tAleS I 41 POLICY RESEARCH WORKING PAPER 1867 Gender Disparity While gender disparities in health and education in South Asia outcomes are higher on average in South Asian than in other countries, the large Comparisons Between within country differences in and Within Countries gender disparity, between Indian states or Pakistani Deon Filmer provinces, demand more Elizabeth M. King local explanations, Lant Pritchett The World Bank Development Research Group Poverty and Human Resources U January 1998 I POLICY RESEARCH WORKING PAPER 1867 Summary findings Using data assembled from the Demographic Health worse than in any country in the world, although in Surveys of over 50 countries and from the National another state (Tamil Nadu) it is lower than in all but Family Health Surveys of individual states in India, three countries. Filmer, King, and Pritchett create a new data set of * Across and within the set of developing nations, comparable indicators of gender disparity. They establish gender disparity is not only a phenomenon of poverty. three findings: There is almost no correlation between per capita - As is by now well-known, the level of gender income and the gender disparities in health and disparities in health and education outcomes for girls in education outcomes. So although absolute levels of South Asia is the highest in the world. health and education outcomes for girls are strongly * Even within South Asia, and within India or related to economic conditions, the disparities between Pakistan, there are huge variations in gender disparity. outcomes for girls and boys are not. Differences in gender disparity among Indian states or Understanding what causes such great gender disparity among provinces of Pakistan are typically greater than within South Asia is the next pressing question for those among the world's nations. The ratio of female to researchers. male child mortality in one Indian state (Haryana) is This paper-a product of Poverty and Human Resources, Development Research Group- is part of a larger effort in the group to understand the determinants of gender differentials. The study was funded by the Bank's Research Support Budget under the research project "Explaining Gender Disparity in South Asia" (RPO 681-29). Copies of this paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Sheila Fallon, room MC3-638, telephone 202-473-8009, fax 202-522-11S3, Internet address sfallon@worldbank.org. January 1998. (59 pages) 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 he 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. Produced by the Policy Research Dissemination Center Gender Disparity in South Asia: Compariscns between and within Countries Deon Filmer Elizabeth M. King Lant Pritchett The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necess2rily represent the views of the World Bank, its Executive Directors, or the countries thtey represent. The paper should not be cited without the permission of the authors Thlis research was funded by a grant from the World Bank research committee (RPO 681-26). The maps in this paper are used for illustrative purposes only. The boundaries, denominations, and classifications do not imply any judgement on the legal or other status of any territory, or any endorsement or acceptance of any boundary. Gcnder Disparity in South Asia: Comparisons between and within Nations Introduction On one level, gender d:isparity can be narrowly defined as the purely descriptive observation of different outcomes between males and females. However, to move beyond the descriptive level to ask what night cause gender disparities reaches into the complex interplay of the possible sources. Discrimination (the differential treatment of individuals because of their gender), biological differences, individual and societal beliefs and attitudes about appropriate gender-specific roles, and the choices of individuals and households based on all of these factors (and more, such as an individual's own circumstances) all play a role in determining gender disparities. These factors are causally interrelated and it is very difficult to disentangle what are the underlying causes and what are merely proximate indicators or symptoms. Our objective in this paper is very modest. We will not propose a theory of the causes; we remain entirely at the descriptive level of differences in outcomes. While the data contain a wide range of indicators, we focus on and document gender disparities principally for children in two main areas: health outcomes, including treatment of illness, and educational enrollments'. We focus on these outcomes for children for two main reasons. First, the data are available and generally comparable across countries. Second, outcomes for children, such as child mortality, are relatively less influenced by choices of the lThe full set of indicators is described and summarized in Appendices 1 and 3. 2 children themselves and potentially more indicative of differential treatment by their parents (and other adults) and hence may indicate more clearly one particular source of gender disparity2. This approach of using the rich data provided by the Demographic and Health Surveys (DHS) to examine gender disparity has been used in past studies (Arnold, 1992, Hill and Upchurch, 1995) which also examined gender differentials in outcomes using these data. This study updates theirs and includes the relatively unique feature of combining data at the country level with data from individual states of India and provinces of Pakistan. What's the value added of this innovation? In comparative data each country is typically treated as a single unit, irrespective of its size. India is enormous, with a population of over 900 million which is twice as large as all countries in Africa combined, or all countries in Latin America (South and Central America and the Caribbean) combined. The state of Uttar Pradesh has a population of over 140 million, which would make it the world's sixth largest country. But size alone is not the right criterion for determining the value of cross-country (versus within-country) comparisons, as it depends on the empirical characteristics of the phenomena to be examined. Since countries, almost by definition, share a common currency and have free internal trade, one might expect economic phenomenon like inflation to be similar in all parts of the same country. In examining the relationship between money supply growth and inflation, little would be gained by considering Minneapolis and Miami separately, 2VWhile some might argue that some gender disparities, like differential participation in the paid labor force or the gender division of labor, are the result of a joint, voluntary, and optimal decision-making on the part of a couple as a household unit, it would be very difficult to argue that female children voluntarily assume higher mortality risks. 3 as over the long-run they wouild be expected to have nearly the same inflation rate3. If, on the other hand, one were examining an economic outcome determined by local conditions, like expenditures on heating, little would be gained by aggregating cities like Minneapolis where the average January temperatire is 11.8F (-11.2C) and Miami which basks in 67.2F (19.6C) temperatures even in January simply because they are in the same country. The value of disaggregating data on gender disparities to the subnational level depends on the extent to which they are "national" phenomena determined by conditions associated with a particular national government, versus the extent to which they are local phenomena, determined by social, cultural, environmental, economic or political conditions that vary sharply within a country. A,s documented below, our fnding is that while gender disparities in South Asia are partly "natioral", but there are also enormous variations within countries. I) The data A perennial problem with cross-national comparisons is without doubt the degree of comparability and reliability of the data across countries. In order to avoid these problems, we use data drawn from a ccllection of household surveys that used a nearly identical survey instrument and methodology in each country (or area) -- the Demographic and Health Surveys (DHS) and the National Family Health Surveys (NFHS)4. 3 Between 1980 and 1992 cumulative inflation was 66 percent in Miami and 71 percent in Minneapolis (Statistical Abstract of the United States, table 759). Details on the questionnaires and surveys are in Appendix 1. 4 The DHS are nationally representative random samples of households, from which a woman of reproductive age was interviewed5. The data used here are from 69 surveys from 52 countries carried out between 1985 and 1995, with sample sizes varying from 1,623 (Nepal 1987) to 28,168 women (Indonesia 1994). The measures of gender disparity are derived from the Final Report issued for each survey. The questionnaire has been revised several times and not all questions are asked in all countries, so that jiot all data are available for all countries. Moreover, certain of the statistics are reported for slightly different populations; for example, the coverage for Acute Respiratory Infections ranges from children under three years old in some countries to children under five in others. While these may affect levels, we hope that these differences do not substantively affect the comparisons of female to male ratios of the indicators. In general, however, these are some of the most comparable data of this type, and certainly the most widely available. The NFHS were carried out in 26 Indian States between 1992 and 1993 using a format very similar to that of the DHS. The sample sizes in the states range from 882 (Arunachal Pradesh) to 8,722 women (Uttar Pradesh). Although the survey instruments were very similar across states, some of the Final Reports do not report gender-disaggregated results for certain indicators (primarily because of small sample sizes). Hence, even within India, not all the data are available for all states. National representativeness is usually achieved through weighting as the samples are often stratified random samples. 6 The sample used in Arnold (1990) includes 26 surveys and countries, that in Hill and Upchurch (1995) includes 38 surveys from 35 countries. 5 The 1990/91 DHS Firnal Report for Pakistan does not report gender-disaggregated outcome measures by province. In order to provide comparable information for these provinces, we use the DHS raw data and reconstruct the outcome measures. Because of the small sample sizes in Baluchistan, additional information was used to create accurate measures of gender disparity in that province7. 11) Gender disparities betweei and within countries in South Asia We discuss two broad types of indicators: those related to health, which include data on mortality, morbidity, health treatments, and anthropometrics; and educational enrollments at various ages. A) Health indicators Mortality outcomes. As an indicator of gender disparity, the ratio of female to male child mortality, defined as the chance of a child dying after turning one but before turning five, is preferable to other mortality indicators. Unlike other studies, we do not normalize mortality rates to a reference population but report actual ratios, a point we return to below. Under-five mortality (5qO) is conventionally divided into neonatal (less than one month), post- neonatal infant (between one month and the first birthday), and child (4ql) mortality8 as deaths at these various ages are typically due to different causes and reflect different disease conditions and health-seeking behaviors by parents. There are two reasons to believe that 4ql ' See Appendix 1 for details. The notation 4ql deniotes the probability of a child dying between exact age one and the fifth birthday, and 5qO is the probability of dying between birth and the fifth birthday. 