank.org.

We express our gratitude to the social protection and education anchor units (HDNSP and HDNED) and the research group (DECRG) of the World Bank for their financial support for this work. We thank Margaret Grosh and Donald Bundy for valuable comments on an earlier draft, and to Rodica Cnobloch and Janmejay Singh for excellent research assistance. We also thank participants in the Peace Corps Town Hall meeting in October 2001 and in a World Bank seminar in January 2002 for their comments on preliminary results, and Ken Wachter and participants at the PAA meetings in May 2002 for comments on the first draft.

Contents

Abstract..................................................................................................................................iii 1. Introduction....................................................................................................................1 2. Country coverage, data, definitions, and methodology.......................................................2

Source of data.................................................................................................................4 Definitions......................................................................................................................4 3. Results............................................................................................................................6

How prevalent are orphaned children and with whom do they live?....................................6 Are orphans more likely to be poor?...............................................................................13 Are orphans under-enrolled?..........................................................................................16 Is the gender gap in enrollment larger for orphans?..........................................................22 4. Conclusions..................................................................................................................27 References..............................................................................................................................29

Appendices

1. Data sets and sample sizes 2A. Orphan rates, ages 7-14 2B. Orphan rates, ages 15-17

3. Relationship to head among two-parent orphans, ages 7-14

4. Enrollment rates by orphan status and household wealth, ages 7-14

5. Changes in enrollment over time, by orphan status and household welfare 6. Enrollment rates by orphan status and household wealth, ages 15-17 7. Enrollment rates by orphan status and gender, ages 7-14

Tables

1. Poverty, schooling, and HIV/AIDS in the countries studied

2. Classification of countries by overall enrollment rates and difference in enrollment rates between

orphans and non-orphans, most recent survey

Figures

1. Percent of children 7-14 with missing orphan status 2. Percent of children orphaned by age, Mozambique 1997 3. Percent of children 7-14 who are orphans, West Africa

4. Percent of children 7-14 who are orphans, Eastern and Southern Africa 5. Percent of children 7-14 who are orphans, Latin America and Asia 6. Relation between two-parent orphan rate and HIV infection

7. Percent of single-parent orphans living with the surviving parent, West Africa

8. Percent of single-parent orphans living with the surviving parent, Eastern and Southern Africa 9. Percent of single-parent orphans living with the surviving parent, Latin America and Asia 10. Percent of households with an orphan aged 7-14

11. Percent of the wealthiest and poorest households with an orphan aged 7-14 12. Percent of 7-14 year olds who are orphans

13. Relation between enrollment rates and HIV prevalence, countries surveyed since 199514. Enrollment differentials by orphan status, ages 7-14

15. Enrollment rate by orphan status in lowest and highest quintiles, Zambia 1998 16. Changes in enrollment rate by orphan status and wealth, Uganda 1995-2000 17. Changes in enrollment rate by orphan status and wealth, Kenya 1993-98 18. The gender gap in enrollment, all children

19. The gender gap in enrollment among orphans and non-orphans, selected countries 20. Gender differences in enrollment, orphans and non-orphans compared

21. Gender gap in enrollment for orphans and non-orphans in the poorest and richest of the sample

Abstract

This paper analyzes the relationship between orphan status, household wealth, and child school enrollment using data collected in the 1990s from 28 countries in Sub-Saharan Africa, Latin America, the Caribbean, with one country in Southeast Asia. The findings point to considerable diversity—so much so that generalizations are not possible. While there are some examples of large differentials in enrollment by orphan status, in the

majority of cases the orphan enrollment gap is dwarfed by the gap between children from richer and poorer households. In some cases, even children from the top of the wealth distribution have low enrollments, pointing to fundamental issues in the supply or demand for schooling that are a constraint to higher enrollments of all children. The gap in

enrollment between female and male orphans is not much different than the gap between girls and boys with living parents, suggesting that female orphans are not

disproportionately affected in terms of their enrollment in most countries. These diverse findings demonstrate that the extent to which orphans are under-enrolled relative to other children is country-specific, at least in part because the correlation between orphan status and poverty is not consistent across countries. Social protection and schooling policies need to assess the specific country situation before considering mitigation measures.

iii

1. Introduction

Two decades into the AIDS pandemic, a cure for AIDS is still not at hand and the international community is becoming increasingly concerned with the impact of high adult AIDS mortality on child welfare, particularly on the welfare of orphans. In addition, many countries are suffering from civil unrest and post-conflict situations, resulting in war orphans and displaced children. AIDS and conflict are adding to an already elevated number of orphans from high adult mortality in developing countries.

While the number of affected children is potentially large, very little is known about the

welfare consequences of being an orphan in developing countries, where poverty is widespread and human capital is low. One of the most frequently expressed concerns is that school-aged orphans will be forced to drop out of school or will never enroll, either because their guardians cannot afford the costs of schooling, the child is needed for income-generating or other

economic activities, or the guardians simply have less interest in the welfare of children who are not their own (Foster and Williamson 2000, Nyambedha, Wandibba, and Aagaard-Hansen 2001, USAID 2000). This has prompted calls for governments to subsidize the schooling of orphans (Subbarao, Mattimore, and Plangemann 2001, USAID 2000, World Bank 2002a). Yet, to the extent that they drop out of school, orphans in the poorest countries will swell the ranks of an already large group of poor children who are not enrolled: In 1997, at least 67.5 million primary-aged children were not in school worldwide, of which 58 million were living in low-income countries, 31.5 million were living in South Asia and 25 million were living in sub-Saharan Africa (World Bank 2000).

The extent to which orphans are under-enrolled relative to other children and the reasons

for non-enrollment have not been systematically reviewed. Most studies have focused

exclusively on orphans with no comparison group of children with living parents, and in many cases analyze the hardest-hit orphans (e.g., Kitonsa and others 2000, Nyambedha, Wandibba, and Aagaard-Hansen 2001). It is not clear, for example, whether orphaned children are worse off than other equally poor children—therefore requiring a targeted intervention linked to their special needs—or whether the impact of becoming an orphan is to swell the already large group of poor or uneducated children.1 In the latter case, one might argue for policies that will raise the levels of schooling of the unenrolled poor, orphan and non-orphan alike. In fact, there are reasons to believe that AIDS orphans may not be worse off than the poorest children and are possibly not as poor as other orphans. While adult mortality from other infectious diseases disproportionately affects the poor, AIDS strikes both the poor and the non-poor. Early in the African epidemic, the adults most likely to be infected were in fact those who were most mobile (traders, businessmen, fishermen, transport workers), not the poorest (World Bank 1999). Thus, orphan status alone may not be a good correlate of poverty or adverse outcomes.

This paper examines the relation between parental survival and two dimensions of

welfare—poverty and school enrollment?to answer the question of whether orphan status is a

1. An exception is the study by Lloyd and Blanc (1995), which uses a multiple regression model that controlled for living standards to predict enrollment of children 10-14 in seven African countries.

good predictor of lower welfare.2 We use large and nationally representative datasets from 28 developing countries and four regions (Africa, Latin America, the Caribbean, and Asia) in a primarily descriptive exercise to examine the welfare correlates of orphan status among children 7-14 and, for a few countries where data permit, those aged 15-17. We anticipate that the impact of being an orphan on welfare will depend on many country-specific factors, including the overall poverty rate, the socioeconomic status of households that experience adult mortality, customs and demographic factors like child fostering and the extended family, existing demand for child schooling, and the public policies already in place. While we can’t explore all of these explanatory factors, we expect that the results will demonstrate considerable diversity in the relation between being an orphan and welfare outcomes and therefore suggest diverse policy responses. This point is important in light of the current tendency to assume that the experience of the hardest-hit countries can be generalized to all countries hit by AIDS, and that there is a single, preferred policy solution based on that example.

The paper is organized into four major sections. Section 2 describes the datasets and

define the key variables. Section 3 contains the findings on the following questions: (1) How prevalent are orphans and with whom do they live? (2) Are orphans more likely to be poor? (3) Are orphans less likely to be enrolled in school? (4)Is the gender gap in enrollment greater for orphans? Section 4 summarizes the results, identifying key policy issues and a future agenda for research.

We find considerable diversity in the relation between orphan status and poverty?so

much so that generalizations are not possible. While there are some examples of large differentials in enrollment by orphan status, in the majority of cases the size of the orphan enrollment gap is dwarfed by the gap in enrollment between children at the bottom and the top of the income distribution. In some cases, even children from the top of the income distribution have low enrollments, pointing to fundamental issues in the supply or demand for schooling that are a constraint to higher enrollments of all children, whether or not their parents are alive. When orphan enrollment gaps persist, even among the non-poor, these differences are very

likely due to factors specific to being an orphan that cannot easily be addressed through policies on subsidizing school fees and uniforms. Finally, we find in most cases that the gap in

enrollment between female and male orphans is not much different than the gap between girls and boys with living parents, suggesting that female orphans are not disproportionately affected in terms of their enrollment in most countries.

2. Country coverage, data, definitions, and methodology

The 28 countries in this study were selected based on data availability. They

nevertheless achieve good geographic coverage within Sub-Saharan Africa and more limited coverage of Latin America, the Caribbean and a single country in Southeast Asia (Table 1).

2. The enrollment rate captures only one dimension of schooling. Even if the enrollment rate were 100 percent, it does not tell us about attendance, repetition rates, completion rates, drop out rates, or the ultimate variables of interest, learning and achievement. These variables may also be affected by orphan status and poverty but they were not available for analysis.

Table 1. Poverty, schooling, and HIV/AIDS in the countries studied

Percent of the Gross

Adult HIV population living on primary

GNP/ less than $1/day enrollment prevalence capita (%) ratio 1998 1999 Year Percent 1997

Male adult Female adult

mortality mortality rate/1000 rate/1000 1998 1998

Country

Western Africa Benin 380 .. 78 2.45 367 308 Burkina Faso 240 1994 61.2 40 6.44 547 522 Cameroon 610 .. 85 7.73 336 303 Central African Rep. 300 1993 66.6 .. 13.84 576 488 Chad 230 .. 58 2.69 454 388 Côte d’Ivoire 700 1995 12.3 71 10.76 526 513 Ghana 390 .. 79 3.60 282 230 Guinea 530 .. 54 1.54 404 404 Mali 250 1994 72.8 49 2.03 404 325 Niger 200 1995 61.4 29 1.35 453 352 Nigeria 300 1997 70.2 98 5.06 401 339 Senegal 520 1995 26.3 71 1.77 456 385 Togo 330 .. 120 5.98 488 444 Eastern Africa Kenya 350 1994 26.5 85 13.95 442 418 Madagascar 260 1993 60.2 92 0.15 273 231 Tanzania 220 1993 19.9 67 8.09 521 482 Uganda 310 1992 36.7 74 8.30 579 615 Southern Africa Malawi 210 .. 134 15.96 464 483 Mozambique 210 1996 37.9 60 13.22 408 364 South Africa 3,310 1993 11.5 133 19.94 282 194 Zambia 330 1996 72.6 89 19.95 521 545 Zimbabwe 620 1990-91 36.0 112 25.06 470 417 Latin America Brazil 4,630 1997 5.1 125 0.57 279 139 Guatemala 1,640 1989 39.8 88 1.38 297 195 Nicaragua 370 1993 3.0 102 0.20 208 139 Caribbean

Dominican Republic 1,770 1996 3.2 94 5.07 153 96 Haiti 410 .. .. 5.17 432 339 Southeast Asia Cambodia 260 .. 113 4.04 357 309 Definitions: Population living on less than $1/day: Percent living at less than $1.08/day at 1993 international prices (corresponding to $1/day in 1985), with prices adjusted for purchasing power parity; Gross primary enrollment ratio (GPER): primary enrollments as a percent of children of primary school age; Adult HIV

prevalence: percent of adults 15-50 infected with HIV and alive; Adult mortality rate: number of people aged 15-60 per thousand who will die between the ages of 15-60 at the current age-specific mortality rates. The GPER can exceed 100 percent because of enrollment of over-age children.

