POLICY RESEARCH WORKING PAPER 1864 Child Mortality and Public Rough!y 95 percent ofcross- national variation in child or Spending on Health infant mortality can be explained by a country's per capita income, the How Much Does Money Matter ? distribution of income, the extent of women's education, Deon Filmer the level of ethnic Lant Pritchett fragmentation, and the predominant religion. Public spending on health has relativefy fittle impact. The World Bank Development Research Group December 1997 POLICY RESEARCH WORKING PAPER 1864 Summary findings Filmer and Pritchett use cross-national data to examine than one-tenth of 1 percent of the observed differences the impact on child (under 5) and infant mortality of in mortality across countries. both nonhealth (economic, cultural, and educational) The estimates imply that for a developing country at factors and public spending on health. They come up average income levels, actual public spending per child with two striking findings: death averted is $50,000 to $100,000. This contrasts * Roughly 95 percent of cross-national variation in markedly with a typical range of estimates for the cost- mortality can be explained by a country's per capita effectiveness of medical interventions to avert the main income, the distribution of income, the extent of causes of child mortality of $10 to $4,000. women's education, the level of ethnic fragmentation, They outline three possible explanations for this and the predominant religion. divergence between the actual and apparent potential of * Public spending on health has relatively little impact, public spending: the allocation of public spending, the with a coefficient that is numerically small and net impact of additional public supply, and public sector statistically insignificant at conventional levels. efficacy. Independent variations in public spending explain less This paper-a product of the Development Research Group-is part of a larger effort in the group to investigate the impact of health sector policies. The study was funded by the Bank's Research Support Budget under the research Project "Primary Health Care: A Critical Evaluation" (RPO 680-29). Copies of this paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Sheila Fallon, room MC3-63 8, telephone 202-473-8009, fax 202-522- 1153, Internet address sfallon@worldbank.org. December 1997. (41 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 be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Produced by the Policy Research Dissemination Center Child Mortality and Public Spending on Health: How Much Does Money Matter? Deon Filmer The World Bank Lant Pritchett The World Bank The authors can be reached at dfilmereworldbank.org and Ipritchett~worldbank.org Child Mortality and Public Spending on Health: How Much Does Money Matter?' In 1995 over 9 million children under five in developing countries died avoidable deaths. This staggering figure is more than the entire population of Sweden or of Zambia.2 The cumulative human suffering in the individual and familial tragedies behind these statistics is overwhelming and creates a powerfiul impetus to action, to do something. This is a laudable impulse but if the desire to do something is allowed to be the enemy of good policy the result can be wasted, and possibly counter-productive, efforts. In this paper we examine cross national differences in the widest and best measured indicators of health status: child (under-5) and infant mortality. We establish two major points about the cross national relationship between health status and public spending on health.3 First, the differences across countries in infant and child mortality are overwhelmingly explained by economic and social factors, that is, "development" broadly taken. The finding that I We would like to thank Nancy Birdsall, Jeffrey Hammer, Maureen Lewis, Samuel Lieberman, and Martin Ravaillon for helpful discussions. The findings, interpretations, and conclusions expressed in this paper 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 paper should not be cited without the permission of the authors. 2 "Avoidable" deaths are defined as the excess of the average death rate for the 0-5 age group in the low- and middle-income countries of 88 per 1000 versus the level in the high- income countries, 9. Using a similar approach Gwatkin (1980) calculated the total number of deaths of under fives to be about 15 million. I We use the perhaps awkward phrase "public spending on health" throughout to avoid the ambiguity in the phrase "public health expenditures" which could mean either "expenditures on those items classified as 'public health"' or "all health expenditures by the public sector." 