6 is a better indicator of gender disparity, especially of the type that reveals behavioral differences, than mortality at other ages. First, there is very little correlation between the gender ratios of neonatal and child mortality; in the sample outside of South Asia, the correlation is even mildly negative. It is reasonable to believe that, aside from outright infanticide, neonatal deaths are unlikely to be influenced by gender, because they are partly genetically determined and partly determined by prenatal care, when few mothers know the sex of the child. Second, the disparities in the mortality ratios by gender grow with age. The median gender ratios in the non-South Asian countries, Indian states, and Pakistani provinces are nearly the same for neonatal mortality (.78, .82 and .75), somewhat different for post-neonatal mortality (.93, 1.13 and 1.02), and, as shown in table 1, very different for child mortality (1.0, 1.43 and 1.52). Similarly, the variance of outcomes is much larger at later ages. The cross-regional standard deviation of the gender ratio of mortality grows from .10 for neonatal to .22 for post-neonatal, and .34 for child mortality. These two facts suggest that the underlying pattern emerges more clearly in the later ages, as mortality then is less dependent on intrinsic genetic conditions and more determined by behavior. 7 Table 1: Ratio of female to male child mortality (4ql) N Med- Std. Range ian Dev Highest Lowest Non-South 58 1.01 .17 Egypt (1992) 1.46 Kazakstan (1995) .47 Asian Paraguay (1990) 1.23 Ghana (1993) .51 countries NE Brazil (1991) 1.22 Colombia (1990) .51 India 19 1.44 .35 Haryana 2.35 Tamil Nadu .80 (States) Punjab 1.81 Kerala .94 Uttar Pradesh 1.70 Goa 1.11 Pakistan 4 1.52 .54 Punjab 2.06 NWFP .86 (Provinces) Baluchistan 1.79 _ Sindh 1.24 South Asia 5 1.33 .25 Pakistan (1990) 1.66 Sri Lanka (1987) .99 India (1992) 1.43 Nepal (1996) 1.24 ______ Bangladesh (1993) 1.33 Notes: A higher ratio indicates that females are MORE likely to die between ages one and four than are males. What do the data on child mortality ratios show? First, that gender disparities are worse in South Asian countries than in the rest of the world. Figure la shows a map of the world where the shading given each country reflects the female to male ratio in child mortality9. The median in tiLe non-South Asian countries is exactly one, meaning that in a typical DHS country outside of South Asia, male and female children are equally likely to die between ages one and five. In Pakistan, India and Bangladesh, however, this is not the case. A girl is between 30 and 50 percent more likely to die between her first and fifth birthdays The cutoff points between the various levels are determined by pooling all the (country, Indian state, and Pakistani province) data and selecting the values at the 15th, 50th, and 85th percentiles. 8 than a boy'0. Sri Lanka, in contrast, is more typical of non-South Asian countries. These findings confirm what has been shown in other data using, for instance, the difference in population sex ratios among countries (Sen, 1989). However, a second important result is that this gender disparity is not uniform within South Asia, nor, more importantly, within the countries in South Asia. Particularly strildng is the band across northern India and Pakistan where this gender disparity is substantially worse than in the south (Figure lb). Moreover, there are much greater differences in gender disparity among the states of India than among countries in the rest of the world. Figure Ic shows the distribution of the mortality ratio within each country or region (as well as the area- specific values at the 25th, 50th, and 75th percentiles). The standard deviation is almost twice as large for Indian states (.35) than for the non-South Asian countries (.17). Some Indian states have quite low mortality differentials, actually slightly favoring girls (Tamil Nadu and Kerala), while many other Indian states have ratios higher than any non-South Asian country in the world. The same is true in Pakistan: some provinces are much worse, such as the Punjab where girls are twice as likely as boys to die, while others, like the Sindh, have mortality ratios that are higher than the international average but not as extreme. This large discrepancy between the "Northern Crescent" (mainly Pakistan and northern India) and the rest of India and South Asia has been noted before (see the discussion below). The country-wide numbers are not just the mean of the state (or province) level numbers as these would need to be population weighted. 9 Table 2: Ratio of fenLale to male children who received no treatment for episodes of fever or acute respiratory infection (ARI) N Med- Std. Range ian Dev Highest Lowest Non-South 34 1.02 .29 Colombia (1990) 2.04 Paraguay (1990) .57 Asian Togo (1988) 1.58 NE Brazil (1991) .66 countries Ondo State (1986) 1.44 Ghana (1988) .71 India 19 1.34 .29 Rajastan 1.79 Karnataka .62 Punjab 1.67 Tamil Nadu .85 Bihar 1.59 Kerala .88 Pakistan 4 1.28 .55 Baluchistan 2.01 Sindh .68 NWFP 1.38 _________ Punjab 1.18 South Asia 3 1.19 .11 India (1992) 1.27 Pakistan (1990) 1.05 I__ _ | Bangladesh (1993) 1.19 Notes: A higher ratio indicates that females are LESS likely to get treatment than men. Health treatment beiavior. While mortality outcomes are clearly different between sexes, can we also see differences in the underlying gender discriminating behaviors that produce those outcomes? lVhile it is relatively straightforward to document gender disparity in mortality ratios, it has been quite difficult to create comparable indicators of differential health treatment. In the DIIS, the women surveyed were asked both whether their children (under a certain age, usually five years) had experienced fever, acute respiratory infection (ARI)," and diarrhea. If a mother reported her child had suffered from one of these conditions, she was then asked about various types of treatment the child had received. We The actual survey question refers to a cough with rapid breathing. 10 focus on the likelihood that, among the alternatives, a female versus a male child received "no treatnent" when suffering from fever or ARI'2. This table confirms both findings on mortality outcomes above. First, in the typical non-South Asian country, there is almost no gender disparity at all in whether children receive "no treatment" (1.02). In contrast, at the national or regional level, girls in South Asia are significantly less likely to receive treatment. The median for Indian states is 1.34, and for Pakistani provinces, 1.28. But again, there is significant variation within the South Asian countries and the differences are as large within India as they are among non-South Asian countries. In certain states girls appear to be more likely to receive treatment (Tamil Nadu and Karnataka), while in others (Rajastan and Punjab) girls are strikingly less likely to receive treatment. There are similarly notable differences across the provinces of Pakistan. Morbidity and anthropometrics. The reported frequencies of episodes of fever, ARI, or diarrhea do not indicate gender disparity. In contrast to either mortality outcomes or health care, these indicators are not higher in South Asia; nor are they correlated with mortality outcomes. This finding is consistent with either of two explanations. Perhaps, actually contracting diseases is equally likely for both genders and that only treatment differs, accounting for the different mortality outcomes. An alternative explanation is that female 12 The data on diarrhea does not appear to be as reliable. In some countries there is not information on both fever and ARI. If only one exists, then the other is inferred; see Appendix 1 for details. 11 morbidity is under-reported; hence, the differentials in actual treatment per episode are even larger than reported. Similarly, the anthropometric estimates do not indicate gender disparities. This could be for a variety of reasons, but one is that the severe malnutrition indicators, such as the fraction of the population that is three (or even two) standard deviations below reference group norms, were very unreliably estimated13. B) Educational enrollment For the households interviewed in the more recent surveys, the DHS reports whether or not each child is "attending" school. Beyond the questions of national (or area) representativeness, the data are potentially better than national statistics both because they are based on reported attendance, not official enrollment, and because they compare children of similar ages rather than by grades. The data are broken down into two age groups, ages 6-10 and 11-14 (with some variation in the cutoff ages across countries). We focus mainly on children in the 11-14 age range, as they are still young enough to be part of the "basic" education cycle of most countries but are reaching the ages when any gender-based discrimination in education may worsen. In contrast, the 6-10 age range is potentially more problematic an indicator as ,so many countries have achieved 100 percent enrollment for both sexes. The correlation between the gender ratios of enrollment for the two age groups is quite high at .84. '3 This lack of association among alternative indicators was found also in the other studies discussed below. 12 The ratio of female to male enrollment confirms the patterns of gender disparity in mortality outcomes and health treatments (see table 3). The ratio for the countries outside of South Asia is .91 (see also figure 3a) which is perhaps surprisingly near gender neutrality, although one might suspect that the real differences emerge more strongly beyond the basic education level, in upper secondary and tertiary education. The average of Pakistani provinces is strikingly lower, with even the highest province-level ratio reaching only .69. In India the median for the 25 states is .86. But since in some of the larger states the ratio is quite low -- for instance, in Uttar Pradesh the ratio is only .6, which is near the lowest for any country in the world -- the overall mean for India is only .72. Again, the disparities within India and Pakistan, and certainly within South Asia, are nearly as large as the differences across countries. India has states with no gender disparity at all (Kerala) and states in which girls are only half as likely to attend school (Rajastan) (figures 3b and 3c). Within South Asia, Bangladesh appears to be doing modestly better. 14 14 This may be due to the fact that Bangladesh has been targeting education subsidies towards girls in an explicit effort to achieve gender equality in schools. In 1992, the government initiated a program of free tuition for all girls attending junior secondary schools, which resulted in a dramatic rise in female enrollment that year. This program was replaced in 1994 by an expanded nationwide program that offers both free tuition and stipends to girls in secondary schools. 