Source: World Bank (2000), tables 1.1, 2.7, 2.10 and 2.18, and UNAIDS (2000).

Twenty-four are low-income countries with GNP per capita of less than US$1,000.

Among the low-income countries, the percent of the population living on less than one U.S. dollar per day, where measured, ranges from 12-73 percent. Gross primary enrollment ratios (GPER)—the number of children in primary school divided by the number of children of

primary age?are also relatively low. Thirteen countries have GPER of less than 80 percent and only seven have ratios of more than 100 percent. Only 7 are “on track” to achieve the

international goal of universal basic education by 2015, and 8 are “seriously off-track” to reach the goal (World Bank 2002b).

Levels of HIV infection are geographically concentrated, with the highest rates of 20

percent or more in Southern Africa and the lowest rates below 1 percent in Latin America. HIV is clearly a contributing factor to high levels of adult mortality in the hardest-hit countries, but not the only factor. Several countries have high adult mortality even with low HIV prevalence (for example, Guinea, Niger, and Mali) while countries like the Dominican Republic and South Africa have relatively lower adult mortality despite high HIV infection rates. Thus, AIDS is only one of several causes of the adult mortality that creates orphans; in some of the countries it is likely the major cause, while in others orphans are created by high levels of baseline adult mortality. It is also worth noting that in 24 of the 28 countries, men have higher mortality than women.

Source of data

We use datasets from 39 nationally representative household surveys dating from 1992–

2000 that collected data on orphan status, school enrollment, and variables that measure household living standards. Thirty-four of the datasets are Demographic and Health Surveys (DHS) and five are Living Standards Surveys (see Appendix 1). Eight countries have a survey for more than one year, which permits analysis of trends in enrollment and orphan status. We analyze primarily children in the age group 7-14 because the DHS generally collects orphan status only for children under 15 and a lower boundary of seven years of age enhances cross-national comparability. To the extent that the children in this age group are enrolled, almost all would be enrolled in primary school. The total sample sizes for children 7-14 range from 5,000 – 24,500 but most are on the order of 5,000-10,000 (Appendix 1). All results are weighted to be nationally representative.

Definitions

Orphan. We consider three mutually exclusive types of orphan?a child who has lost

his/her mother only (“maternal orphan”), his/her father only (“paternal orphan”), or both parents (“two-parent orphan”). Because the data are from household surveys, institutionalized orphans or children not living in households are not included in this analysis. In addition, between 0 and 7 percent of children age 7-14 could not be classified according to their orphan status because respondents were not certain about the survival of at least one parent, usually the father (Figure 1). 3 For 18 of the countries, between 1-3 percent of the children had missing orphan status.

3. Excluding Nigeria, where 7 percent of children could not be classified, the range was between 0-4.4 percent. Sensitivity analysis was carried out on the missing orphans category. While the percentage of children

Figure 1. Percent of children 7-14 years old with missing orphan status 765Percent43210DomUginicaanBun Rda 2rkiep00na 190 Fa96Nicso 1ara99gu3a 1997MaliTo 199gZao 15m998bia Be1998niGhn 19an93aM 199ala8wi 1CA992R Co1te Ch994d'Iadvo 19ire96 1994HaitiCaNig 199mer 4ero199Taon 18nz99Caania8m 19bodia96Gu 19ine99a 1GuBr999aazMtemil 19adala9ag 16as99ca9Ser 19ne97gaMKl 199ozen3amySobiqa 19uthue98 A 19fr9Zimica 17ba998bweNi 199geria9 1999Source: Authors' calculations, DHS and LSMS datasets. Enrollment. The enrollment rate is the percent of children aged 7-14 who are reported

as currently “in school”, irrespective of the grade in which they are enrolled. This enrollment rate cannot exceed 100 percent. Note that this is quite different from the Gross Primary Enrollment ratio, which can exceed 100 percent because older children who started late or

repeated grades are included in the numerator. It is different from the Net Enrollment Rate since it does not take into account the grade attended.

Welfare/poverty. The DHS do not measure household consumption or income, but they

do collect information on the ownership of assets and housing conditions, as do the living standards surveys we use. We have computed for every household a wealth index that is a continuous variable based on the factor loadings from the first component of a principal component analysis of asset ownership and housing characteristics:

• radio, refrigerator, television, bicycle, motorcycle, car • source of drinking water, type of toilet facility

• electricity, number of rooms for sleeping, “finished” flooring or roofing.

We then assigned to every individual in each survey the wealth index for his/her household. Individuals were ordered from the lowest to the highest index in their country and, based on this, we defined quintiles of the wealth index across all individuals. Because of the problem with

small cell sizes on two-parent orphans, we have aggregated children in the lowest 40 percent, the middle 40 percent, and the upper 20 percent of the wealth distribution based on the distribution

who are orphans is affected, the average enrollment rates, or the distribution of orphans by household wealth is not substantially changed. Children with missing orphan status were not included in either the orphan or non-orphan enrollment rates reported here.

of the population. The wealth index is used to place children relative to a given distribution of wealth within a country; it does not map easily into a typical poverty rate, which is usually based on an absolute level of welfare. The wealth index is defined within a country for a given survey; it cannot be compared in an absolute sense across countries or for different surveys in the same country.4 The approach is described more fully in Filmer and Pritchett (2001) and is applied to the analysis of wealth gaps in education in Filmer and Pritchett (1999) and Filmer (2000).

3. Results

How prevalent are orphaned children and with whom do they live?

While the prevalence of orphans varies across countries, in all of them the share of

children who are orphans increases with age. The pattern found in Mozambique is typical

(Figure 2): orphans are relatively rare among pre-school children but rise to much higher levels among school-aged children. In addition, the percent of children who are paternal orphans

generally exceeds the percent who are maternal orphans at all ages, in some countries by a factor of two or three. This reflects the higher age-specific mortality of men and the fact that women usually marry older men. The vast majority of orphans, therefore, have lost one parent. The share who have lost both parents is quite small, particularly in the pre-school age group. Among school-aged children (7-14) in the 28 countries and 39 datasets studied, the percent of children 7-14 who are two-parent orphans ranged from 0.2 percent (Dominican Republic) to a high of 4.5 percent (Uganda).

The small number of two parent orphans poses problems for comparing their welfare

with other children. In the unweighted samples of children used in this study, there were fewer than 20 two-parent orphans aged 7-14 in 2 of the 39 datasets and in 9 other datasets there were fewer than 50. This becomes more of a problem when the samples are disaggregated by level of welfare. In 21 of the 28 countries, we couldn’t compute the enrollment rate for 2-parent orphans in the richest quintile because there were fewer than 20 children who had lost both parents. Aside from these common patterns in all developing countries, there are important

differences across and within regions in the share of children who are orphans and the ratio of paternal to maternal orphans (Appendix 2). In West Africa, 4 to 10 percent of school-aged

children are paternal orphans, roughly twice the proportion who are maternal orphans (Figure 3). Relatively few (1.6 percent or less) are two-parent orphans. Eastern and Southern African levels of paternal orphans are generally higher—6 to 13 percent—while maternal orphan rates are similar to West Africa (Figure 4). As a result, paternal orphan rates are 3 to 5 times higher than maternal rates. The reason for the much higher paternal orphan rate is not known; it could reflect the impact of the AIDS epidemic or higher male mortality from other causes in the region. An exception is Mozambique, which has the highest maternal orphan rate of any of the countries

4. In other words, a child with a value of the wealth index placing him/her in the lowest 40 percent of the distribution in country A, might not necessarily have the same level of welfare of a child in the lowest 40

percent of the distribution in country B. For countries with living standards surveys, the ranking of children by this 40/40/20 distribution was compared, using measures of household consumption per adult and the wealth index. There is substantial overlap in the group classifications, and enrollment rates across groups are very similar when using the different methods to rank individuals. In countries where consumption was available we nevertheless used the wealth index for consistency.

studied, nearly 7 percent. With the exception of three countries—Zambia, Zimbabwe, and

Uganda—the two-parent orphan rate in East Africa is under 2 percent. Finally, in Latin America, the Caribbean and Cambodia, all orphan rates are substantially lower (4-5 percent paternal, 1-2 percent maternal and 1 percent or less two parent orphans). A notable exception is Haiti, where the pattern and level are closer to those found in West Africa.

Figure 2. Percent of children orphaned by age,

Mozambique 1997 302520Percent15105001234567891011121314Age in yearsPaternal orphanMaternal orphanTwo-parent orphanTotal orphansFigure 3. Percent of children 7-14 who are orphaned, West Africa West Africa141210Percent86420Nigeria 1999Ghana 1998Mali 1995Niger 1998Senegal 1Co99te 3d'Ivoire 1994BeninBu 19rki93na Faso 1993Guinea 1999Chad 19Ca96meroon 1998Togo 1998CAR 1994Two-parent orphanMaternal orphanPaternal orphan 7

Figure 4. Percent of children 7-14 who are orphans,

Eastern & Southern Africa

& Southern Africa141210Percent86420Malawi 1992Kenya 1998Zambia 1998Tanzania 1996Uganda 2000Mozambique 1997South Africa 1998Madagascar 1997Zimbabwe 1999Two-parent orphansMaternal orphansPaternal orphansFigure 5. Percent of children 7-14 who are orphans, Latin America and Asia

141210Percent86420Dominican Rep.Nicaragua 19971996Guatemala1999Brazil 1996Cambodia 1999Paternal orphanHaiti 1994Two-parent orphansMaternal orphan What accounts for the variation in orphan rates? There is generally a positive correlation

between orphan rates and HIV prevalence (the percent of people living with HIV), but with a great deal of variation (Figure 6). This is because orphan rates are affected by AIDS through cumulative AIDS deaths, while HIV prevalence is a measure of the percent of the population that is infected and still alive. Because of the long asymptomatic period between HIV infection and

AIDS mortality, countries where HIV has increased rapidly and recently may have high HIV prevalence but low AIDS mortality and therefore only a small impact on orphan rates (e.g., South Africa). In countries with mature epidemics, HIV prevalence may have declined or

stabilized in part because of high mortality rates (e.g., Uganda). Thus, the percentage of children orphaned may be high even though HIV prevalence has declined. Moreover, orphan rates also reflect adult mortality from causes other than AIDS (occupation-related, war-related, maternal causes).

5Figure 6. Relation between two-parent orphan rate and HIV infection parent orphan rate3020315210150NGuigeratemalaBrazilMaliGhMaadanagascarDominican ReNicparaguaCCahadmbodiaTCaogomerooTannSozaniauth AfricaNigeriaKenyaGMouinzaeambiqZimuebabweZambiaUganda0Two-parent orphansAdult HIV (%) 1999Source: Authors' calculations for countries surveyed since 1995; UNAIDS data for HIV prevalence.Pursuing this point further, a regression of the two-parent orphan rate for the 28

countries in Table 1 on the HIV infection rate in 1999 and the 1998 female adult mortality rate (amr) reveals the following result (t-statistics in parentheses, adjusted R2 = .5014)

(1)

Two-parent orphan rate = 0.055 * [1999 HIV rate] + 0.0037 * [1998 female amr].