2 "development" is strongly associated with improvements in mortality is neither surprising nor new (Caldwell, 1986, World Bank 1993). What is perhaps surprising is the strength of the relationship as essentially all (95 percent) of the cross national variations in either under-5 or infant mortality can be explained by five factors: the level of income and its distribution, the extent of female education, the extent of ethnolinguistic differences within a country, and whether it is predominately Muslim. While there are poor countries with exceptionally good health status, properly accounting for determninants besides income reduces the unexplained differences in health outcomes and leaves little to be explained by independent variations in health policy. Second, there is an enormous gap between the apparent potential of public spending to improve health status and the actual performance. Reviews of the cost effectiveness of preventive and primary curative interventions suggest that a significant fraction of under five deaths could be avoided for as little as SI 0, and in many cases under $1000, per death averted (Jamison and others, 1993). However, differences in public spending on health account for essentially none (0.15 percent) of the cross-national differences in health status. The extremely small actual impact of public spending that we estimate from the cross-national data implies that the typical public spending on health per child death averted is $50,000 to $100,000, a striking discrepancy between the apparent potential and actual performance. Why is public spending on health ineffective at improving health status even though relatively cheap and effective medical interventions exist? Addressing that complicated question in detail is left to a companion paper (Filmer, Hammer, and Pritchett, 1997) and here we only suggest three likely explanations: (1) cross-national differences in the efficacy of the public 3 sector mean that public spending on health does not always translate into a larger supply of effective health services, (2) the impact of a greater supply of effective health services in the public sector on health status depends on individual demand and market supply, and (3) public monies are spent on expensive, but ineffective, curative services. I) Explaining cross-national variation in health status Much of the intuitive appeal behind many proposed strategies to improve health status, such as Primary Health Care (PHC) or a "basic package" of cost-effective services, comes from the simple but powerful observation that there are countries with exceptionally good health status for their level of income (World Bank, 1 997a). The relatively good health of Sri Lanka, China, Costa Rica, and Kerala, India is frequently cited as an indication of the potential benefits from PHC.4 However, it is impossible to jump from some countries' good health outcomes to the conclusion that all (or even that any) of the unexplained differences in mortality are due to health policy. While it is possible that these countries' good health outcomes is due to health sector strategy, it is equally plausible that they share non-health characteristics like high levels of female education (King and Hill, 1992), better nutrition, more equal income distribution (Bidani and Ravallion, 1997) that explain their better outcomes. To assess the maximum amount of the variation in health outcomes that can be explained by independent variations in health sector expenditures or policy, and so properly identify the "outliers," we estimate a multivariate regression that explains health outcomes without any ' These countries (regions) were highlighted at a conference sponsored by the Rockefeller Foundation (Halstead et al, 1985). 