13 Table 3: Ratio of female to male enrollment, children aged 11-14 N Med- Std. Range ian Dev Highest Lowest Non-South 36 .91 .19 Zambia (1992) 1.09 Jordan (1990) .33 Asian NE Brazil (1991) 1.09 Yemen (1991) .37 countries Dom. Rep. (1991) 1.05 Niger (1992) .50 India 25 .86 .14 Nagaland 1.01 Rajastan .49 Kerala 1.00 Bihar .55 DeLhi 1.00 Uttar Pradesh .60 Pakistan 4 .52 .19 Punjab .69 Baluchistan .34 Sindh .66 NWFP .37 South Asia 4 .70 .13 Bangladesh (1993) .93 Pakistan (1990) .64 India (1992) .72 Nepal (1996) .67 Notes: A higher ratio indicates that females are MORE likely to be enrolled than are males. III) Patterns of overall gender disparity The first question that might occur to one when observing cross-national differences in any indicator of the standard of living is to ask to what extent the differences are associated with differences in the overall standard of living, say, as measured by overall per-capita income. The levels of man) socio-economic indicators are strongly associated with per-capita income, like infant or under-five mortality (Pritchett and Summers, 1996, Filmer and Pritchett, 1997), male and female educational attainment and enrollment levels (King and Hill, 1993; Ahuja and Filmer, 1996), malnutrition, and the fraction of population in poverty (Bruno, Ravallion, and Squire, 1996). This is true in our data as well: the level of income has 14 a strong relationship with the level of child mortality and the enrollment rate. However, other indicators, particularly of the distribution of the standard of living, are not at all correlated with the average level of income (Bruno, Ravallion, and Squire, 1996). Our indicators of gender disparity also do not appear to be at all correlated with the general standard of living, as proxied by per-capita GDP, either across countries or within countries in South Asia. Table 4 reports the results of regressing three gender disparity measures on per-capita GDP"5. The coefficients are very small. For instance, the coefficient of .031 in the child mortality regression suggests that an increase of 100 percent in per-capita income from the median would raise the female to male ratio by only roughly 3 percent at the median. Moreover, the t-statistics are consistently less than one; hence, the estimates are imprecisely estimated and decidedly statistically insignificant. There is no particular pattern to the coefficients, either across indicators or regions. This lack of a relationship has several implications. First, if one is seeking to understand and explain the high levels of gender disparity in South Asia, low income is not one of the answers. In the sample, poorer countries do not, on average, have worse gender disparity than high-income countries. Moreover, within India the high-income areas also do not, on average, have less gender disparity. Gender disparity does not appear to be something that economies "grow out of."16 15 Or, in the case of the states of India or provinces of Pakistan, a proxy for GDP per capita, see Appendix 1. 16 Easterly (1997), using a broader sample that includes the richer countries, finds a relationship with gender specific enrollments. 15 Table 4: Coefficient of per-capita income in regressions of gender disparity on per- capita incorrmes Gender disparity Female level (female relative to male) Indicators All* Non South All* Non South South Asia* South Asia* Asia Asia Child mortality .031 .022 .117 -47.9 -50.33 -24.36 (4ql) (.68) (.72) (0.51) (9.89) (8.85) (2.52) 82 56 26 81 56 25 Enrolled in .056 .043 .133 13.04 13.14 12.44 school (ages (1.41) (1.01) (1.26) (2.78) (2.52) (1.03) 11-14(15)) 67 36 31 67 36 31 No treatment .052 .064 -.010 -6.17 -4.34 -15.04 for fever or (.65) (.78) (.05) (2.11) (1.28) (2.36) ARI 55 31 24 55 31 24 Notes: Each cell entry includes the coefficient (and t-statistic) for the natural log of GDP per capita, and the number of observations in the OLS regression. * Includes regional subgro'pings within India and Pakistan but excludes observations at the national level for those countries. That said, the absolute level of female mortality is highly correlated with income both across countries and within India. In table 4, the coefficient of 47.9 in the child mortality regression suggests that a 14)-percent increase in per-capita income from the median would lead to a 12-percent fall in the female child mortality rate. Therefore, rising income levels that do not worsen the gender disparity will tend to reduce both female and male mortality, and the disparity itself, the ratio of girls to boys that die, will not improve. An overall strategy for 16 improving female health status might then involve actions both to improve overall standards of living together with measures aimed at the disparity"7. IV) Relationship to previous work We are obviously not the first to notice gender disparity in South Asia or its striking variation across states of India. Our modest contribution in this work is to bring together a new set of indicators that are comparable among nations and regions within South Asia. In this section we compare our results first to other studies within India, and then to other cross- national analyses. A) Results for within-India First, however, how do our indicators compare with other rankings? Table 5 displays the raw data on various indicators and how the various Indian states rank on each indicator (with higher ranks representing less gender disparity). This comparison shows how difficult analyses such as this can be. Some findings are robust: Kerala, reassuringly, is consistently near the top, while Uttar Pradesh is consistently near the bottom. However, Tamil Nadu is consistently near the top for the health indicators, but towards the bottom for education, and vice versa for Haryana. These differences could be a sign of the data being of dubious quality, or that the causes of the gender disparities vary across the types of outcomes under study. ' The results for the effect of income on the pooled male and female mortality rates are in Appendix table A2. 1. 17 Table 5: Indicators of gender disparity at the level of Indian States from various sources. State Child mortality' Female / Male Female / male Female / Male ratio Male female Rural female Difference of Status of women' Population Ratiob ratio of no of school enrollment difference in rural labor force Gender and treatment for ARI (ages 11-14)' literacyb participationb unadjusted HDId or fevera Level Rank Level Rank Level Rank Level Rank Level Rank Level Rank Level Rank Level Rank Andrha Pradesh 1.28 7 973 3 1.45 15 0.67 22 19.3 7 46.60 8 7.3% 6 34.6 10 Arunachal Pradesh 861 23 0.92 8 18.3 16 67.10 2 27.0 19 Assam 1.12 4 925 13 1.27 7 0.90 10 15.7 19 55.40 6 8.4% 8 24.2 22 Bihar 1.55 13 912 18 1.59 17 0.55 24 24.1 21 15.30 18 13.6% 13 20.2 24 Delhi 1.55 12 1.16 4 1.00 3 3 25 36.9 6 Goa 1.10 3 1.38 13 0.98 5 24 49.0 1 Gujarat 1.42 9 936 10 1.39 14 0.79 16 23.4 9 20.20 14 6.4% 4 34.2 12 Haryana 2.34 19 874 22 1.28 8 0.83 14 24.7 12 7.60 23 24.3% 16 30.6 15 Himachal Pradesh 1.43 10 996 2 1.28 9 0.92 7 19.0 6 29.20 12 4.8% 1 34.7 9 Janmuu and Kashmir 1.69 16 923 15 1.34 10 0.86 13 13 9.20 22 30.1 16 Karnataka 1.30 8 961 6 0.62 1 0.77 19 20.9 8 33.40 11 6.9% 5 34.6 11 Kerala 0.94 2 1040 1 0.87 3 1.00 2 5.8 20.20 13 6.3% 3 47.6 2 Madhya Pradesh 1.21 5 932 12 1.19 5 0.74 21 24.9 17 39.70 10 10.6% 9 25.5 20 Maharashtra 1.24 6 935 11 1.58 16 0.81 15 23.9 4 47.30 7 5.9% 2 36.9 7 Manipur 961 7 0.88 11 18.3 61.20 3 41.6 3 Meghalaya 947 8 0.99 4 4.8 11 60.80 4 33.7 14 Mizoram 924 14 0.96 6 8.4 1 60.60 5 40.9 4 Nagaland 890 19 1.01 1 11.0 2 72.60 1 40.8 5 Orissa 1.45 11 972 5 1.37 12 0.75 20 23.8 20 16.70 15 11.8% 10 24.0 23 Punjab 1.81 18 888 20 1.67 18 0.91 9 13.7 10 16.10 16 19.8% 15 33.7 13 Rajastan 1.59 14 913 17 1.79 19 0.49 25 28.8 18 16.10 17 13.2% 12 25.2 21 Tamil Nadu 0.80 1 972 4 0.85 2 0.79 18 21.5 5 39.80 9 8.2% 7 36.1 8 Tripura 946 9 0.86 12 18.8 14 14.30 19 29.3 17 Uttar Pradesh 1.70 17 881 21 1.34 11 0.60 23 25.8 22 9.40 21 15.8% 14 19.9 25 West Bengal 1.63 15 917 16 1.22 6 0.79 17 19.0 15 10.00 20 13.1% 11 28.3 18 Sources: (a) This study (b) Agarwal (1997) (c) Srivastava (1997) (d) Kumar (1996) 18 Table 6 shows the (rank) correlations of the various indicators to gauge the internal coherence of the set. The various indicators of gender differentials in health status are reasonably highly correlated: .77 between child mortality and the population sex ratio, and .36 and .40 between the health treatment disparity and child mortality and the population sex ratio, respectively. Between the enrollment ratio and the male-female literacy gap, the correlation is .57. The labor force participation of rural women is quite highly correlated with child mortality. The two indicators which are aggregates of various others, in particular, the "status of women" indicator which was created from various parts of the NFHS data, are (unsurprisingly) quite highly correlated with several of the other measures. Table 6: Rank correlations among the various indicators of female I male disparity I II m IV v VI vii VIIi I Child mortality' 1 11 F I M Population Ratiob .77 i m F / M no treatment for ARI or fever' .36 .40 1 IV F I M school enrollment (ages 11-14)' .22 .06 .31 1 V M-F difference in rural literacyb .31 .33 .23 .57 1 VI Rural female labor force participationb .61 .23 .08 .20 .32 1 VII Status of women; .75 .83 .28 .39 .64 .72 1 vii Difference of Gender and unadjusted HDId 50 .45 .25 .64 1 .22 .68 1 Sources: (a) This study (b) Agarwal (1997) (c) Srivastava (1997) (d) Kumar (1996) There has been a fairly large body of literature which argues about the patterns and causes of gender differentials in outcomes within India, and we will not delve much into the 19 debate here beyond stating that our findings are in line with those found elsewhere."8 The most robust finding is that of a band across the North-Western States of India (and which extends into several provinces in Pakistan) in which there are large disparities in child mortality rates. The high gender disparities in Northern India and Pakistan have been pointed out before, perhaps most notably in Miller (1982, 1989, 1993), Murthi et al (1995), and Agarwal (1997) in which the focus was on juvenile sex ratios. Moreover, Miller (1982) points out the persistence of the disparity, going back to evidence from the 1872 census. Miller (1982) attributes the mortality differential to the relative neglect of girls in the allocation of food, medical care, and "love and walmth." Others have focused more specifically on the role of the quantity or quality of medical care (e.g. Das Gupta, 1989, Wyon and Gordon, 1971). In our results, however, there is a puzzle as to how these differentials come about: although we find that female disadvantage in absence of health care is higher in South Asia than elsewhere, neither differences in absence of health care nor differences in nutritional status (as reflected by wasting and stunting) show nearly as pronounced a band across the Northern States as mortality.19 Our finding that gender disparities are not systematically decreasing with income is also in line with other studies (see Murthi et al, 1995, for a description of this debate), although the result is perhaps not surprising given that gender disparities are observed to be A useful summary of the claims and counter-claims and an assessment of these can be found in Murthi et al (1995). 9 As pointed out before, the latter of these non-findings might be due to the fact that one is essentially looking at the tails of distributions of which may be hard to get reliable estimates. 