(2.49)

(3.26)

We would expect that the HIV infection rate contributes to the 2-parent orphan rate

through the adult mortality rate, but when we control for HIV infection, the adult mortality rate (net of the influence of HIV) is still significantly associated with the orphan rate, indicating that there is substantial adult mortality not accounted for by the contemporaneous HIV infection rate. At the mean values for this 28-country sample, a 1 percent proportionate increase in HIV infection (from 7.4 to 7.5 percent) is associated with an increase of 0.32 in the two-parent

orphan rate, while a 1 percent proportionate increase in the female adult mortality rate (from 356

Adult HIV prevalence (%), 19994Percent of 7-14 year olds who are 2-parent orphans25to 359) is associated with an increase in the mean two-parent orphan rate of 1.05 (the mean two-parent orphan rate in the 28 countries was 1.26 percent). When HIV is not controlled for (results not shown here), a 1 percent proportionate increase in the adult mortality rate is associated with an increase of 1.38 percent in the two-parent orphan rate.

Another way of gauging the contribution of the AIDS epidemic to the orphan rates is to

compare orphan rates over time, before and after the AIDS epidemic. Unfortunately, data are not available for the orphan rate for both maternal and paternal orphans for school-aged children (7-14) before the AIDS epidemic. However, the share of children 0-14 who had lost their mothers or both parents was about 2 percent in East Africa before the AIDS epidemic?1.91 percent in Kenya and 2.44 percent in Uganda in the 1969 censuses and 2.23 percent in Tanzania in the 1978 census (World Bank 1999). The rate in Kenya was basically unchanged as of the 1993 DHS (1.8 percent) but had risen by 50 percent (to 2.7 percent) in the1998 DHS. In Tanzania, the maternal and two-parent orphan rate for children 0-14 actually declined between the 1978 and 1988 censuses (to 1.96 percent) before rising by 40 percent (to 2.8 percent) by the time of the 1994 DHS. In Uganda the rate had doubled by 1995 (to 5 percent) and reached 5.7 percent by the 1999/2000 National Household Survey (a 130 percent increase since 1969). Thus, in these three East African countries, the maternal and two-parent orphan rates have risen by 40-130 percent since the onset of the AIDS epidemic. We have no information on the pre-AIDS orphan rates in similar age groups for other regions of Africa or the world, but they would have reflected the prevailing adult mortality rates due to other causes.

In the most recent surveys for the 28 countries in this study, most orphans aged 7-14 are

single-parent orphans and most single-parent orphans live with the surviving parent (Figures 7-9). In West Africa, between 50 and 75 percent live with the surviving parent and this is roughly the same for paternal and maternal orphans. Interestingly, a relatively high proportion of maternal orphans live with their father. In East Africa, in all but Madagascar and Zambia, paternal orphans are much more likely to live with their mother compared to West Africa, and maternal orphans are much less likely to live with their father. It is unclear why. In Nicaragua, Guatemala, Cambodia, and Brazil, 80-90 percent of paternal orphans live with their mother. Nicaragua and Haiti seem to have a pattern similar to Eastern and Southern Africa, while the Dominican Republic has a pattern similar to that in West Africa.

Figure 7. Percent of single-parent orphans living with surviving parent,

West Africa

90807060Percent50403020100Chad 1996Guinea 1999Ghana 199Bu8rkina Faso 1993Cameroon 1998Cote d'Ivoire 1994Mali 1995Senegal 1993Kenya 1998CAR 1994Maternal orphanPaternal orphan Figure 8. Percent of single-parent orphans living with surviving parent, Eastern and Southern Africa 80706050403020100Uganda 2000Zambia 1996Mozambique 1997Zimbabwe 1999Tanzania 1996Malawi 1992Madagascar 1997South Africa 1995PercentMaternal orphanPaternal orphan 11

Nigeria 1999Benin 1993Niger 1998Togo 1998Figure 9. Percent of single-parent orphans living with the surviving parent,

Latin America and Asia 90807060Percent50403020100DominicanRep. 1996Haiti 1994Nicaragua1997Guatemala1999Paternal orphanCambodia1999Brazil 1996Maternal orphan Where an orphan lives is likely to be influenced by available alternatives. For example,

in West Africa, and to a lesser extent in East Africa, child fostering within the extended family is relatively common, and thus single-parent orphans are less likely to live with a surviving parent. By contrast, in Cambodia, where previous regimes demolished the extended family structure, orphans may have no choice but to live with a surviving parent. The large degree of mobility among men engaged in mining in Southern Africa may explain why so few maternal orphans live with their fathers. These are all hypotheses that warrant investigation to fully understand the reasons for and welfare consequences of these observed patterns of living arrangements. Most of the household surveys collected information on the relation of every child to the

head of the household. Two parent orphans, by definition, are not living with their parents but usually are living with a relative (Appendix 3).5 Unfortunately, interpretation of the results of the “relation to head” question in these surveys is complicated by the fact that “adopted/foster child” was included as a category in nearly all of them and it is not mutually exclusive with the other categories. Many of the “adopted/foster” children of the head may be the grandchild, sibling or niece or nephew of the head, while it is probable that many of the two-parent orphans living with other relatives have effectively been adopted, if not formally. Further, foster and adopted children were recorded in a single category, yet the two terms often have different meanings, with fostering being a temporary situation and adoption being permanent, and fostering frequently occurring between families of relatives (e.g., Ainsworth 1996). This category

probably was likely defined and interpreted in the cultural context of each country and probably

5. Note that the number of two-parent orphans aged 7-14 in these samples ranged from fewer than 20 in the Dominican Republic to more than 700 in Zambia (1998). In 25 of the datasets there were fewer than 100.

not strictly comparable across countries. If we assume that most children in the “adopted/foster child” category are in fact related to the head (probably a good assumption in the African

countries, at any rate), then at least 90 percent of two-parent orphans in 28 of the 36 datasets for which information is available were living with relatives. The notable exceptions are in Haiti, Guatemala, Madagascar, Benin, Brazil, and Senegal, where from 12-26 percent of two parent orphans of primary school age (7-14) were not related to the head of household. Because of the overlap between ‘adopted/foster’ and other categories, the percent of children listed as living with a grandparent should be interpreted as a lower bound. With this in mind, at least half of two parent orphans in Guatemala, Malawi, Nicaragua, and Zimbabwe were living in grandparent-headed households and at least 40 percent in South Africa and Uganda. In most countries, at least 10 percent of two parent orphans aged 7-14 lived in a household headed by a sibling. It was extremely rare for two-parent orphans in this age group to be listed as the household head (only 4 countries registered any cases), although it is possible that the DHS (with the main objective of interviewing adult women) may have excluded households comprising only children in some countries (Bicego, Rutstein, and Johnson 2002).6 However, systematic investigations in several countries have confirmed that child-headed households are rare (Ainsworth, Ghosh and Semali 1995, Gilborn and others 2001).

Are orphans more likely to be poor?

The relation between orphan status and poverty can be viewed from the perspective of

whether poor or non-poor households are more likely to have resident orphans or whether orphans are more likely to live in poor or non-poor households compared with non-orphans. There are at least two reasons why non-poor households may be more likely to have orphaned children than poor households; first, the orphan’s parents may have been from among the non-poor and, second, orphans may be sent to the homes of relatives most capable of caring for them.7

Figure 10 shows the percent of households with an orphan aged 7-14 in the most recent

survey for each of the 28 countries. With the exception of two outliers (Zambia and Uganda, with 16.5 and 19.7 percent of households with orphans, respectively), between 4 and 13 percent of households have a school-aged orphan. This is an enormous range, affected not only by adult mortality from AIDS and other causes, but also the extent to which orphans are concentrated in a few households or distributed over a larger number of households. The extent of

institutionalization of orphans could also be a factor reducing the share of households with an orphan, although we have no information on the percent of children who are in orphanages in these countries.

6. An alternative explanation might be that two parent orphans who head households are in that role for a very short time before they are absorbed by the extended family.

7. Ainsworth, Beegle, and Koda (2002) find that the deceased parents of orphans had roughly one more year of schooling, on average, than did the living parents of non-orphans in the Kagera region of Tanzania in the early 1990s. Gilborn and others (2001) find that current and prospective guardians of orphans had higher socioeconomic status than parents living with AIDS in Luwero and Tororo Districts of Uganda.

Figure 10. Percent of households with an orphan aged 7-14

16.519.715Percent10.811.911.411.611.611.711.811.812.513.21050

53.63.85.25.86.46.57.288.28.28.78.98.99.19.8If a program were to target interventions to households with resident orphans, would it

be channeling resources to the poorest households? In Figure 11, we plot the share of the richest 20 percent of households with an orphan 7-14 (on the y-axis) against the share of the poorest 40 percent of households with an orphan (on the x-axis). A 45-degree line from the origin indicates the points where exactly the same share of households in the poor and non-poor have orphans. In countries located above the 45-degree line non-poor households are more likely than poor households to have an orphan; in countries below the line poor households are more likely to have an orphan. These results show that, poor households are equally likely to have an orphan as non-poor in 9 cases. In 10 cases, poor households were more likely to have an orphan than were non-poor households (e.g., Senegal, Zimbabwe, Cambodia) , and in 9 cases non-poor

households (the top 20 percent) were more likely to have an orphan. In Uganda in 1999/2000, for example, 17 percent of the poorest 40 percent of households had an orphan, while 23 percent of the households in the highest fifth of the welfare distribution had an orphan. In contrast, in South Africa in 1995 poor households were three times more likely to have an orphan than were non-poor households (nearly 15 percent of the poorest households had an orphan compared to only about 5 percent of the least poor households).

Dominican RBrazepubil 1996lic 1Nic996aragua 19Gh98Guana 1Caatembma998odla 1ia SE999S 19Nig99eria Ma1999li 1995-Nig96er Ke1998Mandaya 1ga9Coscar98te 19d'Iv97oire 1994ChadMa 1998laTawi 19nzan92ia 19Be96nCainme 199roo6BuC.An 19rkin.R98a F. 19as94-o 195992-93ToZimgoba 199bw8RSe 19A O99SeHS 1nega995l 1992Gu-93ineaH 19Moaiti 1999zamb94-9Z5aUgiqanmbiaue 1da ULCM997NHS 1S 1999989-200014

Figure 11. Percent of wealthiest and poorest households

with an orphan aged 7-14

Mozambique 1997Zambia 1998

Uganda 1999/00Rich households are aslikely as poor

households to have anorphan (45oline)

25Percent of the richest 20% of households with an orphan aged 7-14C.A.R. 1994/520

Haiti 1994/5Burkina Faso 1992/3Cote d’Ivoire 199415

Senegal 1992/3Zimbabwe 199910

South Africa 19955

Cambodia 199900

510152025

Percent of the poorest 40% of households with an

orphan aged 7-14

Note: Solid symbol indicates that the difference between rich and poor households is significant at 10 percent level

Figure 11 speaks to the distribution of households according to whether they have an

orphan, but not the distribution of orphans across households. Both poor and non-poor

households could have equal probabilities of having an orphan, but poor households could have a greater number per household. Figure 12 show the orphan rate (the percent of children who are orphans) in the poorest 40 percent and richest 20 percent of households, using the wealth index. Along the 45-degree line, the share is equal; above the line non-poor households have a higher orphan rate and below the line poor households have a higher orphan rate. The data points with open circles indicate that the difference in orphan rates between the two groups was not statistically significant. In 16 of the 28 countries for the latest year there is no statistically significant difference in the orphan rates for poor and non-poor households. In Uganda and Haiti—both of which are heavily hit by the AIDS epidemic—the orphan rate in non-poor households seems substantially higher than in poor households, but the results are not

statistically significant.8 On the other hand, for 12 countries poor households have higher orphan rates than non-poor households and in a few countries this is large. In particular, we see that many of the same countries where poor households are more likely to have an orphan, they also have higher orphan rates, for example, South Africa, Cambodia, and Zimbabwe.