4 health sector variables: Health Status, =f(Incomej, Female Education,, Income Distribution;, Z,) where Z refers to a vector of additional country specific non-health related variables. At this stage we resolve the problem of attribution of effects in a multivariate regression by assuming that our non-health factors may cause better health policy but that better health policy does not independently affect the included factors. That is, we assume more female education might lead to better health policy but not vice versa. This means we attribute to female education, for example, both whatever health improvements it may cause directly through behavioral changes and whatever indirect impact it may have through better health policy. This assumption about the causal ordering of health policy and non-health factors is seems reasonable for nearly all of the variables and we return to the only problematic case of the joint determination of health status and income below.5 The child mortality figures used are from a UNICEF publication (UNICEF, 1992).6 While there are difficulties with measuring child mortality, it is arguably superior to alternative measures. Life expectancy is not reliably measured in many countries and many of the figures reported in official sources are not based actual data, but are extrapolations from child mortality and assumed life tables. More comprehensive measures of health status that go beyond mortality 5 While there are a number of difficult details about the data and the empirical estimation of this equation we will not deal in detail with them here, as its main purpose is to place a (possibly quite strict) upper bound on the magnitude of the health variation across countries that can be attributed to differences in health care strategies (see appendix 1 for the details on the data). 6Based on background work done by demographer Ken Hill (see United Nations, 1992). S (such as QALYs or DALYs) are even less solidly based for cross-national comparisons. Infant mortality is perhaps more reliably measured, but fails to capture mortality from many of the health conditions of concem which are responsive to health care, such as diarrhea and respiratory infections. Especially at moderate to low levels, infant mortality is dominated by perinatal mortality. The first empirical result is the variation in mortality associated with income. Column I of Table 1 shows that a large part of the variation in (the natural log of) under-5 mortality can be "explained" by (the natural log of) GDP per capita and a set of region dummy variables only.7' Even excluding the regional dummy variables, 84 percent of mortality differences can be "explained" by income alone. The second, even more striking, result in column 2 of Table 1, is that when a few other variables are included over 94 percent of the variation in the under-5 mortality rate is explained. The level of female education, an estimate of the income distribution, a binary variable for whether the population of the country is predominantly Muslim, and the degree of "ethnolinguistic fractionalization" are each statistically significant and of plausible sign and magnitude.9 ' Dummy variables are based on World Bank regions: East Asia and Pacific, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa, and (excluded from the regression) "Rest of World." 8 In order to ensure the robustness of these, and subsequent regressions (except the median regression), the two observations with the largest impact on the parameter vector are dropped from the sample. 9 When female education, income inequality, ethnolinguistic fractionalization, or access to safe water are missing, the variable is set to zero and a dummy variable equal to one is 6 Table 1: Dependent variables: Under-5 mortality rate (natural log), 1990. Column: I 2 3 4 Dependent variable Under-5 Mortality Rate Infant M.R. Method OLS OLS Two-stage least Two-stage least squares squares GDP per capita (In) -.661 -.598 -.645* -A62r (10.04) (9.28) (3.48) (2.78) Female education -.102 -.098 -.067 _______________ (4.03) (3.73) (2.46) Income inequality .012 .013 .003 (2.26) (2.08) (.505) Percent urban .003 .004 .001 (1.12) (.990) (.292) Predominantly Muslim .408 .399 .199 (3.13) (2.85) (1.53) Ethnolinguistic .625 .626 .314 fractionalization (4.14) (4.20) (2.12) Tropical country -.006 -.004 -.046 (.063) (.043) (.478) Access to safe water -.003 -.002 -.003 (1.16) (-.777) (1.44) Additional variables Dummy variables Dummy variables for area and a constant term. Dummy for area and a variables for when female education, income inequality, constant term. ethnolinguistic fractionalization, or access to safe water are missing. R-Squared .9023* .9454r .9452 .9369 Number of observations 109 104 104 104 Notes: White heteroskedasticity-corrected t-statistics are in parentheses. Countries with largest influence on parameter vector on OLS case and hence dropped are for column (I) Hong Kong and Turkey, for columns (2) and (3) Rwanda and Korea (South), and for column (4) Sri Lanka and Turkey. *When dummy variables for area are excluded, the R-squared for column (1) is equal to .8534. Instruments are, whether or not the country's main export is oil, and years since 1776 that the country has been independent. included in the regression. Of the 104 countries included in the regression in column 2 of Table 1, 61 are not missing any of these variables, and 16 are missing only the inequality variable. 