20 highest in the generally wealthier Northern States. We do not find, however, that gender disparities are statistically significantly higher in states with higher income. There is no consensus on the underlying determinants of the greater gender disparities found in the Northern Indian States. Much of the debate centers around the "worth" of female children, with the obviously problematic definition of "worth" being at the contentious center of the discussion. Miller (1982), for example, emphasizes the different roles of women in agriculture in the North versus the South. In the North, dry-field wheat cultivation is hypothesized to lead to a low demand for female labor relative to the South where wet-rice cultivation leads to a high demand for such labor. Rosenzweig and Schultz (1982), analyzing data for rural India, find that where the female employment rate was higher the sex difference in survival probabilities was somewhat smaller. Another hypothesis suggests that the increased relative bargaining power of women in contexts where their economic opportunities are higher, combined with a preference of women for investing in the human resources of their daughters (relative to the preference of men for investing in their daughters), leads to higher relative survival rates of female children (Folbre, 1984). In contrast, Das Gupta (1987) emphasizes the cultural rights and obligations which lead to a higher long-term value of a son relative to a daughter which she argues leads to dramatically high death rates for higher birth order daughters. B) Other results across countries. Normalization. Our findings are also not the first to document cross-country differences. However, to compare the results of this study to others, a short digression on normalization is in order. Many other studies of gender disparities in mortality normalize all 21 of the mortality ratios to a reference population. Hill and Upchurch (1995) use mortality rates in six Northern European cc'untries from 1820-1963 and adjust the differential to the average level of mortality. Svedberg (1990) uses mortality from Sweden in the 1980s. Klasen (1996) and Svedberg (1996) debate the use of mortality in Sweden in the 1980s versus that in the "North" and "West" Model Life Tables of Coale, Demeny and Vaughn (1983). Since girls appear to exhibit biologically higher survival capability a normalization to underlying "nratural" mortality rates would imply that an equal mortality ratio is an indicator of female disadvantage. We report acLual mortality differentials because we feel that any given choice of normalization and its interpietation are problematic. Reference populatiors are presumably chosen in order to answer a question like: "If there were absolutely no geinder-based differential treatment, what would the observed mortality ratios be?" One way to answer this question is to use the mortality rates of a historical reference population, say Northern Europe in the 19th century, not necessarily because there was no gender discrimination, but because medical technology was (at best) ineffective, implying that mortality rates would not reflect differences in health-seeking behaviors based on gender. However, this seems problematic as medical advances that led to mortality gains in some causes of death and not others might then appear as increases in discrimination. For exampl,., suppose in the historical period that one in ten boys died of disease and one in ten died of unavoidable accidents, and no girls died of disease and one in ten died of unavoidable accidents. Now suppose that advances in medical technology made it possible to cure all diseases, A country in which children only died of unavoidable accidents would have equal mortality Dutcomes, but when normed to the historical reference period, 22 would appear to have an enormous gender disparity of 2. It hardly seems right that medical progress that leads to better survival chances be called an increase in gender disparity. A second possibility is to norm to recent mortality rates in a richer, low mortality country. This represents the mortality when medical resources are effective and plentiful, which might represent "natural" mortality. As Svedberg points out, in one setting where medical technology is as advanced as possible (Sweden in the 1980s), "the child mortality rates are almost identical [for boys and girls], .29 and .28 respectively" (Svedberg, 1990, fn. 3). But this is not always the case. In the U.S., the death rate in ages 1 to 4 is .52 for boys and .41 for girls, so the gender ratio is .78. Hence, a country with equal male and female death rates normed to the U.S. ratios would have a gender disparity of 1.27. However, 62 percent of the gender difference in mortality is due to a greater frequency of accidental death from accidents, as the gender mortality ratio for accidents is .68 versus .86 for all other causes of death. If a country with equal death rates were normed to the U.S. death rates from non- accidental causes alone, the gender disparity would be 1.15. It seems extremely odd to argue that the reported gender disparity for a country with gender equal death rates should be so strongly influenced by the propensity of a reference population of U.S. boys to die of fatal accidents. Results. The recent exchange between Svedberg and Klasen (Svedberg, 1990, 1996, Klasen, 1996) highlights the potential importance of normalizing. When normalizing mortality rates using the Northern European countries from 1820-1963 as the reference, Klasen finds that within Sub-Saharan Africa DHS samples, nine out of 14 countries exhibit excess female child mortality. Using a collection of 32 surveys from 20 countries compiled by Svedberg, he 23 finds excess female child mortality in 20 of the 32 surveys. Using Svedberg's normalization on the other hand (modemr day Sweden as the normalization, which comes close to no normalization at all for child mortality), the number of samples with excess female mortality is roughly equal to that with excess male mortality. Beyond these disagreements though, the authors concur that anthropometric indicators do not show anti-female bias in the Sub-Saharan African samples. Perhaps more importantly for our paper, the authors both state that the differentials considered in the Sub-Saharan African context are much smaller than those found in South Asia. For the cross-country analysis, our data source is most similar to Hill and Upchurch (1995) as they also use DHR data to construct their index of gender differentials. They find a female disadvantage in under-five mortality in 31 out of 38 surveys relative to the female disadvantage in a set of Northern European countries from 1820-1963 with matched average mortality. As mentioned above, such an interpretation must be made with caution. Consistent with our findings, however., they find that the region with the highest gender disparity in child mortality is the "Middle East Crescent" (which includes Egypt, Jordan, Morocco, Pakistan, and Tunisia). As in our findings, Hill and Upchurch find that expanding the set of indicators does not necessarily reinforce the gender disadvantage result. For example, they find a disadvantage in the percent of females who are immunized in 58 percent of the samples, who are stunted and wasted in only 17 and 24 percent of the samples, who have had diarrhea or ARI in 9 and 26 percent of the samples, and who receive treatment of fever or ARI in 43 and 30 percent of the samples. 24 In an earlier study, Arnold (1992) also used DHS data to assess the prevalence of female disadvantage across countries (not normalized by reference population mortality). He found that the female child mortality rate was equal or higher to that of males in 18 of the 26 countries he included. Similarly to Hill and Upchurch, Arnold found that there was no clear pattern of female disadvantage in the prevalence or treatment of diarrhea, fever, and ARI, nor in the nutritional status indicators available. Using the precursor surveys to the DHS, the World Fertility Surveys (WFS) carried out between 1974 and 1980, Rutstein (1984) reports the female and male mortality between the ages of two and five (3q2). Of 40 countries, Rutstein finds a higher female than male mortality rate in 25 countries.20 The median female to male mortality ratio was 1.05 for the 36 non-South Asian countries, with a mean of 1.05 and a standard deviation of .23. The median female to male ratio was 1.17 for the four South Asian countries (Bangladesh, Sri Lanka, Nepal, and Pakistan) with a mean of 1.22 and a standard deviation of .18. Although Rutstein's findings indicate a somewhat higher rate for the non-South Asian countries than we do, the much higher level in the South Asian countries is consistent with our results. Conclusion This descriptive work is a first step in a research agenda that aims to examine the causes of gender disparity, and where possible, to suggest policies aimed at reducing it. However, even from this preliminary work there are four conclusions. 20 Portugal was dropped in this assessment. 25 First, South Asia is the region of the world in which gender disparities are noticeably the worst and for which this issue clearly constitutes a crucial part of the development agenda. While child mortality in countries outside of South Asia has been nearly equal between the sexes, it is 30 to 50 percenit higher for female than male children in South Asia. Second, even within South Asia, and even within India or Pakistan, there are huge variations in gender disparity. On some indicators of gender disparity, an Indian state may be very near the best or very near the worst observed in the rest of the world. In child mortality, some Indian states like Tainl Nadu and Kerala have much lower gender disparity than the average of non-South Asia countries (with a female to male ratio below 1), while others have a higher gender disparity than any other country in the world. The ratios of 2.35 (Haryana), 2.06 (Punjab-Pakistan), 1.81 (Punjab-India), 1.79 (Baluchistan), and 1.70 (Uttar Pradesh) are all more than a standard dMviation higher than the highest in any non-South Asian country (Egypt, 1.46). Third, unlike many indicators of standard of living and even many social indicators such as enrollment ratios, gender disparity is not correlated with level of income in this set of countries or across region; within South Asia. While the level of female mortality falls with rising incomes around the world, including in South Asia, the ratio between male and female child mortality does not appear to be related to income. Gender disparity is not a problem of poverty. The fourth conclusion, which follows from the above three findings, is that understanding the causes of the large variations in gender disparity within South Asia is a pressing question for research. First, if research into causes of gender disparity could be at all 26 useful in devising remedies when the gaps are so large, this is practically important. Second, the large variation within countries suggests that the underlying causes of gender disparity differ sharply. This variation makes studies within a single country attractive. Third, the fact that some countries in the region and individual units within nations have achieved much lower levels of gender disparity shows that greater gender equality is possible even within the South Asian context. 27 References Agarwal, Bina, 1997. "Gender, Environment, and Poverty Interlinks: Regional Variations and Temporal Shifts in Rural ILdia, 1971-91," World Devel mn, 25(1):23-52. Ahuja, Vinod, and Deon Filmer, 1996. "Educational Attainment in Developing Countries: New Estimates and Projections Disaggregated by Gender," Journal of Educational Planning and,,Administration, Vol3 ,4:229-254. Arnold, Fred, 1992. "Sex preference and its Demographic and Health Implications," In al Family Planning Perspectivs, 18:93-101. Bruno, Michael, Martin Ra,vallion, and Lyn Squire, 1996. "Equity and Growth in Developing Countries," Policy Research Working Paper No. 1563, The World Bank. Washington, DC. Coale, Ansley J, Paul Demeny, and Barbara Vaughn, 1983. Regional Model Life Tables and Stable oulations, New York: Academic Pres. Das Gupta, Monica, 1987. "Selective Discrimination against Female Children in Rural Punjab," Poulation and Develo nt Review, 13(1):77-100. Easterly, William, 1997. " Life During Growth," mimeo, Development Research Group, The World Bank, Washington, DC. Filmer, Deon, and Lant Pritchett, 1997. "Child Mortality and Public Spending on Health: How Much does Money Matter?" mimeo, Development Research Group, The World Bank, Washington, DC. Folbre, Nancy, 1984. "Market Opportunities, Genetic Endowments, and Intrafamily Resource Distribution: Child Survival in Rural India: Comment," American Economic Review, 74(3):518-520. Government of India, 1994. Economic Survey, Ministry of Finance, Economic Division. Hill, Kenneth, and Dawn M. Upchurch, 1995, "Gender Differences in Child Health: Evidence from the Demographic and Health Surveys," PpuIation and Develoment Review, 21(1): 127- 151. King, Elizabeth M, and M Anne Hill, 1993. Women's Education in Developing Countries: Barriers, Benefits and Polie. Baltimore MD.: Johns Hopkins Press. Klasen, Stephan, 1996. "Nutrition, Health and Mortality in Sub-Saharan Africa: Is There a Gender Bias," Journal of I)evelopment Studies, 32(6):913-932, 944-948. 