8. Bicego, Rutstein, and Johnson (2001) found, similarly, that double orphans in the age group 0-14 were less likely than non-orphans to be living in poor households in Niger, Kenya, and Tanzania, using DHS data.

Figure 12. Percent of 7-14 year olds who are orphans

Uganda 1999/00 (UNHS)Haiti 1994/5Orphans are equally as likely to be in poor as in rich households(45oline)25Percent of 7-14 year olds who are orphans in the richest 20% of householdsBurkina Faso 1992/320

Zimbabwe 1999Mozambique 199710

Cameroon 19985

Ghana 1998Cambodia 1999South Africa 1995Madagascar 199700

510152025Percent of 7-14 year olds who are orphans in the

poorest 40% of households

Note: Solid symbol indicates that the difference between rich and poor households is significant at 10 percent level

In summary, orphans live in both poor and non-poor households. Households with orphans are not necessarily the poorest households, and in some countries the poorest households are less likely to have orphans because of the natural coping processes in which those with the most resources take in orphaned children or because of the socioeconomic distribution of HIV

infection. In more than half of the countries in this study, children in poor families are no more likely to be orphans than are children in non-poor families, while in the remainder poor children are more likely to be orphans.

Are orphans under-enrolled?

The countries most affected by the AIDS epidemic in Sub-Saharan Africa have among

the lowest enrollment rates in the world. Estimates are that by 2015 half of countries in sub-Saharan Africa will not reach the Education for All goals.9 In a quarter of the 28 countries

9. The Education for All goals are (1) expanding and improving comprehensive early childhood care and education, especially for the most vulnerable and disadvantaged children; (2) ensuring that by 2015 all

children, particularly girls, children in difficult circumstances and those belonging to ethnic minorities, have access to and complete free and compulsory primary education of good quality; (3) ensuring that the learning needs of all young people and adults are met through equitable access to appropriate learning and life skills programmes; (4) achieving a 50 per cent improvement in levels of adult literacy by 2015, especially for

women, and equitable access to basic and continuing education for all adults; (5) eliminating gender disparities in primary and secondary education by 2005, and achieving gender equality in education by 2015, with a focus on ensuring girls’ full and equal access to and achievement in basic education of good quality; (6) improving

studied, fewer than 50 percent of 7-14 year olds are enrolled in school in the most recent household survey. In about half, 50-80 percent are enrolled and in the remaining quarter,

enrollment exceeds 80 percent. Aggregate enrollment rates are affected by many economic and policy factors governing the supply and demand for education as well as labor market conditions that are only indirectly affected by the AIDS epidemic, so it is not surprising that there is no correlation between adult HIV prevalence and enrollment across countries (Figure 13). Nevertheless, within countries and particularly in those hardest hit by AIDS or conflict, policymakers are concerns that orphans may be under-enrolled.10 If true, then the growing number of orphans might pose special challenges for achievement of education for all at the national level and may lead to lower human capital and greater poverty among orphans when they reach adulthood.

Figure 13. Relation between enrollment rates and HIV prevalence,

countries surveyed since 1995

Percent enrolled (%) Latest survey available100908070605040302010

Dominican Republic C.A.R.Haiti Madagascar Mozambique South Africa Zimbabwe Uganda Nigeria Guatemala Cameroon Cambodia Nicaragua Tanzania Zambia Guinea Ghana Kenya Benin Brazil Mali Chad Niger Togo 30

Adult HIV prevalence (%) 19992520151050

Enrollment 7-14Adult HIV prevalence (%) 1999

all aspects of the quality of education and ensuring excellence of all so that recognized and measurable learning outcomes are achieved by all, especially in literacy, numeracy and essential life skills (UNESCO 2002). The Millennium Development Goals set precise targets for completion and gender equity: (1) ensure that, by 2015, children everywhere, boys and girls alike, will be able to complete a full course of primary schooling; and (2) that girls and boys will have equal access to all levels of education (United Nations 2002). 10. Even if not under-enrolled, orphans could be disadvantaged in terms of hours of attendance and

ultimately achievement and learning outcomes because of lower investments in complementary inputs (health care, text books), greater demand for their time in economic activities, lack of parental attention, and psychological stress.

Are orphans of primary school age (7-14) less likely to be enrolled in school than

children with living parents? Population-weighted enrollment rates for children by orphan status for all 39 datasets and 28 countries are presented in Appendix 4. Tests of statistical significance of the enrollment rate of each category of orphan compared with children with two living

parents are presented. These tests are useful, but it is often the case that the sample size was very small for two-parent orphans resulting in a lack of significance for what appears to be large differentials or that two rates are highly statistically significant from a large sample size but the size of the differential is small.

The results show substantial heterogeneity in terms of enrollment differentials among

orphans and non-orphans in the 28 countries with very different overall levels of enrollment among children with living parents. For example, in both Chad (with overall enrollment rates of less than 40 percent ) and South Africa (with overall enrollment rates greater than 90 percent) we see no significant difference in enrollment between orphans and children with living parents (Figure 14, panel A). In contrast, in both Benin and Kenya single- and two-parent orphans all have lower enrollment rates than children with living parents (Figure 14, panel B). The overall enrollment rate for children with living parents in Kenya is nearly twice that of Benin. In

Burkina Faso and Haiti, maternal orphans and two-parent orphans are disadvantaged in terms of enrollment, while in Tanzania and Nigeria orphans have higher enrollment than children with living parents (Figure 14, panels C and D). The situation in all 28 countries is summarized in Table 2 according to the overall 7-14 enrollment rate.

Figure 14. Enrollment differentials by orphan status, ages 7-14

A. No significant enrollment differentials by orphan status100Percent of children enrolled806040200Chad 1996100806040200South Africa 1998100806040200Benin 1993B. Lower enrollment for all orphans100806040200Kenya 1998C. Low enrollment for some orphans100Percent of children enrolled806040200Burkina Faso 1993100806040200Haiti 1993100806040200D. Countries with higher enrollment for orphans100806040200Tanzania 1996Nigeria 1999Both AlivePaternal Orphans Maternal OrphansTwo-parent Orphans 18

Table 2. Classification of countries by overall enrollment rates and difference in enrollment rates between orphans and non-orphans, most recent survey

Orphan enrollment relative to children with living parents

Mean enrollment rate for children 7-14

Low (<50%) Medium (50-80%) High (>80%)

Lower enrollment All orphans Benin 1996 Cambodia 1999 Brazil 1996

CAR 1994/5 Kenya 1998 Côte d’Ivoire 1994 Guatemala 1999 Madagascar 1997 Malawi 1992 Nicaragua 1997/8

Maternal and 2-parent orphans Burkina Faso Cameroon 1998 Zimbabwe 1999

1992/93 Haiti 1994/5

Maternal orphans only Guinea 1999 Dominican Republic

1996*

Paternal and 2-parent orphans Senegal 1992/93 Togo 1998 Ghana 1998 Paternal orphans only Uganda 1999/00 Only 2-parent orphans Mozambique 1997

Zambia 1998

Equally likely to be enrolled Chad 1996/97 South Africa 1998

Mali 1995/96 Niger 1998

Higher enrollment Nigeria 1999

Tanzania 1996

* Enrollment rates could not be computed for two-parent orphans because there were fewer than 20 children.

One possible explanation for these observed differentials is the correlation between

poverty and orphan status. Of the 28 countries, 25 have large differences in enrollment rates between children from the poorest and wealthiest families (see Appendix 4). Orphan enrollment may be lower in some cases because orphans are more likely to be poor. If we control for the effects of poverty, do differences in enrollment by orphan status persist? In Figure 15 we show the enrollment rate by orphan status for the lowest 40 percent and highest 20 percent of the

wealth distribution in Zambia. Within the poorest and richest households, orphans are less likely to attend school but particularly among the poor. Reasons for this “orphan effect” may include a greater demand placed on children’s time at home; grief that prevents a child from attending school; or other factors. However, the greatest differentials in school enrollments are between the poor and the non-poor, including orphans in these groups. Many of the reasons that poor orphans are not in school are the same as those that prevent other poor children from attending.

Figure 15. Enrollment rate by orphans status in lowest and highest quintiles,

Zambia 1998

1009080Percent enrolled706050403020100Both alivePaternal orphanMaternal orphanTwo-parent orphanLowest 40%Highest 20%Source: Authors' calculations, 1998 Zambia Living Conditions Monitoring SurveyThe large differentials between poor and non-poor enrollments in many countries

suggest that policies to raise enrollment among the poor will have a large impact on the most disadvantaged orphans. This can be seen most clearly by the case of Uganda, where we have surveys from both 1995 and 2000 (Figure 16). In 1995, there was a roughly 20 percent

differential between the poor and the non-poor in enrollment. In 1997, the government launched a large scale “universal enrollment” program that included the abolishing of fees for primary school that resulted in a surge in enrollments, particularly among the poor. By 2000, enrollment among the poor—including orphans—had increased by roughly 20 percentage points, reducing this gap (this result is explored in Deininger, Crommelynck and Kempaka 2001). There has been a similar large increase in enrollment of the poor in the Dominican Republic, which could be due to specific school policies or simply to growth in incomes among the poor (see Appendix 5). In Tanzania, enrollment of two-parent orphans has risen among the poor to the same low level as other poor children, eliminating orphan differentials. However, the large gap between all poor and non-poor children persists.

Figure 16. Changes in enrollment rate by orphan status and household wealth,

Uganda 1995-2000 Both alive100908070Enrollment rate (%)6050403020100Lowest 40%1995Highest 20%Lowest 40%2000Highest 20%Paternal orphanMaternal orphanTwo-parent orphanSource: Authors' calculations, 1995 Uganda DHS, 2000 Uganda National Household In contrast, in countries like Kenya enrollment differentials according to household

wealth are small (Figure 17). Yet within the poorest and richest households, enrollment does differ according to orphan status. Reducing poor-non-poor disparities in enrollment in Kenya is unlikely to raise orphan enrollment by much. This finding suggests that addressing issues related to specific problems faced by orphans in schools may help to further reduce enrollment disparities.

Figure 17. Changes in enrollment rate by orphan status and household wealth,

Kenya 1993-98

Both alive1009080Enrollment rate (%)706050403020100Lowest 40%Paternal orphanMaternal orphanTwo-parent orphan***1993Highest 20%Lowest 40%1998Highest 20%Source: Authors' calculations, Demographic and Health Surveys.* <20 two-parent orphans in this wealth category. 21

Finally, in seven countries enrollment data for orphans and non-orphans is available for

children aged 15-17—four in Africa (Cameroon, South Africa, Uganda, and Zambia), two in Central America/Caribbean (Dominican Republic and Nicaragua) and Cambodia (Appendix 6). Enrollment rates for these age groups are generally lower than for children 7-14, but still demonstrate diversity in terms of enrollment differentials for orphans and non-orphans. All orphans are significantly less likely to be enrolled in Cameroon (1998), certain categories of orphan are under-enrolled in the Dominican Republic (1997), Nicaragua (1996), and Cambodia (1999), and there are no significant differences between the enrollment of orphans and non-orphans in South Africa (1998), Uganda (2000), and Zambia (1998). It appears that the orphan enrollment inequalities among 15-17 year olds in Cameroon can be largely explained by large gaps in enrollment between the poor and the non-poor, while the lack of orphan enrollment inequities in Uganda also reflects similar enrollment rates among the poor and the non-poor. Nicaragua, in contrast, has both high differentials among the poor and non-poor and, within each welfare group, lower enrollment among orphans than non-orphans.

Is the gender gap in enrollment larger for orphans?