7 The high explanatory power of these regressions is even more impressive when one considers the role of measurement error, both in mortality and in the independent variables. Imagine the hypothetical case that one was "explaining" an individual's height with a measure of height taken one minute later. If both measures were completely accurate to the recorded significant digits the regression R-squared would be 1, but if there were random error in the first measurement with variance a,2 then the R-squared would be I - a,2 / ( ay2 + av2 ) where ay2 is the true variance of height in the sampled population. In the case of child mortality we can use a recent compilation of mortality estimates to gauge the degree of measurement error as there are multiple estimates for the same country for the same year (United Nations, 1992). The differences across estimates using different surveys and different methods for the same periods are substantial: the four estimates of child mortality per 1000 in Egypt in 1980 are 203, 167, 171, and 97, the four estimates for Ghana in 1975 are 187, 171, 130 and 125. The average coefficient of variation of these different measurements for a sample of countries with repeated measurements is 0.129 compared to a coefficient of variation for the sample as a whole of 0.880, which would suggest the maximum achievable R-squared, if all the true variation were explained, of 0.978.'° The effect of measurement error in the independent variables has a similar effect in lowering the feasible R-squared." '° We use the coefficient of variation since the mean of the small sample for which we have repeated measurements is much larger (142) than the mean across all countries (87). " At low levels, measurement error in the dependent and independent variables are roughly additive in lowering the R-squared. 8 Many researchers do not trust cross-national comparisons because they doubt the data are sufficiently reliable (Srinivasan, 1994). While it is true that there are difficulties in accurately measuring child mortality, incomes, and educational levels across countries, this cannot be an explanation for a high degree of explanatory power; the more strongly one believes the data is "bad" the more puzzling is the high explanatory power of the regression. In addition to the high explanatory power, the regression is impressive as the direction and magnitude of the estimates on the variables are consistent with aggregate and household results elsewhere. The estimated elasticity of mortality with respect to income of around -0.6 is consistent with what has been found elsewhere, either using cross-sectional or time series evidence. Kakwani (1993) uses functional forms that allow for varying income elasticity in cross-national data and finds a range of elasticities between -0.5 and -0.6. Pritchett and Summers (1996) use time series on changes in income and under-S mortality from 1960 to 1980 and find the long-run elasticity to be between -0.43 and -0.76 (depending on the instruments used in the instrumental variables estimation). Pritchett (1997) uses time-series of 22 countries with data going back to 1870 to do fixed effects estimation and finds an infant mortality elasticity with income of-0.59. Jamison, Wang, Hill and Londono (1996) combine cross section and time series data and find an income elasticity of-.65 in 199012. Anand and Ravallion (1993) find that average income does have an important impact on health status, but that it operates only through its effect on the share of the population in poverty (less than a --1985 PPP-- dollar a day) while 12 They allow the income elasticity to vary across periods and find that the estimate increases from -.40 in 1960 to -.65 in 1990. 