28 Kumar, A.K. Shiva, 1996. "UNDP's Gender Related Development Index: A Computation for Indian States," Economic and Political Weedky, April 6, 1996. Miller, Barbara D., 1982. The Endangered Sex, Neglect of Female Children in Rural North lIndia. Ithaca and London: Cornell University Press. Miller, Barbara D., 1989. "Changing Patterns of Juvenile Sex Ratios in Rural India, 1961 to 1971," Economic and Political Wekly, June 3, 1989. Miller, Barbara D., 1993. "On Poverty, Child Survival and Gender: Models and Mis- perceptions," Third World Planning Review 15(3):6-13. Murthi, Mamta, Anne-Catherine Guio, and Jean Dreze, 1995. "Mortality, Fertility, and Gender Bias in India; A District Level Analysis," Popuation and Developen Review 21(4):745-782 Pritchett, Lant, and Laurence H. Summers, 1996. "Wealthier is Healthier," Joumnal of Human Resources, 31(4):841-868. Rosenzweig, Mark R., and T. Paul Schultz, 1982. "Market Opportunities, Genetic Endowments, and Intrafamily Resource Distribution: Child Survival in Rural India," American Economic Review, 72(4):803-815. Rutstein, Shea Oscar, 1984. "Infant and Child Mortality: Levels, Trends, and Demographic Differentials," World Fertility Surveys Comparative Studies No. 43. Voorburg, Netherlands: International Statistical Institute. Sen, Amartya, 1989. "Women's Survival as a Development Problem," Bulletin of he American Academy of Arts and Sciences, 43. Srivastava, Dr. B.P. 1997. Personal communication on work undertaken at the Institute of Population, Environment and Development, Lucknow, India. Svedberg, Peter, 1990. "Undernutrition in Sub-Saharan Africa: Is There a Gender Bias?" Journal of Development Studies, 26:469-486. Svedberg, Peter, 1996. "Gender Biases in Sub-Saharan Africa: Reply and Further Evidence" Journal of Development Studies, 32(4):933-943. Wyon, John B., and John E. Gordon, 1971. The Khanna Study: Population Problems in the Rural Punjah, Cambridge, MA: Harvard University Press. 29 Appedix 1: Data sources and description The data used in this study are compiled from (1) DHS Final Reports (2) NFHS Final Reports and (3) Reconstructed gender arLd provincial disaggregated outcomes for Pakistan. Except for a very few exceptions, the transfonnati ons required to go from the published data to that used here involve no more than taking the ratio of the female to the male value. The exceptions are discussed below. Baluchistan province of Pakistan Because of the limited data from the Baluchistan province of Pakistan, the data were adjusted as follows. The female/male child mortality ratio (4q1) is not that derived from the DHS but is calculated from the Pakistan Integrated Household Survey (PIHS) which was carried out in 1991. The ratio as calculated from the DHS data is 8.92 (4q1 for males is 6.4, for females it is 57.1, which seems implausible). The ratio calculated from the PIHS is equal to 1.79. In the calculation of consultation and no treatment the ratios are again implausible when including all children who suffered from diarrhea, ARI, and fever. These ratio are replaced by the ratio including only those who had a sample weight of less than 1. The corresponding changes in the data are as follow: Changes made to Baluchistan female / male ratios for 6 variables Name Raw ratio Ratio using only those observations with weight less Description than I Female relative to male: percent with ARI who were taken for consultation (usually includes hospital, health aritd .713 1.00 center, clinic, doctor, or other health professional) Female relative to male: percent wilh fever who were fevtd .519 .984 taken for consultation Female relative to male: percent with diarrhea who diatd 2.82 .271 were taken for consultation Female relative to male: percent wilh ARI who arind 9.09 2.60 received no treatment Female relative to male percent wilh fever who fevnd 2.32 1.42 received no treatment Female relative to male percent wilh diarrhea who diand 5.09 3.26 received no treatment Consultation and no treatment of Fever or ARI Conditional on a child having suffered from fever or ARI respondents were asked whether the child was taken to a health facility or provider for a consultation. However, not all countries have this number for both fever and ARI. In order to create and "index" of the female/male ratio of consultation for fever or ARI, the data is "filled in" by predicting the ratio for ARI (from a regression of ARI on fever) and using the predicted value when the actual value for ARI is missing and that for fever is not, 30 and vice versa for fever. The "index" is just the mean of these two variables. An equivalent method is used to generate an "index" for the ratio for no treatment of children with fever or ARI. Description of data The list of variables and their detailed descriptions, the number of non-missing values, their means and standard deviations across all observations, are in table Al-1. The list countries, year of the surveys, sample sizes (i.e. the number of women interviewed), and the types of data available by country are given in table Al-2. In addition, the summary statistics for each ofthe indicators in the data are given in Appendix 2, Table A2- 1. Income data The data used for income across different countries are from the Penn World Tables 5.6 (PWT) and the variable used is the real per capita GDP per capita expressed in 1985 international dollars (i.e. these are purchasing power adjusted quantities). For countries which do not have data up until the date of the survey, the data are extrapolated from the last two years for which actual data exist. The income data for states India are derived from Government of India's 1993-94 Economic Surve (Government of India, 1994) which reports state level per capita net State Domestic Product for 1991-92. These are "converted" into 1985 international dollars and scaled for the difference between net State product and GDP, using the conversion implied by the comparison of the (weighted) average Indian net state product to the Indian real GDP per capita from the PWT. The income data for the provinces of Pakistan are derived from household expenditures per capita from the 1991 PIHS. The Province level per capita expenditures are "converted" into 1985 international dollars and scaled for the difference between per capita expenditures and GDP per capita, using the conversion implied by the comparison of the average Pakistani household per capita expenditures to real GDP per capita from the PWT. 31 Table Al-i: Variables Description Name Number Mean Std. of Dev. obser- vations Marriage and fertility Percent who have never married: women aged 15-19 w15 19nn 94 75.01 14.36 Percent who have never married: women aged 20-24 w2024nm 94 33A9 15.23 Median age at first marriage: women 20-49 w2049mm 53 17.61 1.506 Median age at first marriage: women 25 -49 w2549mm 82 18.42 1.993 Percent who have never had a birth womnen 15-19 wlSl9nb 94 84.34 8.763 Percent who have never had a birth: women 20-24 w2024nb 94 40.97 15.003 Median age at first birth : women 20-49 w2049mb 30 19.12 0.553 Median age at first birth: women 25-49 w2549mb 79 20.41 1.342 Education Female relative to male:percent of house hold population in school: ages 6-10 (4 observations are either 6-11, 7-10, or 7-12) enrld 68 0.8908 0.15 Female relative to male:percent of household population in school: ages 11-14 (15) (4 observations are either 12-14, 13-15, or 13-16) enr2d 68 0.8199 0.184 Infant and child mortality Female relative to male: neonatal mortality mnrtnnd 66 0.802 0.114 Female relative to male: post-neonatal mortality mrtpnnd 65 1.0011 0.188 Female relative to male: infant mortality mrtlqOd 88 0.8705 0.109 Female relative to male: child mortality nrt4qld 85 1.1365 0.3 17 Female relative to male: under-five morality mrtSqOd 87 0.9422 0.119 Vaccinations Female relative to male : percent with all vaccinations (i.e. BCG, measles, and three doses of DPT and pol io vaccine) from vaccination card or mother's report: children aged 12 to 23 months (I observation is for 0 to 59 months, I observation is for 12 to 59 months) vacalld 71 0.9679 0.124 Female relative to male : percent with no vaccinations vacnond 61 1.1727 0.493 Continued... 32 Table Al-I continued: Variables Description Name Number Mean Std. of obser- Dev. ________ vations Incidence of illness Female relative to male percent with ARI in the past 2 weeks (6 observations refer to 4 weeks): children under 5 (27 observations refer to under 4, 8 to under 3) arid 80 0.8958 0.145 Female relative to male :percent with fever in the past 2 weeks (6 observations refer to 4 weeks): (27 observations refer to under 4, 6 to under 3) fevd 71 0.9485 0.075 Female relative to male : percent with diarrhea in the past 2 weeks (1 observation refers to 1 week, 1 to 4 weeks) : children under 5 (27 observations refer to under 4, 8 to under 3) diad 89 0.9266 0.132 Consultation and no treatment of illness Female relative to male : percent with ARI who were taken for consultation (usually includes hospital, health center, clinic, doctor, or other health professional) aritd 69 0.9528 0.10 Female relative to male percent with fever who were taken for consultation fevtd 52 0.9436 0.076 Female relative to male percent with diarrhea who were taken for consultation diatd 77 0.9702 0.17 Female relative to male percent with ARI who received no treatment arind 55 1.1588 0.467 Female relative to male percent with fever who received no treatment fevnd 53 1.149 0.372 Female relative to male percent with fever or ARI who received no treatment (this is a constructed variable, see text) f and 60 1.1547 0.316 Female relative to male: percent with diarrhea who received no treatment I diand 70 1.1548 0.381 Anthropometrics Female relative to male percent whose weight-for-age is below 3 standard deviations of the reference population wfa3sdd 75 0.9824 0.208 Female relative to male: percent whose weight-for-age is below 2 standard deviations of the reference population wfa2sdd 79 0.9765 0.113 Female relative to male: percent whose height-for-age is below 3 standard deviations of the reference population hfa3sdd 70 1.0289 0.448 Female relative to male : percent whose height-for-age is below 2 standard deviations of the reference population hfa2sdd 74 0.948 0.073 Female relative to male : percent whose weight-for-height is below 3 standard deviations of the reference population wfh3sdd 68 0.7911 0.352 Female relative to male: percent whose weight-for-height is below 2 standard deviations of the reference population wfh2sdd 74 0.8377 0.206 33 Table Al-2: Data availability cell is marked with an -x- if at least one of the variables in the category is available. Year of Number Marriage Educa- Infant / Vaccina- Incidence Consulta- Anthrop- survey of women and tion Child tions of illness tion / No ometrics Fertility mortality treatment of illness South Asia Bangladesh 1993/94 9640 x x x x x x India 1992/93 89777 x x x x x x x SriLanka 1987 5865 x x x x x Nepal 1987 16231 Pakistan 1990/91 6611 x x x x x x x Non-South Asian countries Burundi 1987 3970 x x x x x BurkinaFaso 1992/93 6354 x x x x x x x Bolivia 1989 7923 x x x x x Bolivia 1993/94 8603 x x x x x x x Brazil 1986 5892 Botswana 1988 4368 x x x x x Central African Rep. 1994/95 6000 x x x x x x x Cote d'Ivoire 1994 8099 x x x x x x x Cameroon 1991 3871 x x x x x x x Colombia 1986 5329 x x x Colombia 1990 8644 x x x x x x Colombia 1995 11140 x x x x x x x Dominican Rep. 1986 7649 x x x x Dominican Rep. 1991 7320 x x x x x x x Ecuador 1987 4713 x x Egypt 1988/89 8911 x x x x x x Egypt 1992 9864 x x x x x x x Egypt 1995 14779 x x x x x x x Ghana 1988 4488 x x x x x x Ghana 1993 4562 x x x x x x x Guatemala 1987 5160 x x x x x Guatemala 1995 12403 x x x x x x x Haiti 1994/95 5709 x x x x x x x Indonesia 1987 11884 x x Indonesia 1991 22909 x x x x x x Indonesia 1994 28168 x x x x x x Jordan 1990 6461 x x x x x x x Kazakstan 1995 3771 x x x x Kenya 1989 7150 x x x x x Kenya 1993 7540 x x x x x x x Continued... 