There is a frequently voiced concern that the schooling of girls who are orphaned may

suffer more than the schooling of boys who are orphaned, exacerbating existing inequalities in male-female enrollment rates (Subbarao, Mattimore, and Plangemann 2001, World Bank 2002a). There are a variety of reasons why the school enrollment of orphaned girls might be more affected than that of boys, including increased responsibilities in caring for siblings and higher demand for their time in household chores following the loss of an adult (if females are specialized in these tasks).

Before considering the gap among orphans, it is important to note that in many countries

there are significant gaps in enrollment between boys and girls overall, including among

children with living parents. Figure 18 shows a scatter-plot of the enrollment of girls against the enrollment of boys, regardless of orphan status. Children 7-14 are plotted as circles and children aged 15-17 are plotted as squares. Symbols that are solid indicate that the difference in male and female enrollment is statistically significantly different at the 10 percent level. A 45-degree line is drawn to indicate where male and female enrollment rates are the same; above the line girls have higher enrollment and below the line boys have higher enrollment. In countries where boys’ enrollment is relatively high (over 75 percent), girls’ enrollment is typically high as well and the differences that are statistically significant are small in magnitude. Togo is the

exception, with boys’ enrollment at 81 percent and girls’ at 66 percent. Among countries with boys’ enrollment rates between 50 and 75 percent, girls have substantially lower enrollment among 15 to 19 year olds but typically no lower enrollment among 7 to 14 year olds. An

exception is the Central African Republic (CAR), where boys’ enrollment is 70 percent among those 7 to 14 compared to 52 percent among girls. Last, among countries with boys’ enrollment below 50 percent there appears to be a consistent shortfall of about 9 percentage points among girls, and an even greater gap in some cases (e.g., 17 percentage points in Chad).

Figure 18. The gender gap in enrollment, all children Male enrollment equalTo female enrollment(45oline)10080Female enrollmentTogoCARZambia 1998Cote d’IvoireCambodiaCameroon 1998Benin60402000Age 7-14Chad2040Age 15-196080100Male enrollmentSolid symbol indicates that the male-female gap significant at 10 percent level Is the gender gap in enrollment—usually a disadvantage for girls—greater for orphans

than for non-orphans? Analysis of the data from these 28 countries shows that the answer to this question is not generalizeable (Appendix 7). There are four different categories of countries (Figure 19). First are countries like Chad and Senegal, where girls have lower enrollment and the gender gap between boys and girls is worse among orphans than among non-orphans (Panel A). Second is the more typical case, in which the gender gap in enrollment —be it at a low level (e.g., Kenya) or at a high level (e.g., Guinea)—is similar for orphans and non-orphans (Panel B). Twenty-one of the 28 countries had similar gender gaps for orphans and non-orphans among children 7-14 and all seven for which there were data for children 15-17 had similar gender gaps for orphans and non-orphans. A third category of countries has a smaller gender gap in

enrollment among orphans than non-orphans (e.g. Burkina Faso and Nigeria, Panel C). A fourth category includes several countries where female orphans have higher enrollment than male orphans, while among non-orphans this is not the case (e.g., Tanzania and Nicaragua, Panel D).

Figure 19. The gender gap in enrollment among orphans and non-orphans, selected countries (ages 7-14)A. Female disadvantage in enrollment is larger among orphans than non-orphansChad 1998Senegal 1992-93100100808060604040202000MaleFemaleMaleFemaleMaleFemaleMaleFemalenon-non-orphanorphannon-non-orphanorphanorphanorphanorphanorphanC. Other scenarios –e.g. the male-female difference in enrollment is smaller among among orphans than non-orphansBurkina Faso 1992-93Nigeria 1999100100808060604040202000MaleFemaleMaleFemaleMaleFemaleMaleFemalenon-non-orphanorphannon-non-orphanorphanorphanorphanorphanorphan24

B. Male-female difference in enrollment is similar among orphans and non-orphansKenya 1998Guinea 1999100100808060604040202000MaleFemaleMaleFemaleMaleFemaleMaleFemalenon-non-orphanorphannon-non-orphanorphanorphanorphanorphanorphanD. Other scenarios -a female “advantage”among non-orphans which decreases or increases among orphansTanzania 1996Nicaragua 1998100100808060604040202000MaleFemaleMaleFemaleMaleFemaleMaleFemalenon-non-orphanorphannon-non-orphanorphanorphanorphanorphanorphan Figure 20 plots of the gender difference in enrollment among orphans (maternal,

paternal, and both parent) on the Y-axis against the gender difference in enrollment among non-orphans on the X-axis. Differences that are statistically significant from zero are again shown using a solid symbol. Most countries correspond to the second category described above where girls are disadvantaged but the gender differential in enrollment among orphans mirrors that among non-orphans. There are only three countries in which female orphans have a disadvantage in enrollment that is greater for orphans than among non-orphans and in which this gap is significantly different from zero: Chad and Senegal for children aged 7 to 14, and Uganda for children aged 15 to 19.11 In Burkina Faso (for 7-14 year olds) and Zambia (for 15-19 year olds) the gender gap among is significantly smaller among orphans than among non-orphans, and in three other countries a female disadvantage in enrollment among non-orphans becomes a female advantage among orphans (Nigeria and Malawi among 7 to 14 year olds, and Dominican Republic among 15 year olds). Last, in Tanzania a female advantage in enrollment among non-orphans becomes a disadvantage among orphans and in Nicaragua a female advantage is larger among orphans than non-orphans.12

Figure 20. Gender differences in enrollment, orphans and non-orphans compared

Male-Female difference equalamong orphans and non-orphans(45oline)Uganda 1999SenegalChad25Tanzania 1996Male-Female gap in enrollment among orphans155-5-15NicaraguaZambia 1998Burkina FasoNigeria 1999Dominican Rep. 1997Malawi-25-25-15-551525Male-Female gap in enrollment among children with both parents aliveAge 7-14Age 15-19Solid symbol indicates that the male-female gap among orphans is significantly different from the male-female gap among non orphans at the 10 percent level 11. The difference in gender gap between orphans and non-orphans is also statistically significant in Cameroon, although the magnitude of the difference is extremely small.

12. In Nicaragua a female advantage among non-orphans is significantly reduced, although the magnitudes are miniscule.

While the results so far suggest that there is very little consistency across countries

with respect to the relationship between orphan status and the gender gap in enrollment, it is possible that the differential would only manifest itself among poorer households. This would be the case if girls from poor households were especially likely to need to take care of their orphaned siblings, for example. Figure 21 plots the gender gap in enrollment between orphans and non-orphans according to whether the child is from a household in the poorest 40 percent, or the richest 20 percent of the sample. The results for the poorest 40 percent are similar to the overall sample. Chad and Senegal have a female disadvantage among the poor that is

significantly larger for poor orphans, and Nicaragua has a female advantage among poor non-orphans that is larger among poor orphans. All the other the differences that were significant in the sample as a whole no longer are when focused on the poorest. Conversely, in Cambodia there was not a significant difference in the gender gap between orphans and non-orphans in the overall sample but there is a female disadvantage among non-orphans that is significantly (and substantially) larger among poor orphans. Interestingly, there are several countries where a female disadvantage among non-orphans is statistically significantly larger among orphans among children from the richest 20 percent of households: Benin, Côte d’Ivoire, Mali, Ghana, and Cameroon.

Figure 21. Gender gap in enrollment for orphans and non-orphans

in the poorest and richest households

Poorest 40%Richest 20%ChadMale-Female gap in enrollment among orphansCambodiaSenegal453525155453525155-5NicaraguaTanzania 1996GhanaMaliCote d’IvoireBenin-5Niger-15-15-15-5515253545Male-Female gap in enrollment among children with both parents alive-15-5515253545Male-Female gap in enrollment among children with both parents aliveAge 7-14Age 15-19Solid symbol indicates that the male-female gap among orphans is significantly different from the male-female gap among non orphans at the 10 percent level 26

4. Conclusions

These diverse findings demonstrate that the extent to which orphans are under-enrolled

relative to other children is country-specific, at least in part because the correlation between orphan status and poverty is not consistent across countries. Indeed, it cannot be assumed that enrollment differentials exist between orphans and non-orphans or, when they exist, why. On the other hand, all but a handful of the countries studied have sharp differentials in enrollment between children in poor and non-poor households and several have very low enrollments for both poor and non-poor children. Social protection and schooling policies need to take a close look at the specific situation in a country before considering mitigation measures.

• In countries like Benin, Burkina Faso, Guinea, and Senegal, the extent of under-enrollment of orphans is dwarfed by the enormous shortfall in overall enrollment evident among poor and non-poor children alike. This suggests that the key to raising enrollment among orphans is to pursue sectoral and economic policies to raise enrollment among all children, including orphans. • In the group of countries with moderate overall enrollment rates there are often very

large gaps between enrollment of poor and non-poor children. The most disadvantaged children are the poor, including poor orphans. Policies to reduce the gap in enrollment between poor and non-poor will contribute significantly to raising enrollment among the neediest orphans without any orphan-specific targeting. As was shown, in the Dominican Republic, Kenya, and Uganda, improvements in enrollments among the poor through rises in income or specific policies to improve the access of the poor have substantially raised the enrollment of orphans. • In countries like Brazil, Dominican Republic, and Zimbabwe where overall enrollment

rates are high even among the poor, lower enrollment of orphans is likely related to problems specific to being an orphan, some of which may not be school-related. The reasons for persistent enrollment gaps need to be carefully explored?policies that subsidize fees or school uniforms may not be effective in reducing this gap while potentially transferring funds to orphans who might otherwise already be enrolled. The diversity of conditions dictates mitigation measures that are tailored to the needs of specific countries; policymakers need to resist the temptation to advocate a single ‘best practice’ model for all countries regardless of the extent or source of orphan enrollment differentials.

A more general conclusion from this study is that orphan status in most countries (there

are some exceptions) is not good targeting criterion for “traditional” programs aimed at raising enrollment rates?like subsidies for school fees, text books, and uniforms. Orphans are not universally in need of assistance. Further, opportunistic redistribution of orphans is likely to occur when the benefits being channeled to orphans are things that other children or other household members lack?like textbooks, uniforms, school fees, free medical care, or

supplemental feeding. Indeed, in much of Africa there is a strong tradition of redistributing children across households through child fostering (Ainsworth 1996). A concentration of orphans in some households could result from orphan targeting that may or may not result in

their improved welfare. On the other hand, interventions linked solely to the special needs of orphans (for example, grief counseling or health services for HIV-infected children) are

unlikely to create incentives for opportunistic responses by households, as the benefits are not easily shared by other household members. Policies and programs aimed at improving the

welfare of the poorest households will help the poorest children, including the poorest orphans, without creating incentives to redistribute children in ways that may adversely affect their welfare.

This analysis has focused on enrollments, which is a necessary but not sufficient

condition for learning. The objective of “Education for All” is learning. We have not been able to explore delayed enrollment, completion rates, and the determinants of learning outcomes for orphans, the poor, and poor orphans—a high priority for research. Equally if not more

important is greater research on the reasons why differences in enrollment among orphans and non-orphans persist and pilot field tests of alternative mitigation measures. In fact, child

schooling may be affected before a parent dies, during the time when there is a sick adult who must be cared for and for whom many resources may be spent for medical treatment. By focusing exclusively on orphans—after a parental death—researchers may be neglecting the largest impacts, and those that may be amenable through short-term support for households with terminally ill adults.13 Thus, the impacts on child schooling before parents and other adults die of AIDS are also a high priority for research.