9 we find that adding an estimate of the proportion of population in poverty leaves our income estimate unaffected.'3 In interpreting estimates of the impact of income on health there is a potentially serious econometric problem of reverse causation, as better health status might cause higher average income. This is related to the problem of the attribution of effect between income and health policy, as better policy might cause better health which might lead to higher income. While it is almost certainly true that better health leads to higher income at the individual level (Strauss and Thomas, 1995), the effect is less clear at the aggregate level. Moreover, it is unlikely that the mortality rate of children under five would effect higher incomes contemporaneously. Pritchett and Summers (1996), using instrumental variables and fixed effects estimation on a panel of data show that wealthier is causally healthier for the cases of infant mortality, under-5 mortality, and life expectancy. While not going to the same level of detail in the estimation, we use two-stage least squares estimation (column 3 of Table 1) using as instruments for income whether or not a country's primary export is oil, and the percentage of years since 1776 that a country has been independent. Other than the much larger standard errors the results are largely unaffected, suggesting that whatever role health might have in causing higher incomes it does not affect the estimation of the cross-national impact of income on child mortality. The results on female education are consistent with both aggregate and household level studies (King and Hill, 1993, Subbarao and Raney, 1995, Caldwell 1986, 1990, and Hobcraft, 13 In any case, the difference in the two results is not about whether income affects health status- but about the particular functional form of the specification for estimating the income effect. 10 1983). Table 2 presents the differences in under-5 mortality by educational status of the mother derived from forty-five Demographic and Health Surveys which imply that, for the average of this sample, mothers who have secondary schooling in addition to primary schooling (with usually about four years between levels) have child mortality rates 35.8 percent lower. Our cross country results, where mean female schooling is 4.97 years, imply increasing female schooling by 4 years would lead to 39.2 percent fall in under-5 mortality. Table 2: Mean (standard deviations) under-5 mortality by education level of the mother: DHS results Mothers Mothers Change Mothers with Change Number of.. with no with (percent) secondary (percent) countries schooling primary schooling schooling East Asia and 119.0 69.9 41.3 39.2 43.9 3 Pacific (40.3) (18.1) (16.0) ___ Latin America and 110.9 79.3 28.5 49.4 37.7 10 Caribbean (52.5) (43.1) (26.1) Middle East and 94.5 62.1 34.3 41.4 33.3 6 North Africa (34.3) (14.7) (15.6) South Asia 127.7 84.8 33.6 70.9 16.4 4 (41.5) (29.9) (22.9) Sub-Saharan Africa 179.8 139.9 22.2 87.5 37.5 22 (60.8) (43.0) (24.0) All 144.4 106.5 26.2 68.4 35.8 45 (62.8) (49.9) . (30.1) Notes: Compiled from a set of DHS Final Reports from 1987 to 1995. Where there was a choice, the value for "4completed primary" and/or "completed secondary" was chosen. In other cases the group may include mothers who have attended but not completed the educational level. Income inequality and ethnolinguistic fractionalization are each associated with worse under-5 mortality. Flegg (1982) finds that the elasticity of infant mortality with respect to the Gini coefficient is 0.77 when female illiteracy is controlled for. Our results suggest that the 11 elasticity is equal to 0.51 at mean income inequality. Our estimates imply that a country with the high inequality of Brazil (Gini of .596) compared to that of Sri Lanka (Gini of .301) could expect mortality to be 38 percent lower. While it might appear that the cumulative empirical evidence of a redistributive impact on mortality is weaker than that of the level of income or female education, this is merely because reliable data on income distribution are scarce and recent. Nearly all estimates of the relationship between mortality and income use non-linear functional forms which imply a redistribution from rich to poor would increase the average level of mortality. For instance, Bidani and Ravallion (1997) estimate the level (not the log) of life expectancy and infant mortality on a non-linear functional form for income that allows different elasticities for the poor and the non-poor and find strong confirmation of different impacts. The rejection of a linear functional form constitutes powerful evidence that income distribution matters. Second, in estimating a functional form that is non-linear in income with aggregate data the specification of the relationship itself implies that an income inequality term must be included or else the regression is mis-specified. For instance, our estimates are typical in estimating a relationship between the log of average mortality and the log of average income.'4 But the aggregate data are averages of individual data and the log of average income is not the same as the average of log incomes and the difference between these two grows as the variance 14 A variety of non-linear specifications have been used in this literature, sometimes a logistic form is imposed on mortality as the dependent variable with income in logs, sometimes just income is in logs, sometimes other non-linear specifications in income are used (e.g. quadratic). 