34 Table A1-2 continued: Data availability: cell is marked with an -x- if at least one of the variables in the category is available. Liberia 1986 5239 x x x x x Morocco 1987 5982 x x x x x x Morocco 1992 9256 x x x x x x x Madagascar 1992 6260 x x x x x x x Mexico 1987 -9310 x x x x Mah 1987 3200 x x x x x Mali 1995/96 -9000 x x x x x x x Malawi 1992 4850 x x x x x x x Namibia 1992 5421 x x x x x x x Brazil (NE) 1991 6222 x x x x x x Niger 1992 6503 x x x x x x x Nigeria 1990 8781 x x x x x x x Ondo State, Nigeria 1986/87 4213 x x x x x Peru 1986 4999 x x Peru 1991/92 15882 x x x x x x x Philippines 1993 15029 x x x x x x Paraguay 1990 5827 x x x x x x x Rwanda 1992 6551 x x x x x x x Sudan (Northern) 1989/90 5860 x x x x x Senegal 1986 4415 x x x x x Senegal 1992/93 6310 x x x x x x x El Salvador 1985 5207 x Togo 1988 3360 x x x x x Thailand 1987 6775 x x x x x Trinidad/Tobago 1987 3806 x x x x Tunisia 1988 4184 x x x x x x Turkey 1993 6519 x x x x x x x Tanzania 1991/92 9238 x x x x x x x Uganda 1988/89 4730 Uganda 1995 7070 x x x x x x x Yemen 1991/92 5687 x x x x x x Zambia 1992 7060 x x x x x x x Zimbabwe 1988/89 4201 x x x x x x Zimbabwe 1994 6128 x x x x x x x Continued... 35 Table Al-2 continued: Data availability: celi is marked with an -x- if at lesat one of the variables in the category is available. Year of Number Marriage Educa- Infant / Vaccina- Incidence Consulta- Anthrop- survey of women and tion Child tions of illness tion / No ometrics Fertility mortality treatmnent _____________________ j ________ _ j __________________________________ o f illn ess Pakistan Baluchistan 1990/91 941 x x x x x x x NWFP 1990/91 1665 x x x x x x x Punjab 1990/91 2207 x x x x x x x Sindh . 1990/91 1798 x x x x x x x India _ AndhraPradesh 1992/3 4276 x x x x x x x Arunachal Pradesh 1992 882 x x x x Assam 1992/93 3006 x x x x x x x Bihar 1993 2067 x x x x x x x Delhi 1993 3457 x x x x x x x Goa 1992/93 3141 x x x x x x x Gujarat 1993 3832 x x x x x x x Himachal Pradesh 1992 2962 x x x x x x x Haryana 1993 2846 x x x x x x x Jammu region of J&K 1993 2766 x x x x x x x Karnataka 1992/93 4413 x x x x x x x Kerala 1992/93 4332 x x x x x x x Madhya Pradesh 1992 4283 x x x x x x x Meghalaya 1992/93 1137 x x x x Manipur 1993 953 x x x x Maharashtra 1992/93 4106 x x x x x x x Mizoram 1993 1045 x x x x Nagaland 1993 1149 x x x x Orissa 1993 4257 x x x x x x x Punjab 1993 2995 x x x x x x x Rajasthan 1992/93 5211 x x x x x x x Tamil Nadu 1992 3948 x x x x x x x Tripura 1993 1100 x x x x Uttar Pradesh 1992/93 8722 x x x x x x x West Bengal 1992 1036 x x x x x x x 36 Appendix 2: Table A2-1 mrtmnd : Female/Male : neonatal mortality Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 0.70 0.19 4 0.87 0.82 0.75 0.58 0.43 India 0.83 0.10 19 0.99 0.93 0.82 0.76 0.62 5th Asia 0.79 0.04 4 0.84 0.81 0.78 0.77 0.77 R.O.W. 0.80 0.12 43 1.05 0.87 0.78 0.75 0.54 mrtpmd : Female/Male : post-neonatal mortality Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 1.02 0.20 4 1.20 1.19 1.02 0.84 0.82 India 1.13 0.22 19 1.59 1.24 1.14 0.96 0.68 Sth Asia 1.00 0.10 4 1.13 1.08 0.98 0.93 0.92 R.O.W. 0.93 0.14 42 1.40 0.99 0.94 0.84 0.69 mrt1qOd : Female/Male : infant mortality Mean S.D. N Maximun P75 Median P25 Miniaum Pakistan 0.80 0.15 4 1.00 0.90 0.79 0.71 0.63 India 0.93 0.12 19 1.14 1.07 0.91 0.83 0.77 5th Asia 0.82 0.12 5 0.95 0.87 0.84 0.82 0.63 R.O.W. 0.86 0.09 64 1.16 0.90 0.86 0.81 0.66 mrt4q1d : Female/Male : child mortality Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 1.49 0.54 4 2.07 1.93 1.52 1.06 0.86 India 1.43 0.35 19 2.35 1.63 1.44 1.22 0.80 Sth Asia 1.33 0.25 5 1.66 1.43 1.33 1.24 0.99 R.O.W. 1.00 0.17 61 1.47 1.10 1.01 0.91 0.47 mrt5qOd : Female/Male . under-five mortality Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 1.00 0.15 4 1.17 1.11 1.02 0.90 0.80 India 1.04 0.15 19 1.32 1.15 1.04 0.94 0.79 Sth Asia 0.94 0.14 5 1.06 1.01 0.98 0.95 0.70 R.O.W. 0.91 0.08 63 1.15 0.95 0.92 0.86 0.69 enrid * Female/Male * percent hh population in school * age group 6-10 Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 0.60 0.17 4 0.75 0.74 0.62 0.45 0.40 India 0.89 0.11 25 1.04 0.96 0.92 0.84 0.59 Sth Asia 0.82 0.11 4 0.97 0.89 0.79 0.74 0.72 R.O.W. 0.92 0.14 39 1.14 1.02 0.97 0.87 0.48 enr2d : Female/Male : percent hh population in school : age group 11-14 . ...... ........... .. ... ... ...... ........ ..... ........ ...... .......... ..... ....... Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 0.52 0.19 4 0.70 0.68 0.52 0.36 0.34 India 0.83 0.14 25 1.01 0.92 0.86 0.77 0.49 Sth Asia 0.74 0.13 4 0.93 0.83 0.70 0.65 0.64 R.O.W. 0.85 0.19 39 1.10 0.98 0.90 0.71 0.33 Cont... 37 vacalid : Female/Male : percent with all vaccinations Mean S.D. N Maximun P75 Median P25 Miniuni Pakistan 0.77 0.12 4 0.93 0.85 0.74 0.68 0.67 India 0.91 0.13 19 1.15 1.01 0.91 0.79 0.70 Sth Asia 0.87 0.06 4 0.93 0.91 0.87 0.83 0.80 R.O.W. 1.01 0.09 48 1.32 1.04 1.01 0.96 0.84 vacnond : Female/Male : percent with no vaccinations Mean S.D. N Maximun P75 Median P25 Minimua Pakistan 1.12 0.20 4 1.32 1.27 1.13 0.96 0.88 India 1.33 0.65 19 3.38 1.26 1.16 1.01 0.68 Sth Asia 1.18 0.25 4 1.48 1.35 1.19 1.02 0.87 R.O.W. 1.07 0.41 38 2.75 1.22 1.06 0.84 0.25 arid : FemaLe/MaLe : percent with ari Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 0.90 0.23 4 1.06 1.06 0.99 0.74 0.57 India 0.81 0.18 25 1.15 0.92 0.80 0.73 0.46 5th Asia 0.91 0.11 4 1.03 1.00 0.90 0.81 0.80 R.O.W. 0.94 0.09 51 1.16 1.00 0.95 0.90 0.50 fevd : Female/Male : percent with fever Mean S.D. N Maximun P75 Median P25 Mininmu Pakistan 1.00 0.02 4 1.02 1.02 1.00 0.99 0.98 India 0.91 0.08 25 1.06 0.96 0.91 0.89 0.74 Sth Asia 0.96 0.04 3 0.99 0.99 0.98 0.92 0.92 R.O.U. 0.97 0.07 43 1.17 1.00 0.97 0.94 0.72 diad : Female/Male : percent with diarrhea Mean S.D. N Maximun P75 Median P25 Minimun Pakistan 1.01 0.13 4 1.21 1.08 0.95 0.94 0.93 India 0.92 0.22 25 1.54 0.99 0.88 0.80 0.55 Sth Asia 0.94 0.09 5 1.08 0.95 0.94 0.91 0.83 R.O.W. 0.92 0.07 59 1.09 0.96 0.92 0.87 0.73 aritd : Female/Male : percent wl ari : medicaL consultation Mean S.D. N Maximun P75 Median P25 Minimun Pakistan 0.99 0.11 4 1.15 1.08 0.97 0.91 0.89 India 0.91 0.11 17 1.11 0.99 0.90 0.82 0.72 5th Asia 0.92 0.08 4 1.00 0.98 0.91 0.85 0.84 R.o.W. 0.97 0.10 48 1.16 1.05 0.97 0.90 0.69 fevtd : Female/Male : percent w/ fev : medical consultation Mean S.D. N Maximun P75 Median P25 Minimun Pakistan 0.97 0.08 4 1.08 1.03 0.95 0.91 0.90 India 0.91 0.07 19 1.04 0.96 0.93 0.84 0.76 Sth Asia 0.92 0.03 2 0.94 0.94 0.92 0.90 0.90 R.o.W. 0.96 0.08 27 1.12 1.02 0.98 0.90 0.82 diatd : FemaLe/MaLe:percent w/ dia : medicaL consuLtation Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 0.98 0.56 4 1.65 1.32 1.00 0.64 0.27 India 0.95 0.08 18 1.06 1.01 0.95 0.88 0.84 Sth Asia 1.01 0.15 5 1.24 1.06 0.94 0.91 0.88 R.O.W. 0.98 0.14 54 1.40 1.06 0.99 0.90 0.57 arind : Female/Male : percent w/ ari : no treatment Mean S.D. N Maximun P75 Median P25 Minimun Pakistan 1.45 0.84 4 2.60 2.06 1.27 0.84 0.66 India 1.23 0.50 16 1.96 1.63 1.29 0.76 0.43 Sth Asia 1.16 0.18 3 1.29 1.29 1.23 0.95 0.95 R.O.W. 1.09 0.41 32 2.69 1.19 1.03 0.81 0.58 Cont... 38 fevnd FemaLelNale:percent w/ fev : no treatment …-- - - - - - - - - - - - - - Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 1.18 0.32 4 1.42 1.38 1.28 0.97 0.71 India 1.36 0.36 19 1.92 1.69 1.23 1.17 0.71 5th Asia 1.19 0.06 2 1.24 1.24 1.19 1.15 1.15 R.O.W. 1.00 0.33 28 1.51 1.19 1.05 0.84 0.14 f_and : FemaLe/MaLe : percent fev/ari : no treatment (fitLed) Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 1.31 0.55 4 2.01 1.69 1.28 0.93 0.68 India 1.30 0.29 19 1.79 1.45 1.34 1.19 0.62 5th Asia 1.17 0.11 3 1.27 1.27 1.19 1.05 1.05 R.O.W. 1.06 0.29 34 2.04 1.16 1.02 0.88 0.57 diand : Female/MaLe : percent w/ dia : no treatment Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 1.50 1.21 4 3.26 2.29 1.05 0.70 0.63 India 1.31 0.34 18 1.93 1.59 1.34 1.10 0.78 Sth Asia 1.12 0.22 5 1.45 1.18 1.13 0.95 0.89 R.O.U. 1.05 0.22 47 2.00 1.14 1.02 0.91 0.60 wfa2sdd : FemaLe/Ma(e : percent wefght-for-age below 2 SD of reference medfan mean S.D. N Maximun P75 Median P25 Minimum Pakistan 0.95 0.07 4 1.03 1.00 0.94 0.90 0.88 India 1.01 0.10 25 1.19 1.08 1.00 0.93 0.82 5th Asia 1.01 0.03 4 1.05 1.04 1.01 0.99 0.98 R.O.W. 0.97 0.13 50 1.24 1.04 0.96 0.89 0.64 hfa2sdd : Female/Male : percent height-for-age beLow 2 SD of reference median Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 0.94 0.06 4 1.00 0.98 0.94 0.89 0.86 India 0.98 0.06 20 1.12 1.04 0.98 0.94 0.89 Sth Asia 1.03 0.07 4 1.09 1.08 1.03 0.97 0.96 R.O.W. 0.93 0.08 50 1.11 0.98 0.94 0.88 0.70 wfh2sdd : FemaLe/Male : percent weight-for-height betow 2 SD of reference median Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 0.83 0.24 4 1.08 1.02 0.84 0.64 0.56 India 0.84 0.18 20 1.22 0.97 0.86 0.69 0.57 5th Asia 0.89 0.13 4 1.09 0.97 0.84 0.82 0.80 R.O.W. 0.84 0.22 50 1.50 0.93 0.83 0.75 0.20 Cont... 39 wl5l9nm : Percent never married: women 15-19 Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 71.55 6.79 4 75.40 75.35 39.55 67.75 61.40 India 71.57 15.55 25 -96.90 82.00 25.90 61.30 36.00 Sth Asia 67.00 17.02 5 92.70 75.10 18.50 56.00 50.50 R.O.W. 76.77 13.69 64 95.60 85.95 35.15 71.30 24.60 w2024nm : Percent never married: women 20-24 Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 34.80 10.07 4 40.40 40.00 39.55 29.60 19.70 India 28.59 15.71 25 71.00 35.30 25.90 18.10 8.60 5th Asia 28.44 19.26 5 57.10 39.40 18.50 14.80 12.40 R.O.W. 35.25 14.79 64 69.70 44.15 35.15 24.70 2.00 w2049mm : Median age 1st marriage: women 20-49 Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 0.00 0 India 17.12 1.50 20 19.50 18.30 17.20 15.85 14.70 Sth Asia 16.52 1.84 4 18.90 17.65 16.40 15.40 14.40 R.O.W. 18.02 1.31 32 20.80 18.95 18.30 17.25 15.10 w2549m : Median age 1st marriage: women 25-49 Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 18.27 1.05 4 19.30 19.15 18.35 17.40 17.10 India 17.57 2.06 25 21.70 19.00 17.70 16.00 14.50 Sth Asia 16.33 1.73 4 18.60 17.40 16.15 15.25 14.40 R.O.W. 18.96 1.81 53 24.80 20.50 18.80 17.80 15.10 wl5l9nb : Percent with no birth: women 15-19 Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 87.10 1.64 4 88.30 88.15 87.70 86.05 84.70 India 86.20 7.61 25 98.30 92.40 86.00 79.30 72.10 Sth Asia 83.90 8.83 5 96.40 87.80 81.40 81.30 72.60 R.O.W. 83.34 9.16 64 98.00 91.00 85.35 76.85 55.50 w2024nb : Percent with no birth: women 20-24 Mean S.D. N Maximun P75 Median P25 Minimusn Pakistan 48.62 10.73 4 56.60 54.90 52.55 42.35 32.80 India 42.70 14.79 25 80.90 50.70 42.40 31.20 22.40 Sth Asia 40.80 19.09 5 67.00 54.30 33.50 26.90 22.30 R.O.W. 39.48 14.87 64 71.30 50.80 42.05 27.05 15.00 w2049mb : Median age 1st birth: women 20-49 Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 0.00 0 India 19.24 0.57 11 20.00 19.80 19.20 18.90 18.00 Sth Asia 19.03 1.16 3 19.80 19.80 19.60 17.70 17.70 R.O.W. 19.14 0.45 18 19.70 19.40 19.25 18.90 18.00 w2549mb : Median age 1st birth: women 25-49 Mean S.D. N Maximun P75 Median P25 Minimum Pakistan 20.98 0.62 4 21.60 21.50 21.00 20.45 20.30 India 19.83 1.32 19 23.70 20.20 19.70 18.90 17.90 Sth Asia 20.40 2.43 5 24.00 21.30 19.80 19.40 17.50 R.O.W. 20.60 1.22 55 23.10 21.50 20.80 19.60 18.50 40 Appendix 3: Gender disparity in South Asia: A note on additional regression results' Deon Filmer, Elizabeth M. King, Lant Pritchett In the course of carrying out the work for "Gender disparity in South Asia: Comparisons across and within countries2" we assembled a database which can be used to investigate the correlates of gender disparities. In the paper we reported that income, (or a proxy thereof) is not a good predictor of the degree of gender disparity, within or outside of South Asia. This note summarizes the results of introducing other variables into the regression, and is a companion to the earlier paper. The basic regression results, reported in Table 1, restate our earlier results: Gender disparity is not explained by a proxy for income. Controlling for income, a significant difference in the means is found for the South Asian areas (at least for child mortality and the proportion of ARI and fever episodes that resulted in no treatment). Moreover, for these same outcomes, a fair amount of the cross-area variation is explained by the South Asia dummy variable alone. Can we identify the characteristics which explain the effect of the South Asia dummy variable and the variation within South Asia? As discussed in the paper, many of the theories of the low level of investment in girls rest on the economic role of women, particularly in 'The findings, interpretations, and conclusions expressed in this note are entirely those of the authors. They do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent. The note should not be cited without the permission of the authors. 