Finally, while we have focused on the impact of orphan status on enrollment, we

shouldn’t lose sight of the fact that Education for All is a major policy to reduce the spread of HIV/AIDS. There is a well-established positive correlation between educational attainment and safer sexual behavior, which will translate into lower rates of new infection. Further, schools are an important point for providing information on HIV prevention. In many of the hardest-hit countries, young adults still have shockingly low levels of knowledge of how HIV is

transmitted. In many of the countries studied, policies to raise enrollments among the poor will both raise enrollment among orphans and ensure that more poor children are given the tools to prevent HIV as they transition to adulthood.

13. Gilborn and others (2001) found that enrollment of two-parent orphans and of children of people living with HIV/AIDS exceeded 90 percent in Uganda, but that older children (13-17) in households with a sick parent had lower school attendance (80 percent) than orphans (89 percent). Roughly one fourth of the children of people living with HIV/AIDS reported a decline in attendance and performance because of their parents’ illness. Older two-parent orphans reported that their attendance improved after moving in with a guardian following the parent’s death.

References

The word “processed” describes informally reproduced works that may not be commonly available through library systems.

Ainsworth, Martha. 1996. “Economic aspects of child fostering in Côte d’Ivoire.” In T. Paul Schultz, ed.,

Research in Population Economics 8. Greenwich, CT: JAI Press.

Ainsworth, Martha, Susmita Ghosh, and Innocent Semali. 1995. “The impact of adult deaths on household

composition in Kagera Region, Tanzania.”. Development Research Group, World Bank, August. Processed.

Ainsworth, Martha, Kathleen Beegle, and Godlike Koda. 2002. “The impact of adult mortality on primary

school enrollment in Northwestern Tanzania.” Africa Region Human Development Working Paper Series. Washington, D.C.: World Bank, Africa Region.

Bicego, George, Shea Rutstein and Kiersten Johnson. 2002. “Dimensions of the Emerging Orphan Crisis in

Sub-Saharan Africa.” Calverton, MD: Macro International.

Deininger, Klaus, Anja Crommelynck and Gloria Kempaka. 2001. “Long-term welfare and investment

impact of AIDS-related changes in family composition: Evidence from Uganda.” Development Research Group, World Bank. Processed.

Filmer, Deon. 2000. “The Structure of Social Disparities in Education: Gender and Wealth.” World Bank

Policy Research Working Paper No. 2268. Development Research Group, World Bank. Washington, D.C.

Filmer, Deon, and Lant Pritchett. 1999. “The effect of household wealth on educational attainment: Evidence

from 35 countries.” Population and Development Review 25(1, March): 85-120.

Filmer, Deon, and Lant Pritchett. 2001. “Estimating wealth effects without expenditure data – or tears: An

application to educational enrollments in states of India.” Demography 38(1): 115-132.

Foster, Geoff, and John Williamson. 2000. “A review of current literature on the impact of HIV/AIDS on

children in sub-Saharan Africa.” AIDS 14 (suppl 3): S275-S284.

Gilborn, Laelia Z., Rebecca Nyonyintono, Robert Kabumbuli, and Gabriel Jagwe-Wadda. 2001. “Making a

difference for children affected by AIDS: Baseline findings from operational research in Uganda.” Horizons Program/Population Council and Makerere University. Processed.

Kitonsa, E.N.K., L. Antivelink, C.A. Kajura, P. Kaleebu, and J.A. Opolot. 2000. “The needs and coping

mechanisms of children orphaned by AIDS in semi-urban south Uganda: Implications for policy makers.” Presented at the XIIIth International Conference on AIDS, Durban, South Africa, July.

Lloyd, C., and A.K. Blanc. 1995. “Children’s schooling in Sub-Saharan Africa: The roles of fathers, mothers

and others.” The Population Council Working Papers, no. 78. New York: The Population Council. Nyambedha, Erick Otieno, Simiyu Wandibba, and Jens Aagaard-Hansen. 2001. “Policy implications of the

inadequate support systems for orphans in Western Kenya.” Health Policy 58(1): 83-96.

Subbarao, K., Angel Mattimore, and Kathrin Plangemann. 2001. “Social protection of Africa’s orphans and

other vulnerable children: Issues and good practice program options.” Africa Region Human Development Working Paper Series. Africa Region, World Bank, Washington, D.C. UNESCO. 2002. http://www.unesco.org/education/efa/ed_for_all/faq.shtml as of July 2002. United Nations. 2002. Millennium Report of the Secretary-General of the United Nations.

http://www.un.org/millennium/sg/report/ as of July 2002.

U.S. Agency for International Development. 2000. Children on the Brink. Washington, D.C.

World Bank. 1999. Confronting AIDS: Public Priorities in a Global Epidemic. Revised edition. Washington,

D.C.: Oxford University Press.

World Bank. 2000. World Development Indicators 2000. Washington, D.C.

World Bank 2002a. Education and AIDS: A window of hope. Report 24059. Human Development Network,

World Bank, Washington, D.C.

World Bank 2002b. “Achieving Education For All: Simulation Results for 47 Low-Income Countries”

Human Development Network, World Bank, Washington, D.C. Processed.

Appendix 1. Data sets and sample sizes

Number of Number of Number of

Number of paternal maternal 2-parent

Country Survey Year children 7-14 orphans orphans orphans Benin 6,455 393 226 36 DHS 1996 Brazil 10,601 550 129 47 DHS 1996 Burkina Faso 7,933 537 267 139 DHS 1992/3 Cambodia 7,463 399 87 69 SES 1999 Cameroon 4,391 293 118 32 DHS 1991 Cameroon 5,835 513 189 58 DHS 1998 Central African Rep. 5,996 576 277 90 DHS 1994/5 Chad 8,459 639 237 86 DHS 1996/7 Côte d’Ivoire 8,497 512 209 57 DHS 1994 Dominican Republic 6,684 221 135 17 DHS 1991 Dominican Republic 7,504 294 162 16 DHS 1996 Ghana 5,156 292 135 76 DHS 1993 Ghana 5,131 277 149 37 DHS 1998 Guatemala 6,760 360 169 23 DHS 1999 Guinea 8,202 564 246 112 DHS 1999 Haiti 5,242 461 252 115 DHS 1994/5 Kenya 9,705 649 200 43 DHS 1993 Kenya 9,159 814 219 119 DHS 1998 Madagascar 7,127 525 295 55 DHS 1997 Malawi 5,924 626 311 75 DHS 1992 Mali 11,298 362 250 75 DHS 1995/6 Mozambique 10,257 1054 665 165 DHS 1997 Nicaragua 14,276 690 177 36 DHS 1997/8 Niger 8,194 460 259 36 DHS 1998 Nigeria 8,136 360 225 94 DHS 1999 Senegal 7,103 407 194 33 DHS 1992/3 South Africa 24,559 2,861 383 402 OHS 1995 South Africa 15,927 1,667 299 174 OHS 1998 Tanzania 10,189 695 306 67 DHS 1991/2 Tanzania 8,660 671 305 80 DHS 1996 Togo 11,176 989 402 104 DHS 1998 Uganda 8,131 967 405 287 DHS 1995 Uganda 15,359 1,765 675 781 UNHS 1999/0 Zambia 7,773 563 252 87 DHS 1992 Zambia 8,881 901 384 217 DHS 1996/7 Zambia 13,248 1,355 488 329 LCMS 1996 Zambia 20,830 2,194 687 748 LCMS 1998 Zimbabwe 7,345 624 198 80 DHS 1994 Zimbabwe 6,783 841 242 201 DHS 1999

Source: DHS: Demographic and Health Survey; LCMS: Living Conditions Measurement Survey; OHS:

October Household Survey; SES: Socio-Economic Survey; UNHS: Uganda National Household Survey.

Appendix 2A. Orphan rates, ages 7-14

Country

Year

Paternal orphans

Maternal orphans 3.41 1.23 3.52 1.10 2.84 3.58 4.60 2.93 2.47 1.67 2.09 2.63 2.84 2.44 3.02 4.91 1.99 2.45 4.37 4.23 2.64 6.74 1.19 3.20 2.74 2.95 2.72 1.63 1.80 2.91 3.68 3.42 4.89 4.06 3.25 4.34 3.60 3.44 2.63 3.67

Two-parent orphans

0.54 0.42 1.62 0.89 0.75 0.99 1.53 0.87 0.68 0.27 0.19 1.48 0.70 0.35 1.32 2.06 0.38 1.26 0.79 1.58 0.67 1.78 0.26 0.40 1.16 1.24 0.47 1.64 0.97 0.81 1.01 0.99 3.26 4.54 1.07 2.57 2.42 3.54 1.10 3.11

Missing 1.17 2.43 0.63 2.15 1.49 2.16 1.35 1.61 1.63 1.88 0.24 1.25 1.20 2.42 2.30 1.64 3.13 2.91 2.59 1.24 0.86 3.10 0.68 1.99 7.00 - 2.71 n/a 3.98 3.60 2.22 0.94 2.32 0.22 1.24 1.99 2.33 1.03 2.61 4.37

Benin DHS 1993 6.15 Brazil DHS 1996 5.10 Burkina Faso DHS 1993 6.37 Cambodia SES 1999 5.18 Cameroon DHS 1991 6.66 Cameroon DHS 1998 8.87 Central African Rep. DHS 1994 9.62 Chad DHS 1996 7.25 Côte d’Ivoire DHS 1994 5.88 Dominican Republic DHS 1991 3.54 Dominican Republic DHS 1996 3.73 Ghana DHS 1993 5.65 Ghana DHS 1998 5.10 Guatemala DHS 1999 5.02 Guinea DHS 1999 6.88 Haiti DHS 1993 8.56 Kenya DHS 1993 6.60 Kenya DHS 1998 8.77 Madagascar DHS 1997 7.60 Malawi DHS 1992 6.07 Mali DHS 1995 5.15 Mozambique DHS 1997 9.66 Nicaragua DHS 1997 4.75 Niger DHS 1998 5.25 Nigeria DHS 1999 4.31 Nigeria DHS^ 1999 4.63 Senegal DHS 1993 5.71 South Africa OHS 1995 12.48 South Africa OHS 1998 10.61 Tanzania DHS 1991 6.66 Tanzania DHS 1996 8.04 Togo DHS 1998 8.87 Uganda DHS 1995 11.87 Uganda NHS 1999/00 11.10 Zambia DHS 1992 7.17 Zambia DHS 1996 10.58 Zambia LCMS 1996 10.41 Zambia LCMS 1998 10.75 Zimbabwe DHS 1994 8.75 Zimbabwe DHS 1999 12.59 ^ Percentages omitting missing orphan status category.