12 of income grows."5 Therefore, if the non-linear functional form assumed for aggregates also holds in the individual level data (and there is no reason to assume that the form of the relationship between income and mortality within countries is different than the form across countries) then this implies that a measure of the distribution should necessarily be included in the regression.'6 The fact that a country is predominantly Muslim is significantly positively associated with higher under-5 mortality, a result that has been suggested previously in the literature. Caldwell (1986) finds that no exceptionally good infant mortality health performers relative to income are predominantly Islamic, whereas many of the poor performers are. In our results, the coefficient on "predominantly Muslim" while strong for child mortality, is only insignificantly associated with higher infant mortality. This pattern of higher child mortality is consistent with the pattern of higher mortality for girls aged 14 in some Muslim countries such as Pakistan and Egypt (Filmer, King, and Pritchett, 1997). The exact causal mechanisms behind either of these relationships have yet to be fully investigated. " Mathematically, if one has a monotonically increasing (f 0) but concave function f' _ | ' ~~~~~~~~~~~~~OrganizationI ,EHS Efficacy of and incentives I PubZ7 4- -public sector in the public I ~~sector Level of publicly provided effective health services/.------ 'Extent to which Demand for public provision treatment. 8 *HS7~ 4 displaces 4 Existing and private potential provision I private supply. Consumption of effective health services a 7 Clinical {iMedical and 8 Consg < -" i effectivness of 4 biological the services processes g ~~Health V status J < ~~~(HS) 35 Appendix 1: Cross national data availability and reliability This appendix provides details on four aspects of the cross national regressions: data on health status, data on independent variables, data on health sector strategy, and functional form. a) Data on health status The data used here are as reported by UNICEF (1992) except for the mortality rates for Zaire which are unbelievable: under-5 mortality rate for 1990 is reported as 130 and infant mortality rate as 79. These are replaced with the, more reliable, estimates reported in United Nations (1992) for 1984. These are 200 for under-5 mortality and 126 for infant mortality. b) Data on non-health sector variables Income Real GDP per capita in 1995 international dollars (i.e. adjusted for Purchasing Power Parity) are from the Penn World Tables 5.6. Education Average education levels for men and women over 15 are from Barro and Lee (1996). Income inequality Gini coefficient as calculated by Deininger and Squire (1996) multiplied by 100. Percent urban Percent of the country's population that lives in urban areas. From the World Bank's Social Indicators of Development database. Predominantly Dummy equal to one if over 90 percent of the country's population is Muslim Muslim. Ethnolinguistic Index of ethnolinguistic fractionalization for 1960. Measures the fractionalization probability that two randomly selected people from a given country will not belong to the same ethnolinguistic group as reported in Easterly and Levine (1996). Tropical country Dummy equal to one if part of the country's territory lies within 20 degrees of the equator. Access to safe Percent of population with access to safe water. From the World water Bank's Social Indicators of Development database. Oil exporter Dummy equal to one if the country primary export is fuels (mainly oil) as classified by the World Bank's World Development Indicators (1996) plus Kuwait. 