2 "Gender disparity in South Asia: Comparisons across and within countries," Deon Filmer, Elizabeth M. King, Lant Pritchett, DECRG mimeo, The World Bank. 41 agricultural production. Table 2 reports the results of the regression which includes a set of observable variables related to religion and production in rural and agricultural areas (see Table 3 for summary statistics and Annex 2 for the sources of these data). About 45 percent of the variation in mortality and in enrollment disparities within South Asia is accounted for by this set of variables. Adding these variables, however, does not eliminate the significance of the dummy variable for South Asia in the all-countries regression, implying that there is still something fixed and common to at least certain areas of South Asian countries which accounts for the higher gender disparities in that region. The percent of the population that is Muslim has a significantly negative effect on the school enrollment of girls relative to that of boys. This is the case both across states, provinces, and countries within South Asia, as well as across countries outside of that region. However, this variable does not appear to affect the gender differential in child mortality or in the treatment of fever and ARI. Note however that the percent Muslim does have a significantly positive effect on the level of female mortality, and a strong negative effect on the percent of women who receive treatment in countries outside of South Asia (see Annex Table 1). Controlling for income and the rural population density, the share of the labor force employed in agriculture may serve as an indicator of the level of modernization of the society. With this interpretation, one might expect a relatively more agricultural economy to have larger gender disparities. In fact, the two significant coefficients suggest otherwise. Likewise, a larger rural population per area of land implies higher enrollment rates for girls relative to boys in the all-countries regression. 42 One of the most frequently mentioned hypotheses regarding gender disparities in South Asia involves the difference between wheat and rice production. The hypothesis is that women play a much smaller part in the production of wheat and therefore are less "valuable" to households, or have less bargaining power in households, in areas where wheat is the predominant crop. Our results are consistent with this hypothesis. The share of agricultural land that is harvested with wheat has a significantly positive relationship with the mortality differential, but does not appear to be related to enrollment or treatment. To the extent that wheat may be correlated with unobserved factors which are not controlled for by the other included variables, the results do not allow the inference of a causal link between wheat production and gender disparity. However, including a dummy variable for Indian states and for Pakistani provinces in the South Asian regression does not qualitatively alter the results (see Annex Table 2 for those results). In sum, a handful of variables besides income help predict the variation in gender disparity in child mortality and education in South Asia. From the all-countries regression, however, it appears that there are other strong South Asia fixed effects that are not captured by these variables. How much more can be done to unbundle these effects is severely limited by the availability of additional comparable data at the country and state levels, and the remaining degrees of freedom. 43 TabLe 1: Basic regression on the gender disparity in mortality, enroLLment, and treatment VariabLe Child Mortality (4q1) EnroLled in schooL No treatment for fever or ARI ALL* Non South* ALI* Non South* All* Non South* South Asia South Asia South Asia Asia Asia Asia Income 0.031 0.022 0.117 0.056 0.043 0.133 0.052 0.064 -0.010 (In) C 0.68) ( 0.72) t 0.51) ( 1.41) ( 1.01) ( 1.26) ( 0.65) ( 0.78) (-0.05) Dummy for 0.405 . . -0.067 . . 0.230 5th Asia ( 7.04) ( .) ( .) (-1.58) t .) ( .) ( 2.74) ( .) ( .) Cons 0.785 0.847 0.568 0.449 0.543 -0.174 0.690 0.601 1.364 ( 2.39) ( 3.79) ( 0.34) ( 1.56) ( 1.76) (-0.23) ( 1.23) ( 1.04) ( 0.89) R-sq .3891 .0096 .0107 .0639 .0292 .0518 .1491 .0207 .0001 A R-sq .3736 -.0087 -.0305 .0346 .0006 .0191 .1164 -.0131 -.0454 Num. Obs. 82 56 26 67 36 31 55 31 24 t-statistics in parentheses * Includes regional subgroupings within India and Pakistan but excludes observations at the nationaL LeveL for those countries. 44 Table 2: Regression on the gender disparity in mortaLity, enroLlment, and treatment VariabLe Child Mortality (4q1) EnroLLed in school No treatment for fever or ARI All* Non South* ALL* Non South* AlL* Non South* South Asia South Asia South Asia Asia Asia Asia Income -0.070 -0.021 -0.122 0.117 0.114 0.203 0.006 -0.025 -0.128 (ln) (-1.34) (-0.30) (-0.43) ( 2.83) t 1.46) t 1.67) ( 0.06) (-0.12) (-0.36) Dummy for 0.240 . . -0.156 . . 0.279 Sth Asia ( 3.31) t .) ( .) (-3.22) C .) C .) (2.17) ( .) ( .) Rur. Pop. -0.013 0.010 -0.039 0.021 0.026 0.007 -0.012 0.080 -0.030 Ag. land (-1.09) ( 0.57) (-1.38) ( 2.43) ( 1.35) C 0.58) (-0.57) ( 1.49) (-0.83) Labor frce -0.171 -0.031 -0.077 0.177 0.316 0.148 -0.228 -0.801 -0.043 in agric (-1.31) (-0.15) (-0.25) ( 2.01) ( 1.55) ( 1.20) (-0.98) (-1.56) (-0.11) Share 0.021 -0.004 0.540 -0.289 -0.434 0.098 0.016 0.240 -0.002 Muslim C 0.34) (-0.06) ( 1.18) (-5.72) (-5.47) ( 0.49) ( 0.12) ( 1.12) ( 0.00) Share Ag 1.280 1.013 1.409 -0.127 0.561 -0.145 0.197 -2.410 0.405 Wheat ( 5.84) ( 1.76) ( 3.24) (-0.84) ( 0.88) (-0.77) ( 0.53) (-1.30) ( 0.74) Share Ag 0.159 -0.036 0.053 0.077 1.016 0.048 -0.236 -1.614 -0.345 Rice ( 1.11) (-0.12) ( 0.17) ( 0.80) ( 2.28) ( 0.40) (-0.97) (-1.16) (-0.88) Share Ag 0.060 0.102 -1.323 -0.028 -0.667 -0.104 0.507 0.519 0.121 Maize ( 0.19) ( 0.29) (-1.01) (-0.12) (-1.97) (-0.20) ( 0.86) ( 0.50) ( 0.06) Dummy for . . -0.422 . . -0.262 . . -0.481 Pakistan ( .) ( .) (-0.81) C .) ( .) (-1.28) ( .) ( .) (-0.74) Dummy for . . 0.118 . . 0.141 . . -0.355 India C .) C .) (0.46) ( .) ( .) (1.00) ( .) ( .) (-0.42) Cons 1.581 1.131 2.176 -0.027 -0.058 -0.879 1.146 1.648 2.750 ( 3.68) ( 1.85) ( 1.02) (-0.08) (-0.09) (-0.97) (1.35) ( 0.90) ( 0.98) R-sq .6227 .2591 .5099 .4639 .6105 .5074 .1982 .2461 .1762 A R-sq .5796 .1487 .1949 .3859 .5095 .2741 .0490 .0063 -.4416 Nun. Obs. 79 55 24 64 35 29 52 30 22 t-statistics in parentheses * IncLudes regional subgroupings within India and Pakistan but excludes observations at the national Level for those countries. 45 Table 3: Summary statistics: Mean (standard deviation) All* Non South Asia South Asia* Child mortality: gender differential 1.137 (.312) 1.005 (.157) 1.438 (.368) female level 49.12 (40.2) 55.47 (45.4) 33.93 (16.3) Enrolled in school: gender differential .8162 (.184) .8485 (.173) .7773 (.179) female level 61.23 (21.3) 60.47 (22.4) 62.15 (20.2) No treatment for Fever or ARI: gender differential 1.157 (.326) 1.058 (.285) 1.292 (.337) female level 18.06 (10.5) 17.83 (11.9) 18.37 (8.49) Income: natural log 7.230 (.612) 7.240 (.710) 7.213 (.311) level 1670.5 (1179) 1786.5 (1371) 1425.8 (491.3) Dummy for South Asia .3107 (.465) Rural pop. per agric. land 2.198 (2.69) 1.208 (1.81) 4.397 (3.04) Share of labor force in agric. .5564 (.249) .5606 (.241) .5471 (.270) Share muslim .2842 (.373) .2999 (.383) .2494 (.352) Share of agric. land: wheat .0565 (.123) .0203 (.056) .1396 (.183) Share of agric. land: rice .1288 (.242) .0321 (.082) .3434 (.329) Share of agric. land: maize .0509 (.073) .0518 (.076) .0488 (.068) Dummy for Pakistan .1250 (.336) Dummy for India .7188 (.457) Avg years of school of women 15 and over (zero if missing) 2.459 (1.05) 2.632 (1.81) 2.076 (2.48) Dummy for avg years of school of women 15 and over missing .0777 (.269) .1127 (.318) 0 Gini coefficient (zero if missing) .3305 (.186) .3754 (.189) .2308 (.138) Dummy for Gini coefficient missing .1942 (.397) .1690 (.377) .2500 (.440) Number of observations (non dependent variables) 103 71 32 * Includes regional subgroupings within India and Pakistan but excludes observations at the national level for those countries. 46 Annex 1 Annex I Table 1: Regression on the female Level of mortaLity, enrollment, and treatment Variable Child Mortality (4q1) Enrolled in school No treatment for fever or ARI AWl* Non South* ALL* Non South* All* Non South* South Asia South Asia South Asia Asia Asia Asia Lrgdpch -46.800 -33.929 -36.055 18.041 9.217 18.016 -6.189 3.198 -6.862 (-7.01) (-2.60) (-3.02) ( 3.91) ( 1.00) ( 1.37) (-1.98) ( 0.42) (-1.24) sasia -22.789 . . -5.535 . . 1.270 (-2.33) ( .) ( .) (-1.03) ( .) ( .) (0.34) ( .) ( .) rpoppag -0.417 2.014 -1.434 2.348 1.513 1.444 0.265 1.935 -0.375 (-0.27) ( 0.63) (-1.19) ( 2.44) ( 0.66) ( 1.11) ( 0.43) t 1.03) (-0.68) Lfag 11.549 31.602 -2.621 11.591 0.679 20.896 -1.192 21.645 -7.028 ( 0.70) ( 0.85) (-0.20) ( 1.18) ( 0.03) ( 1.56) (-0.18) ( 1.21) (-1.17) shnusl 18.314 47.465 20.171 -37.518 -55.329 -0.792 13.485 19.872 -5.256 X 2.22) ( 3.72) ( 1.04) (-6.67) (-5.95) (-0.04) ( 3.74) ( 2.67) (-0.35) swheat 25.702 -231.773 53.327 -17.198 117.224 -19.548 -12.442 -40.549 -14.132 ( 0.91) (-2.18) ( 2.88) (-1.02) ( 1.57) (-0.95) (-1.18) (-0.63) (-1.66) srice 7.306 -99.172 -1.479 -0.610 109.203 0.017 6.941 -48.701 1.654 ( 0.40) (-1.76) (-0.11) (-0.06) ( 2.08) ( 0.00) ( 1.00) (-1.01) ( 0.27) smaize -59.917 37.358 -94.532 27.427 -46.825 31.386 -24.923 -0.222 -14.474 c(-1.46) ( 0.58) (-1.70) ( 1.08) (-1.18) ( 0.57) (-1.48) (-0.01) (-0.47) dumpak . . -47.355 . . -11.690 . . -17.264 ( .) ( .) (-2.15) ( .) ( .) (-0.53) ( .) ( .) (-1.72) dunind . . -21.520 . . 25.616 . . -24.926 - .) c .) (-2.00) ( .) C .) ( 1.67) ( .) ( .) (-1.88) cons 382.988 270.203 315.717 -67.708 6.301 -101.853 58.588 -25.023 100.176 * 6.98) C 2.38) ( 3.52) (-1.86) C 0.08) (-1.04) ( 2.41) (-0.40) C 2.29) t-statistics in parentheses * Includes regional subgroupings within India and Pakistan but excludes observations at the national Level for those countries. 47 Amex I TabLe 2: Regression on the gender differentiaL in mortality, enrollment, and treatment: IncLuding average years of education of women 15 and over Variable Child Mortality (4ql) EnroLled in school No treatment for fever or ARI ALl* Non South* AIL* Non South* Atl* Non South* South Asia South Asia South Asia Asia Asia Asia --....--.-...------.--------------.-------------.---. ---....------- ...----- ..................... lrgdpch 0.019 -0.044 -0.411 .0.027 0.081 0.157 0.192 -0.124 -0.209 ( 0.25) (-0.54) (-1.18) ( 0.52) ( 0.92) C 2.11) ( 1.26) (-0.43) (-0.51) yrs15 -0.032 -0.003 -0.012 0.030 0.034 0.026 -0.070 -0.046 -0.051 (-1.63) (-0.12) (-0.22) ( 2.36) C 1.13) ( 1.96) (-1.78) (-0.46) (-0.72) yrsl5m -0.095 -0.011 . 0.026 0.068 . -0.141 -0.134 (-1.04) (-0.14) ( .) C 0.36) ( 0.69) C .) (-0.80) (-0.56) ( .) gini 0.068 0.581 -6.454 0.293 0.340 2.641 -0.157 1.254 -7.598 ( 0.18) ( 1.69) (-2.10) ( 1.02) ( 0.92) ( 3.28) (-0.19) ( 1.16) (-2.02) .................................................................................................. ginimis 0.143 0.371 -1.537 0.113 0.094 0.947 -0.109 0.549 -2.331 ( 0.85) ( 2.30) (-1.66) ( 0.97) ( 0.54) ( 3.79) (-0.30) ( 1.09) (-1.96) sasia 0.232 . . -0.124 . . 