Appendix 2B. Orphan rates, ages 15-17

Country

Cambodia SES Cameroon DHS

Dominican Republic DHS Nicaragua DHS South Africa OHS South Africa OHS Uganda NHS Zambia LCMS Zambia LCMS

Year 1999 1998 1996 1997 1995 1998 1999/00 1996 1998

Paternal orphans 8.34 13.20 5.27 7.58 15.48 14.71 15.18 14.03 14.53

Maternal orphans 1.75 4.74 3.02 2.18 2.08 2.41 5.13 4.39 4.80

Two-parent orphans 1.72 2.02 0.59 0.66 2.48 1.61 6.51 3.59 5.27

Missing 3.37 1.81 0.46 0.76 n/a 3.37 0.19 5.67 1.46

Appendix 3. Relationship to head among two-parent orphans, ages 7-14

Other Relation (including

Adopted/ spouse, in-law, foster niece, nephew, No

achild Country/data set/year Head Grandchild Sibling etc.) relation

Benin DHS 1993 0.0 11.6 13.8 6.2 52.2 16.3

Brazil DHS 1996 0.0 23.2 4.1 36.9 21.9 13.9 Burkina Faso DHS 1993 0.0 27.3 11.0 17.1 39.1 5.5 Cambodia SES 1999 0.0 37.0 14.0 27.4 21.1 0.6 Cameroon DHS 1991 0.0 10.4 22.8 .. 48.9 18.0 Cameroon DHS 1998 0.0 21.1 22.1 3.0 46.7 7.2 C.A.R. DHS 1994 0.0 16.5 19.3 3.0 57.1 4.1 Chad DHS 1996 0.8 13.9 9.1 18.2 57.5 0.7 Côte d’Ivoire DHS 1994 0.0 16.4 10.2 0.0 65.8 7.6 Dominican Rep. DHS 1991 0.0 23.1 12.9 15.1 35.4 13.5 Dominican Rep. DHS 1996 0.0 38.5 28.5 12.9 15.3 4.9 Ghana DHS 1993 0.0 37.7 7.8 6.5 44.2 3.9 Ghana DHS 1998 0.0 29.7 8.1 11.5 44.2 6.5 Guatemala DHS 1999 0.0 60.8 3.5 1.3 13.1 21.2 Guinea DHS 1999 0.8 13.9 18.2 31.8 30.7 4.6 Haiti DHS 1997 0.0 28.5 5.5 3.6 36.6 25.9 Kenya DHS 1993 0.0 37.8 13.2 9.5 35.6 3.9 Kenya DHS 1998 0.0 27.2 10.9 12.5 39.5 9.9 Madagascar DHS 1997 0.0 23.9 15.9 25.8 18.2 16.3 Malawi DHS 1997 0.0 54.8 9.7 14.0 11.1 10.4 Mali DHS 1996 0.0 10.2 13.9 36.3 31.1 8.5 Mozambique DHS 1997 0.0 15.2 20.8 5.8 57.3 0.9 Nicaragua DHS 1997 0.0 52.5 9.4 13.0 22.3 2.8 Niger DHS 1998 0.0 36.3 7.0 23.7 27.4 5.6 Nigeria DHS 1999 0.0 36.8 14.3 10.3 31.1 7.5 Senegal DHS 1993 0.0 6.1 3.0 12.1 66.7 12.1 South Africa, OHS 1995 0.0 46.3 10.2 21.9 17.5 4.1 Tanzania DHS 1991 0.0 38.9 13.0 4.4 41.3 2.4 Tanzania DHS 1996 0.0 35.4 13.0 0.6 46.0 5.0 Togo DHS 1998 0.0 30.5 11.2 14.6 34.3 9.4 Uganda DHS 1995 0.1 40.7 9.9 9.6 36.7 3.0 Zambia DHS 1992 0.0 27.5 17.4 2.2 51.3 1.7 Zambia DHS 1996 0.0 30.8 15.7 2.9 49.1 1.5

Zambia, LCMS 1996 0.0 38.1 10.1 8.4 42.6 0.8 Zimbabwe DHS 1994 0.0 46.0 10.3 8.7 35.1 0.0 Zimbabwe DHS 1999 0.4 50.1 13.2 6.0 29.9 0.5 Notes: a. This category may include children who are related biologically to the head, including grandchildren,

siblings, and other relatives. Depending on the country, the response may be adopted and/or fostered and/or stepchild.

b. Of which 11.3 percent are the niece or nephew of the head. c. Ages 7-11.

Appendix 4. Enrollment rates by orphan status and household wealth, ages 7-14

All children

Two- Both Paternal Maternal parent alive orphans orphans orphans Total 47.3 95.3 30.2 74.8 70.7 77.9 63.2 35.6 53.3 73.4 94.2 78.8 80.7 80.6 29.0 77.2 84.3 91.3 62.9 64.5 29.1 61.4 79.5

38.7** 92.6* 31.6 67.3* 76.5* 79.0 53.1** 36.7 44.9** 69.4 92.7 72.9* 68.9** 73.8* 28.0 77.7 83.5 87.2** 53.1** 53.4** 30.0 59.6 73.5**

37.9** 85.5** 22.3** 68.7 69.3 66.6** 55.2* 32.6 44.1** 58.5* 88.5+ 77.0 77.6 69.8* 19.4** 64.3** 77.9+ 84.2* 44.7** 50.8** 26.0 63.8 71.1*

20.1** 87.2 25.5 69.0 66.0 72.5 46.5** 33.8 38.8* # # 68.4+ 73.6 74.4 31.1 59.9** 68+ 72.8** 40.6** 39.0** 24.3 32.1** 73.4

46.0 94.7 29.9 74.1 71.2 77.5 61.1 35.5 52.3 72.6 94.0 78.2 79.8 79.7 28.3 76.0 83.8 90.4 60.8 62.6 29.0 60.1 79.1

Poorest 40 percent

Two- Both Paternal Maternal parent alive orphans orphans orphans Total 27.3 91.8 15.7 64.9 52.0 62.1 44.9 24.4 36.0 56.5 90.2 72.0 71.4 69.5 14.8 60.2 82.7 91.6 49.8 53.1 12.5 46.5 65.7

24.0 91.0 12.1 61.1 58.8 66.2 38.8 24.5 27.6* 54.7 84.5 69.7 64.4 67.0 12.7 55.1 82.5 87.5+ 44.1 42.2+ 10.2 56.5* 61.0

21.8 82.4+ 15.9 64.3 43.0 56.2 38.7 14.3+ 26.0+ 37.0* 82.4 66.7 71.6 57.6* 13.1 50.7 67.6* 91.8 35.1* 37.9* 12.4 52.3 56.0+

# 91.7 18.0 54.1 # 60.9 24.0* # # # # 60.0 # # 10.9 44.4 # 81.7 34.2 61.2 0.9** 25.8* 65.0

26.5 91.1 15.6 64.1 52.3 62.4 42.9 24.1 35.1 55.3 89.7 71.6 70.6 68.6 14.2 58.5 82.2 91.0 48.2 51.6 12.2 47.1 65.2

Richest 20 percent

Two- Both Paternal Maternal parent alive orphans orphans orphans Total 74.3 99.0 67.5 91.6 93.6 94.6 86.2 61.6 77.0 93.7 97.8 92.2 93.6 95.9 54.7 92.1 90.6 94.4 92.7 85.3 66.6 82.5 94.8

48.6** 69.8 # 72.4** 97.6 # # 98.7** 63.3 61.1 46.1** 66.4** 94.4 # # 91.1** 92.0 92.4 # 93.3** 91.0 94.4 # 94.3** 73.4* 83.1 83.2 84.7** 60.1 63.6 47.5 61.3** 58.0** 70.6 # 75.6** 90.9 # # 93.6** 99.5* # # 97.9** 80.4+ 91.7 # 91.5** 94.6 # # 93.1** # # # 95.5** 45.4 49.2 67.1 53.3** 92.4 72.2* 75.8* 90.5** 91.7 93.6 # 89.5** 90.1 82.5 # 93.3+ 83.5+ 81.0 # 91.5** 81.2 71.2* # 83.9** 75.1+ 72.2 47.0+ 66.6** 69.2* 88.4 65.0 80.9** 94.4 # # 94.5**

(Continued on the next page.)

Dataset

Benin DHS Brazil DHS

Burkina Faso DHS Cambodia SES Cameroon DHS Cameroon DHS C.A.R. DHS Chad DHS

Cote d’Ivoire DHS Dominican Rep. DHS Dominican Rep. DHS Ghana DHS Ghana DHS Guatemala DHS Guinea DHS Haití DHS Kenya DHS Kenya DHS

Madagascar DHS Malawi DHS Mali DHS

Mozambique DHS Nicaragua DHS

Year 1996 1996 1992/3 1999 1991 1998 1994/5 1996/7 1994 1991 1996 1993 1998 1999 1999 1994/5 1993 1998 1997 1992 1995/6 1997 1997/8

Appendix 4 (continued). Enrollment rates by orphan status and household wealth, ages 7-14

All children Poorest 40 percent Richest 20 percent

Two- Two- Two- Both Paternal Maternal parent Both Paternal Maternal parent Both Paternal Maternalparent Dataset Year alive orphans orphans orphans Total alive orphans orphans orphans Total alive orphans orphans orphans Total

Niger DHS 1998 26.3 23.6 22.2 22.1 25.7 13.7 9.1 10.2 # 13.1 61.4 53.4 46.2+ # 60.0** Nigeria DHS 1999 67.8 73.7* 71.3 66.5 67.6 41.4 49.5 52.9+ 53.7 42.0 93.7 93.1 82.3+ # 92.2** South Africa OHS 1995 97.0 96.9 93.5* 95.7 96.9 95.8 96.4 93.7 96.4 95.9 99.1 97.8 95.7 97.1 98.9** South Africa OHS 1998 93.3 92.8 95.3 90.6 93.2 92.1 92.4 96.4**88.0 92.2 95.0 97.5 # # 95.1** Senegal DHS 1992/3 35.9 31.2* 39.2 9.1**35.4 15.6 21.3+ 20.2 # 15.8 72.0 57.1+ 68.8 # 70.7** Tanzania DHS 1991/2 53.2 56.6 53.9 37.9* 53.2 47.6 50.4 43.0 21.9**47.7 65.6 74.5 76.4 # 65.6** Tanzania DHS 1996 53.7 59.9** 56.2 60.7 54.3 44.8 56.3**50.1 52.3 46.0 73.1 65.6 75.6 67.8 72.0** Togo DHS 1998 75.1 69.7** 76.9 59.6** 74.2 63.9 64.3 63.8 42.7* 63.5 87.8 76.8+ 96.2** 66.6* 86.7** Uganda DHS 1995 74.9 66.7** 71.0 74.7 73.6 65.5 57.1* 64.0 70.6 64.4 88.2 80.3* 79.1* 86.3 86.2** Uganda UNHS 1999/0 90.4 87.9+ 92.5 88.4 90.1 84.2 77.6* 89.2 88.8 83.8 95.1 93.6 96.9 86.3+ 94.3** Zambia DHS 1992 77.8 72.0** 68.5** 77.0 76.9 61.3 58.8 57.3 69.5 60.7 95.7 93.7 91.1 # 95.3** Zambia DHS 1996/7 68.6 62.0** 66.9 64.4 67.6 56.1 52.8 55.2 56.5 55.5 92.6 90.6 91.2 79.7* 91.9** Zambia LCMS 1996 71.1 70.2 65.0+ 71.8 70.6 56.7 60.1 57.6 38.8* 56.9 92.9 90.5 83.8* 87.0+ 92.0** Zambia LCMS 1998 68.7 69.2 65.9 58.7** 68.3 56.7 58.2 58.7 41.7**56.4 91.9 89.2 82.4** 84.0* 91.0** Zimbabwe DHS 1994 91.0 89.4 85.3* 94.4 90.6 88.7 84.8+ 87.5 91.3 88.1 96.5 97.7 # # 96.2** Zimbabwe DHS 1999 90.0 88.4 85.5+ 80.0** 89.1 88.6 85.7 80.1* 81.7+ 87.5 96.6 99.3* 94.9 77.5+ 96.1** # indicates a cell size of fewer than 20 observations. All significance tests are carried out relative to the “Both alive” category within the wealth level, except for the “Total” column of the “richest 20 percent” level which is relative to the “Total” column for the “poorest 40 percent”. + indicates significance at the 10 percent level, * indicates significance at the 5 percent level, and ** indicates significance at the 1 percent level.