36 Years independent The percentage of years since 1776 that a country has been independent, as reported in Easterly and Levine (1996). Defense spending Defense spending as a share of GDP, as reported in CIA (1994) c) Data on health sector variables Health As part of the World Development Report 1993, Investing in Health, Expenditures figures on the magnitude of health expenditures were generated (Murray, Govindaraj, and Musgrove, 1995). These have undergone various updates, based on country level reports. We use the latest data available in our analysis. Percentage of As reported in the WHO's Health for All Database. The observation national health closest to 1990 in the 1986-1993 period is used. expenditures devoted to local health services Percentage of the As reported in the WHO's Health for All Database. The observation population with closest to 1990 in the 1986-1993 period is used. local health services, including availability of essential drugs, within one hour's walk or travel 37 Table A2- 1: Summary statistics for estimation sample (Number of observations = I 00) Mean Std. Dev. Correlation with GDP per capita (In) Under-5 mortality rate (In) 3.928 1.183 -.9274 Under-I mortality rate (n) 3.605 1.053 -.9200 GDP per capita (In) 7.914 1.151 1.000 Public health expend. (In of share of GDP) -3.769 .7591 .6228 Female education 4.971 2.753 .8115 Income inequality 40.93 8.706 -.2432 Percent urban 49.96 25.01 .8075 Predominantly Muslim .120 .327 .0096 Ethnolinguistic fractionalization .4219 .2871 -.5112 Tropical country .620 .488 -.6445 Access to safe water 67.13 25.73 .6909 East Asia and Pacific .10 .302 .0377 Latin America and Caribbean .21 .409 .0090 Middle East and North Africa .09 .288 .1236 South Asia .05 .219 -.1346 Sub-Saharan Africa .31 .465 -.6812 Female education missing .11 .314 -.2463 Income inequality missing .23 .423 -.2649 Ethnolinguistic fractionalization missing .08 .273 .0390 Access to safe water missing .08 .273 .3405 Under-5 mortality rate 87.75 77.25 Under-I mortality rate 56.96 44.66 GDP per capita 5004 5300 Public health expend. (Share of GDP) .02967 .01998 38 References: Alesina. Alberto, Reza Baqir, and William Easterly, 1997. "Public Goods and Ethnic Divisions," NBER Working Paper No.6009. Cambridge, MA. Anand. Sudhir, and Martin Ravaillon, 1993. "Human Development in Poor Countries: On the Role of Private Incomes and Public Services," Journal of Economic Perspectives, 7(1):133-50. Barro, Robert J. and Jong-Wha Lee. 1996. "International Measures of Schooling Years and Schooling Quality." American Economic Review. Papers and Proceedings 86 (2): 218-23. Bidani, Benu, and Martin Ravaillon, 1997. "Decomposing Social Indicators using Distributional Data," Journal of Econometrics, 77:125-139. Caldwell, John C., 1986. "Routes to Low Mortality in Poor Countries," Population and Development Review, 12(2): 171-220. Caldwell, John C., 1990. "Cultural and Social Factors Influencing Mortality Levels in Developing Countries," The Annals of the American Academy, 510:44-59. CIA, 1994. "The World Factbook 1994" CIA. Deininger, Klaus and Lyn Squire, 1996. "A New Data Set Measuring Income Inequality," I h World Bank Economic Review, 10(3):565-591. Easterly, William and Ross Levine, 1996. "Africa's Growth Tragedy: Policies and Ethnic Divisions," mimeo (September). Washington, DC: The World Bank. Filmer, Deon, Jeffrey Hammer, and Lant Pritchett, 1997. "Health Policy: Weak Links in the Chain?" mimeo, Development Research Group, The World Bank. Washington, DC. Foster, Stanley O., Deborah A. McFarland, and A. Meredith John, 1993. "Measles," in Disease Control Priorities in Developing Countries, Jamison et al eds. Halstead, Scott B., Julia A. Walsh, and Kenneth S. Warren sed., 1985. Good Health at Low Cost (Conference Report), The Rockefeller Foundation. New York: The Rockefeller Foundation. Hammer, Jeffrey S., 1997. "Economic Analysis for Health Projects," World Bank Research Observer, 12(l):47-7 1. 39 Heerink, Nico, and Henk Folmer, 1994. "Income Distribution and the Fulfilment of Basic Needs: Theory and Empirical Evidence," Journal of Policy Modeling, 16:625-652. Hobcraft, John, 1993. "Women's Education, Child Welfare and Child Survival: A Review of the Evidence," Health Transition Review: The Cultural. Social and Behavioral Deterninants of Health, 3:159-75. Jamison, Dean T., W. Henry Mosley, Anthony R. Measham, Jose-Louis Bobadilla eds., 1993. Disease Control Priorities in Developing Countries. Oxford Medical Publication. Jamison, Dean T., Jia Wang, Kenneth Hill, and Juan-Luis Londoflo, 1996. "Income, Mortality and Fertility Control in Latin America: Country-Level Performance 1960-90," mimeo, LAC Technical Department, The World Bank. Washington, DC. Kakwani, Nana, 1993. "Performance in Living Standards: an International Comparison," Jounal of Development Economics. 