0.193 C 2.82) ( .) ( .) (-2.27) ( .) ( .) (1.24) t .) ( .) rpoppag -0.004 0.022 -0.046 0.009 0.042 -0.018 0.023 0.097 0.013 (-0.25) t 1.16) (-1.04) ( 0.85) ( 1.76) (-1.99) ( 0.79) ( 1.48) ( 0.24) lfag -0.152 -0.151 0.035 0.117 0.406 0.018 -0.062 -1.091 0.258 (-1.11) (-0.71) ( 0.12) ( 1.34) ( 1.67) ( 0.23) (-0.25) (-1.83) ( 0.69) shmusl -0.001 0.023 0.980 -0.222 -0.355 0.209 -0.125 0.194 0.172 (-0.02) ( 0.29) ( 2.13) (-4.09) (-3.66) ( 1.70) (-0.85) ( 0.63) C 0.17) swheat 1.146 1.120 1.458 0.055 0.346 0.146 -0.108 -1.896 0.190 ( 4.90) ( 1.97) ( 3.21) ( 0.35) ( 0.50) ( 1.23) (-0.27) C-0.94) ( 0.35) ....... ................... .. ............. ....................... ................................................ srice 0.100 0.188 -0.147 0.187 0.656 0.205 -0.530 -0.697 -0.868 ( 0.63) ( 0.57) (-0.39) ( 1.84) ( 1.11) ( 2.33) (-1.85) (-0.38) (-1.95) smaize -0.032 -0.205 -3.547 0.041 -0.679 0.294 0.597 0.028 -1.815 (-0.10) (-0.56) (-2.42) ( 0.18) (-1.75) ( 0.86) ( 0.99) C 0.02) (-0.81) dwipak . . -0.895 . . -0.345 . . -0.878 .) ( .) (-1.77) ( .) ( .) (-2.74) C .) C .) (-1.46) dumind . . -0.096 . . 0.123 . . -0.633 ( .) ( .) (-0.38) ( .) ( .) (1.43) C .) ( .) (-0.74) -cons 0.975 1.080 6.458 0.432 -0.134 -1.383 0.006 2.082 5.922 ( 1.78) ( 1.73) ( 2.02) ( 1.10) (-0.18) (-2.06) ( 0.00) C 1.00) C 1.50) t-statistics in parentheses * Includes regional subgroupings within India and Pakistan but excludes observations at the nationat level for those countries. 48 Amex 1 Table 3: Regression on the femate level of mortality, enrollment, and treatment: Including average years of education of women 15 and over …----- Variable Child Mortality (4q1) Enrolled in school No treatment for fever or ARI …---- All* Non South* ALL* Non South* ALl* Non South* South Asia South Asia South Asia Asia Asia Asia lrgdpch -40.110 -32.635 -36.768 4.466 0.187 10.396 0.105 3.967 -3.481 (-4.22) (-2.05) (-2.21) ( 0.82) ( 0.02) ( 1.29) ( 0.02) ( 0.40) (-0.45) yrs15 -1.447 1.737 -2.636 4.331 3.714 3.834 -2.103 2.146 -1.279 (-0.58) ( 0.34) (-1.02) ( 3.25) ( 1.13) ( 2.69) (-1.90) ( 0.62) (-0.94) yrsi5m -5.519 3.667 . -2.567 -4.094 . 0.923 9.933 (-0.47) ( 0.23) C .) (-0.34) (-0.38) C .) ( 0.19) ( 1.21) C .) gini -7.452 2.496 -111.054 37.165 30.643 227.012 -2.249 -4.391 -27.341 (-0.16) ( 0.04) (-0.69) ( 1.24) ( 0.75) ( 2.59) (-0.10) (-0.12) (-0.38) ginimis 16.542 15.907 -16.003 12.891 4.670 83.008 -1.472 -4.387 -11.621 ( 0.77) ( 0.50) (-0.34) ( 1.06) ( 0.24) t 3.05) (-0.14) (-0.25) (-0.51) sasia -22.562 . . -2.026 . . -0.076 (-2.07) ( .) ( .) (-0.36) C .) ( .) (-0.02) C .) ( .) rpoppag -0.534 1.869 -0.772 0.511 2.319 -1.731 1.265 2.383 0.606 (-0.29) ( 0.51) (-0.37) ( 0.49) ( 0.88) (-1.72) ( 1.57) ( 1.05) ( 0.59) Lfag 9.390 35.150 2.654 2.776 -0.333 4.564 3.669 29.412 -2.545 ( 0.54) ( 0.85) ( 0.19) ( 0.30) (-0.01) ( 0.54) ( 0.52) ( 1.42) (-0.36) shmusl 18.865 49.016 31.473 -27.937 -45.128 12.066 9.241 24.489 -7.768 ( 2.02) ( 3.17) ( 1.39) (-4.93) (-4.24) ( 0.90) ( 2.22) ( 2.31) (-0.40) swheat 17.549 -211.056 44.523 8.787 108.720 14.406 -22.443 -61.639 -19.692 ( 0.58) (-1.89) ( 2.08) ( 0.53) ( 1.43) ( 1.12) (-1.96) (-0.88) (-1.93) srice 9.612 -95.359 -15.518 14.977 72.712 20.218 -1.314 -81.223 -5.983 ( 0.47) (-1.48) (-0.86) ( 1.41) ( 1.12) ( 2.11) (-0.16) (-1.29) (-0.71) smaize -65.208 23.981 -164.143 34.197 -48.647 74.716 -18.702 10.478 -19.816 (-1.56) ( 0.34) (-2.15) ( 1.45) (-1.14) ( 2.02) (-1.09) ( 0.24) (-0.47) dumpak . . -58.413 . . -22.191 . -17.659 ( ) ( ) ~~(-2.44) ( )( ) (1.62) ) ) ( 1.55) dumind . . -26.388 . - 20.775 m -25.881 ( ) ( .) (-2.18) ( . ( .) ( 2.21) ( . ( .) (-1.60) cons 340.377 250.069 364.100 6.841 48.438 -117.312 17.246 -40.488 85.056 ( 4.85) ( 2.05) ( 2.32) ( 0.17) ( 0.58) (-1.61) ( 0.54) (-0.56) ( 1.13) t-statistics in parentheses * Includes regional subgroupings within India and Pakistan but excludes observations at the national level for those countries. 49 Annex 2: Data sources Annex 2 Table 1: Source (year of data) l Variable Country level Indian states Pakistani provinces Income PWT56 I ES (1990/91) PIHS (1990/91) Percent of population rural XWB SID I SA (1991) P AS (1981) Agricultural land FAO 1 1I SA (1990/91) P AS (1992/93) Share of labor force in agriculture WB SID I LYB (1991) P ES (1991/92) Share muslim I SA (1981) T I Area harvested with wheat FAO 2 1I SA (1990/91) P AS (1992/93) Area harvested with rice FAO 2 I SA (1990/91) P AS (1992/93) Area harvested with maize FAO 2 I SA (1990/91) P AS (1992/93) Average years of schooling of BL NFHS (1992193) DHS (1990) women 15 and over Gin coefficient DS ODR (1993) M (1987/88) PWT56: Penn World Tables Mark 5.6 WB SID: World Bank Social Indicators of Development 1997 FAO 1: FAOSTAT web site "http://apps.fao.orgf' as of 9/1196 FAO 2: FAO statistics as reported in the World Bank's BESD system BL: Barro-Lee education data I ES: Economic Survey, Government of India, 1993/94, "converted" to 1985 international dollars I SA: Statistical Abstract, India 1992 I LYB: Indian Labor Year Book, 1993 NFHS: National Family Health Survey (similar to DHS) PIHS: Pakistan Integrated Household Survey, 1990/91, "converted" to 1985 international dollars P ES: Economic Survey, Government of Pakistan, 1993-94 P AS: Agricultural Statistics of Pakistan 1992-93 DHS: Demographic and Health Survey DS: Deininger and Squire (1996) ODR: Ozler, Datt, and Ravaillon (1996) M: Malik (1996) in Lipton and Van der Gaag (eds) 50 Figure 1A FEMALE RELATIVE TO MALE CHILD MORTALITY mrt4ql d LIZ] co0.886 [m] 0.887 to 1.062 1.0631o 1.448 This map was producod by te Map Design Unit of The Word Bank. The baondori, e.colars, denminotions and anyoAer information nhown I 51.449 iudgmetonth leolsttu d any ritory,ora nyno or L J NO DATA ocepWnen of such boundories. I~~ ATS>>LA NTaE D.IC PACIFIC ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ PAII OCEA N ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ CA ; . >- (~~~A t A PA T I C OCEAN ' '0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1 10~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 Figure 1B IBRD 29033 FEMALE RELATIVE TO MALE CHILD MORTALITY mrtAgld [ <.0.886 0.887 to 1.062 1.063 to 14.48 c72br *- * * J _ >s1 ~~~~~~~~IA49 I ] **;-NO DATA _S e- a I .. - - o, Mi o I -^ f ........... ::' .::- ::.. ____ _ ~~~~~~~~~~.__.-........... B n g' This map wm pro&umd by Met Map Design unit of The Wand Book. The boundarie, cao6-s,denonratios and cnyouber nianaro. sh.w on thJs ta do not imply, on tip port of The World Bonk Group, ony judgent ath legl s otw twian y , oerha~a any endorsement or OCTOBER 1997 Figure lc Plot of mrt4qld: child mortality 2.5 - hya pnj z1) ~2 1.93_ ___ to bic pjb 0- a ~~~pak o .2_ __ 1.63_- __ a 2 ~~~~~1.52- .. ~ 1.5 1.44_§C_--- 1.43.ind egy > ~~~~~~~~~~~~~~~~~~~1.33..bgd -- 4-, n .2-p 4-J sd1.22-Mfl8 ---- 13_g__ 1.06 - ---- @8ti~ __ C- krl lka a) nfp.9_ __ r-o tan E 'L .5- I I I I I . I IOI Pakistan India 5th Asia R.O.W. Figure 2A FEMALE RELATIVE TO MALE NO TREATMENT FOR FEVER / ARI f and 1.28_ _ 1.347_nd ~~~~~~~~~n I 19 1 .19-bgd ---- 1.16--i-__ a) 1.05_pak____ 1.02 ---- X_ . 193- ~ ~ t1.8 l__ sndpo E knk a) pry LL 5 Pakistan India Sth Asia R.O.W. Figure 3A FEMALE RELATIVE TO MALE ENROLLED IN SCHOOL (AGES 1 1-15) enr2d _ c-0.642 0.643 to 0.877 0.878 to 0.986 This maop was produced by the Map Design Unit of The World sank. The boundaries, colors, denominations and any other InFormation shown .0.987 an this map do not inpIy , on the part of The Worild Bonk Group, any [II] NO DATA j~~~~~~~~~~iudg men t on the lea ttu fay lerrtap o r any e ndors ement or [ NO DATA occtance of such boundaries. _ ~~~~~~~~~~~~~~~~~~~PACIFIC )~~~~~~~~~~~~~~~~~~~~~~~~~~ C E A N j A T L A N T I N i 0j)i _.OCEAN N PACIFIC OCEA N Figure 3B IBRD 29035 FEMALE RELATIVE TO MALE ENROLLED IN SCHOOL (AGES 11-15) ~~ - S', enr2d ~~~ ~ ~ ~ ~ ~ =O~~~-.642 '4 ~~~~~f O~~~.64 to 0.877 A r a b iaB n ga 0~~~~~ This map was produced by the Map Design Unit of The World Bonk. The boundaries, colors, denominations and anyother information shown on this map do no imply, on the part of The WorS Bank Group, ony jument on he Igstotus of ony territory, or any endorsement or occeptance of such boundaries. OCTOBER 1997 Figure 3c Plot of enr2d enrolled in school (+/-) 11-15 2.5 - 2- 0 1.5 - *4, wL 12- bgd cliF E eq ,~~~~~~~77_ -___ in 02-- .71 _"-- E ~~~.68- --- an LL .5.52- -- raj O .36O ... Pak I i i I . l l l Pakistan India Sth Asia R.O.W. Policy Research Working Paper Series Contact Title Author Date for paper WPS1842 Motorization and the Provision of Gregory K. Ingram November 1997 J. Ponchamni Roads in Countries and Cities Zhi Liu 31052 WPS1843 Externalities and Bailouts: Hard and David E. Wildasin November 1997 C. Bernardo Soft Budget Constraints in 37699 Intergovernmental Fiscal Relations WPS1844 Child Labor and Schooling in Ghana Sudharshan Canagarajah November 1997 B. Casely-Hayford Harold Coulombe 34672 WPS1845 Employment, Labor Markets, and Sudharshan Canagarajah November 1997 B. Casely-Hayford Poverty in Ghana: A Study of Dipak Mazumdar 34672 Changes during Economic Decline and Recovery WPS1846 Africa's Role in Multilateral Trade Zhen Kun Wang November 1997 J. Ngaine Negotiations L. Alan Winters 37947 WPS1847 Outsiders and Regional Trade Anju Gupta November 1997 J. Ngaine Agreements among Small Countries: Maurice Schiff 37947 The Case of Regional Markets WPS1848 Regional Integration and Commodity Valeria De Bonis November 1997 J. Ngaine Tax Harmonization 37947 WPS1 849 Regional Integration and Factor Valeria De Bonis November 1997 J. Ngaine Income Taxation 37947 WPS1 850 Determinants of Intra-Industry Trade Chonira Aturupane November 1997 J. Ngaine between East and West Europe Simeon Djankov 37947 Bernard Hoekman WPS1851 Transportation Infrastructure Eric W. Bond November 1997 J. Ngaine Investments and Regional Trade 37947 Liberalization WPS1852 Leading Indicators of Currency Graciela Kaminsky November 1997 S. Lizondo Crises Saul Lizondo 85431 Carmen M. Reinhart WPS1853 Pension Reform and Private Pension Monika Queisser November 1997 P. Infante Funds in Peru and Colombia 37642 WPS1 854 Regulatory Tradeoffs in Designing Claude Crampes November 1997 A. Estache Concession Contracts for Antonio Estache 81442 Infrastructure Networks WPS1 855 Stabilization, Adjustment, and Cevdet Denizer November 1997 E. Khine Growth Prospects in Transition 37471 Economies Policy Research Working Paper Series Contact Title Author Date for paper WPS1856 Surviving Success: Policy Reform Susmita Dasgupta November 1997 S. Dasgupta and the Future of Industrial Hua Wang 32679 Pollution in China David Wheeler WPS1857 Leasing to Support Small Businesses Joselito Gallardo December 1997 R. Garner and Microenterprises 37664 WPS1 858 Banking on the Poor? Branch Martin Ravallion December 1997 P. Sader Placement and Nonfarm Rural Quentin Wodon 33902 Development in Bangladesh WPS1859 Lessons from Sao Paulo's Jorge Rebelo December 1997 A. Turner Metropolitan Busway Concessions Pedro Benvenuto 30933 Program WPS1860 The Health Effects of Air Pollution Maureen L. Cropper December 1997 A. Maranon in Delhi, India Nathalie B. Simon 39074 Anna Alberini P. K. Sharma WPS1861 Infrastructure Project Finance and Mansoor Dailami December 1997 M. Dailami Capital Flows: A New Perspective Danny Leipziger 32130 WPS1862 Spatial Poverty Traps? Jyotsna Jalan December 1997 P. Sader Martin Ravallion 33902 WPS1 863 Are the Poor Less Well-Insured? Jyotsna Jalan December 1997 P. Sader Evidence on Vulnerability to Income Martin Ravallion 33902 Risk in Rural China WPS1864 Child Mortality and Public Spending Deon Filmer December 1997 S. Fallon on Health: How Much Does Money Lant Pritchett 38009 Mafter? WPS1865 Pension Reform in Latin America: Sri-Ram Aiyer December 1997 P. Lee Quick Fixes or Sustainable Reform? 37805 WPS1866 Circumstance and Choice: The Role Martha de Melo December 1997 C. Bernardo of Initial Conditions and Policies in Cevdet Denizer 31148 Transition Economies Alan Gelb Stoyan Tenev