Appendix 5. Changes in enrollment over time, by orphan status and household welfare

Changes in enrollment rates by orphan status and household wealth,

Cameroon 1991-98

and wealth, Cameroon 1991-98Paternal orphanMaternal orphanBoth alive1009080Enrollment rate (%)706050403020100Lowest 40%Two-parent orphan****Source: Authors' calculations, Demographic and Health Surveys* < 20 two-parent orphans in this wealth category1991Highest 20%Lowest 40%1998Highest 20%Changes in enrollment rate by orphan status and household wealth, Dominican Republic 1991-96

Both alive1009080Enrollment rate (%)706050403020100Lowest 40%Paternal orphanMaternal orphan**1991Highest 20%Lowest 40%1996Highest 20%Source: Authors' calculations, Demographic and Health Surveys* < 20 children were maternal orphans in the highest 20% or 2-parent orphans in all wealth categories. 36

Changes in enrollment rates by orphan status and household wealth,

Tanzania 1991-96

and wealth, Tanzania 1991-96Both alive1009080Enrollment rate (%)706050403020100Lowest 40%Paternal orphanMaternal orphanTwo-parent orphan*1991Highest 20%Lowest 40%1996Highest 20%Source: Authors' calculations, Demographic and Health Surveys* <20 two-parent orphans in the highest 20%

Changes in enrollment rate by orphan status and household wealth,

Zimbabwe 1994-99

Both alive100908070Enrollment rate (%)Paternal orphanMaternal orphanTwo-parent orphan6050403020100Lowest 40%*1994Highest 20%Lowest 40%1999Highest 20%Source: Authors' calcluations, Demographic and Health Surveys* < 20 children were maternal or two-parent orphans in this wealth category 37

Changes in enrollment rate by orphan status and household wealth,

Zambia 1992-98 and wealth, Zambia 1992-98Both alive1009080Enrollment rate (%)706050403020100Lowest 40%Source: Authors' calculations, 1992 Zambia DHS and 1998 Zambia Living Conditions Monitoring Survey* <20 two-parent orphans in the highest quintilePaternal orphanMaternal orphanTwo-parent orphan*1992Highest 20%Lowest 40%1998Highest 20%

Appendix 6. Enrollment rates by orphan status and household wealth, ages 15-17

All children Poorest quintile Richest quintile Paternal Maternal Two-parent Paternal Maternal Two-parent Paternal Maternal Two-parent

Country Both alive orphans orphans orphans Both alive orphans orphans orphans Both alive orphans orphans orphans Cameroon DHS 1998 54.9 46.5* 40.8* 26.3** 33.1 29.5 # # 75.5 70.7 # # Dominican Rep. DHS 1997 75.2 62.9** 64.2+ 43.4+ 60.9 32.3** 35.3* # 81.8 # # # Nicaragua DHS 1996 53.1 43.6** 31.0** 29.7* 18.8 18.1 # # 82.1 60.6** # # Cambodia SES 1999 57.4 43.5** 55.4 43.9 42.8 30.4+ # # 70.9 74.2 # # Zambia LCMS 1996 60.8 53.9* 58.4 49.3* 40.4 38.4 65.9* 25.3 81.5 77.2 75.2 69.8+ Zambia LCMS 1998 56.2 53.3 54.1 52.8 45.5 43.3 34.0 34.1 80.4 79.9 70.9 72.1 South Africa OHS 1995 92.7 89.3** 86.9* 83.1** 89.9 86.0+ 87.4 74.9* 96.2 92.7 # # South Africa OHS 1998 89.5 86.1* 85.7 88.5 79.9** 77.2+ 78.9 92.4 85.2 # # Uganda UNHS 1999/00 74.1 64.8** 71.6 61.8* 61.1 58.8 52.3 66.7 79.0 70.7 81.1 66.9+ # indicates a cell size of fewer than 20 observations. All significance tests are whether the enrollment for females is different from males, within the orphan status group. + indicates significance at the 10 percent level, * indicates significance at the 5 percent level, and ** indicates significance at the 1 percent level.

Appendix 7. School enrollment by orphan status and gender, ages 7-14

Data set Benin DHS Brazil DHS

Burkina Faso DHS Cambodia SES Cameroon DHS Cameroon DHS C.A.R. DHS Chad DHS

Cote d’Ivoire DHS Dominican Rep. DHS Dominican Rep. DHS Ghana DHS Ghana DHS Guatemala DHS Guinea DHS Haiti DHS Kenya DHS Kenya DHS

Madagascar DHS Malawi DHS Mali DHS

Mozambique DHS Nicaragua DHS

Year 1996 1996 1992/3 1999 1991 1998 1994/5 1996/7 1994 1991 1996 1993 1998 1999 1999 1994/5 1993 1998 1997 1992 1995/6 1997 1997/8

Male 58.3 95.3 35.9 75.8 74.5 80.0 71.7 43.7 61.4 69.7 93.8 81.4 80.7 83.5 33.6 77.3 84.7 91.9 62.0 66.6 33.6 65.7 77.5

Both alive Female 35.6** 95.3 24.5** 73.8+ 67.0** 75.7* 54.1** 27.5** 45.3** 77.1** 94.7 76.0** 80.7 77.6** 24.4** 77.1 83.8 90.7 63.9 62.6* 24.8** 57.1** 81.5**

Total 47.3 95.3 30.2 74.8 70.7 77.9 63.2 35.6 53.3 73.4 94.2 78.8 80.7 80.6 29.0 77.2 84.3 91.3 62.9 64.5 29.1 61.4 79.5

Paternal orphans Male Female Total 49.6 28.5** 38.7 92.1 93.0 92.6 33.8 29.3 31.6 71.1 63.0 67.3 77.6 75.2 76.5 84.0 73.5* 79.0 62.6 43.0** 53.1 47.6 25.8** 36.7 53.8 36.9** 44.9 66.1 72.8 69.4 91.2 94.3 92.7 78.0 67.9+ 72.9 67.9 69.9 68.9 75.1 72.4 73.8 33.5 21.8** 28.0 75.6 79.9 77.7 85.7 81.6 83.5 87.5 86.9 87.2 53.9 52.3 53.1 54.7 52.3 53.4 38.3 22.8** 30.0 66.9 51.3* 59.6 67.3 79.2** 73.5

Maternal orphans Male Female Total 48.1 23.3** 37.9 84.6 86.4 85.5 24.8 19.4 22.3 63.0 77.1 68.7 76.0 62.3 69.3 66.3 67.0 66.6 69.0 40.3** 55.2 44.9 18.9** 32.6 49.9 38.8+ 44.1 57.4 59.8 58.5 85.5 91.3 88.5 81.2 70.9 77.0 78.3 76.9 77.6 82.1 58.3* 69.8 28.9 11.2** 19.4 71.1 58.4+ 64.3 77.6 78.2 77.9 82.9 85.4 84.2 47.8 41.7 44.7 48.3 53.8 50.8 33.0 19.5** 26.0 69.6 58.4+ 63.8 69.0 73.0 71.1

Two-parent orphans

Male Female Total # 10.1 20.1 # 83.0 87.2 31.5 21.0 25.5 70.9 66.6 69.0 # # 66.0 80.6 65.9 72.5 59.3 36.3* 46.5 45.6 19.4* 33.8 42.1 35.9 38.8 # # # # # # 67.6 69.2 68.4 83.6 # 73.6 # # 74.4 27.1 35.2 31.1 51.1 68.2+ 59.9 63.1 74.7 68.0 80.1 68.1 72.8 43.9 37.2 40.6 33.4 47.1 39.0 15.0 30.2 24.3 33.0 31.0 32.1 71.6 # 73.4

Total

Male Female Total 57.1 34.3** 46.0 94.6 94.7 94.7 35.3 25.5** 29.9 74.9 73.1 74.1 74.8 67.7** 71.2 79.9 75.0** 77.5 70.1 51.5** 61.1 44.0 26.9** 35.5 60.2 44.4** 52.3 69.0 76.3** 72.6 93.5 94.5 94.0 81.0 75.2** 78.2 79.9 79.6 79.8 82.7 76.5** 79.7 33.1 23.5** 28.3 75.9 76.0 76.0 84.3 83.3 83.8 91.1 89.7+ 90.4 60.4 61.3 60.8 64.2 61.2+ 62.6 33.7 24.5** 29.0 65.2 55.0** 60.1 76.9 81.3** 79.1 (Continued on the next page.)

Appendix 7 (continued). School enrollment by orphan status and gender ages 7-14

Both alive Paternal orphans Maternal orphans Two-parent orphans Total

Data set Year Male Female Total Male Female Total Male Female Total Male Female Total Male Female Total Niger DHS 1998 31.0 21.4** 26.3 24.6 22.5 23.6 26.9 17.1** 22.2 # # 22.1 30.1 21.0** 25.7 Nigeria DHS 1999 70.7 64.6** 67.8 71.8 75.9 73.7 75.3 65.0 71.3 55.1 75.0+ 66.5 70.3 64.5** 67.6 South Africa OHS 1995 97.1 97.0 97.0 96.7 97.2 96.9 93.0 94.1 93.5 96.5 94.9 95.7 97.0 96.9 96.9 South Africa OHS 1998 92.6 93.9* 93.3 92.8 92.7 92.8 93.5 96.9 95.3 93.2 88.1 90.6 92.6 93.7* 93.2 Senegal DHS 1992/3 40.1 31.7** 35.9 41.6 20.2** 31.2 44.5 32.1** 39.2 # # 9.1 39.9 30.7** 35.4 Tanzania DHS 1991/2 52.7 53.6 53.2 56.1 57.1 56.6 54.3 53.3 53.9 36.3 39.6 37.9 52.8 53.6 53.2 Tanzania DHS 1996 52.1 55.4** 53.7 61.3 58.6 59.9 57.3 55.0 56.2 55.0 66.0 60.7 52.7 55.8* 54.3 Togo DHS 1998 81.7 67.7** 75.1 77.7 60.7** 69.7 85.0 66.9** 76.9 67.5 48.2+ 59.6 81.1 66.5** 74.2 Uganda DHS 1995 77.3 72.5** 74.9 70.5 63.0* 66.7 73.4 68.0 71.0 77.3 72.1 74.7 76.1 71.2** 73.6 Uganda UNHS 1999/0 90.9 89.9 90.4 87.6 88.2 87.9 94.4 90.6 92.5 87.9 89.0 88.4 90.5 89.6 90.1 Zambia DHS 1992 78.2 77.5 77.8 69.1 74.4 72.0 71.9 65.1 68.5 79.5 75.0 77.0 77.4 76.6 76.9 Zambia DHS 1996/7 68.8 68.4 68.6 60.0 64.5 62.0 64.0 70.2 66.9 64.5 64.2 64.4 67.4 67.8 67.6 Zambia LCMS 1996 71.0 71.1 71.1 74.0 66.42* 70.2 65.7 64.2 65.0 72.0 71.6 71.8 71.0 70.2 70.6 Zambia LCMS 1998 68.9 68.4 68.7 70.0 68.3 69.2 65.9 66.0 65.9 57.8 59.5 58.7 68.5 68.1 68.3 Zimbabwe DHS 1994 91.4 90.0 91.0 89.8 89.1 89.4 88.7 82.3 85.3 94.0 94.8 94.4 91.3 89.9+ 90.6 Zimbabwe DHS 1999 90.1 89.9 90.0 88.7 88.0 88.4 87.6 83.8 85.5 82.4 78.0 80.0 89.4 88.9 89.1 # indicates a cell size of fewer than 20 observations. All significance tests are whether the enrollment for females is different from males, within the orphan status group. + indicates significance at the 10 percent level, * indicates significance at the 5 percent level, and ** indicates significance at the 1 percent level.

M:\\Working Papers\\Filmer\\wps.targeting.ainsworth.filmer.aug22.2002.doc August 22, 2002 11:51 AM

因篇幅问题不能全部显示，请点此查看更多更全内容

查看全文

全部栏目

Poverty, AIDS and Children’s Schooling A Targeting Dilemma