41:307-336. Kim, Kwangkee and Philip M. Moody, 1992. "More Resources Better Health? A Cross-National Perspective," Social Science and Medicine 34(8):837-842. Kunstadter, Peter, 1985. "Remarks [on Good Health at Low Cost]," in Haltead et al eds.(l985) Gwatkin, Davidson, R., 1980. "How Many Die? A Set of Demographic Estimates of the Annual Number of Infant and Child Deaths in the World," American Journal of Public Health, 70:1286-1289. Lopez, Alan D., 1993. "Causes of Death in Industrial and Developing Counties: Estimates for 1985-1990," in Disease Control Priorities in Developing Countries, Jamison et al eds. Martines, Jose, Margaret Phillips, and Richard G. Feachem, 1993. "Diarrheal Diseases," in Disease Control Priorities in Developing Countries, Jamison et al eds. Murray, Christopher, Karel Stybolo, and Annik Rouillon, 1993. "Tuberculosis," in Disease Control Priorities in Developing Countries Jamison et al eds. Musgrove, Philip, 1996. "Public and Private Roles in Health: Theory and Financing Patterns," World Bank Discussion Paper No. 339. Washington, DC. Najera, Jose A., Bernhard Liese, and Jeffrey Hammer, 1993. "Malaria," in Disease Control Priorities in Developing Countries Jamison et al eds. Pritchett, Lant, 1997, "Divergence, Big Time," Journal of Economic Perspectives. (Summer) 40 Pritchett, Lant and Lawrence Summers, 1996. "Wealthier is Healthier," Journal of Human ResQurce, 31(4):841-868. 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. Srinivasan, T. N., 1994. "Data base for development analysis: an overview," Joumal of Development Economics, 44:3-27. Stansfield, Sally K., and Donald S. Shepard, 1993. "Acute Respiratory Infection," in ease Control Priorities in Developing Countries, Jamison et al eds. Strauss, John, and Duncan Thomas, 1995. "Human Resources: Empirical Modeling of Household and Family Decisions," in J. Behrnan and T.N. Srinivasan,, eds., Handbook of Development Economics, Vol 3.: Amsterdam: North Holland Press: 1883-2023. Subbarao, K.Y. and Laura Raney, 1995. "Social Gains from Female Education: A Cross- National Study," Economic Development and Cultural Change 44:105-128. UNICEF, 1992. The State of the World's Children: 1992 New York: Oxford University Press. United Nations, 1992. Child Mortality since the 1960s: A Database for Developing Countries New York: United Nations. Vicusi, Kip, 1993. "The Value of Risks to Life and Death," Journal of Economic Literature, 3 1(4): 1912-46. Walsh, Judith A., Chris N. Feifer, Anthony R. Measham, and Paul J. Gertler, 1993. "Maternal and Perinatal Health," in Disease Control Priorities in Developing Countries Jamison et al eds. Wolfe, Barbara, 1986, "Health Status and Medical Expenditures: Is there a Link?" Social Science and Medicine 22(10), pp 993-999. World Bank, 1993. World Development Report 1993: Investing in Health. New York: Oxford University Press. World Bank, 1997a. Sector Strategy: Health. Nutrition. and Population. Human Development Network, The World Bank, Washington, DC. World Bank, 1997b. Social Indicators of Development. New York: Oxford University Press. 41 Policy Research Working Paper Series Contact Title Author Date for paper WPS1 842 Motorization and the Provision of Gregory K. Ingram November 1997 J. Ponchamni Roads in Countries and Cities Zhi Liu 31052 WPS1 843 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 WPS1 846 Africa's Role in Multilateral Trade Zhen Kun Wang November 1997 J. Ngaine Negotiations L. Alan Winters 37947 WPS1 847 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 WPS1849 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 WPS1 852 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 WPS1854 Regulatory Tradeoffs in Designing Claude Crampes November 1997 A. Estache Concession Contracts for Antonio Estache 81442 Infrastructure Networks WPS1855 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 WPSI 857 Leasing to Support Small Businesses Joselito Gallardo December 1997 R. Garner and Microenterprises 37664 WPS1858 Banking on the Poor? Branch Martin Ravallion December 1997 P. Sader Placement and Nonfarm Rural Quentin Wodon 33902 Development in Bangladesh WPS1859 Lessons from S§o 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 WPS1863 Are the Poor Less Well-Insured? Jyotsna Jalan December 1997 P. Sader Evidence on Vulnerability to Income Martin Ravallion 33902 Risk in Rural China