FILt$COpy SWP594 The Effect of Piped Water on Early Childhood Mortality in Urban Brazil, 1970-1976 Thomas Merrick WORLD BANK STAFF WORKING PAPERS Number 594 WORLD BANK STAFF WORKING PAPERS 7 S Number 594 C The Effect of Piped Water on Early Childhood Mortality in Urban Brazil, 1970-1976 Thomas Merrick INTERNATIONAL MONETARY FUND JOINT LIBRARY MAR G 1984 IlTERNATIONAL BANK FOR 1lECONSTllUCTION AN$ DEVELOPMENT WASHIliGTON, D.C. 20431 The World Bank Washington, D.C., U.S.A. Copyright i 1983 The International Bank for Reconstruction and-Development / THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. First printing August 1983 All rights reserved Manufactured in the United States of America This is a working document published informally by the World Bank. To present the results of research with the least possible delay, the typescript has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. The publication is supplied at a token charge to defray part of the cost of manufacture and distribution. The views and interpretations in this document are those of the author(s) and should not be attributed to the World Bank, to its affiliated organizations, or to any individual acting on their behalf. Any maps used have been prepared solely for the convenience of the readers; the denominations used and the boundaries shown do not imply, on the part of the World Bank and its affiliates, any judgment on the legal status of any territory or any endorsement or acceptance of such boundaries. The full range of World Bank publications is described in the Catalog of World Bank Publications; the continuing research program of the Bank is outlined in World Bank Research Program: Abstracts of Current Studies. Both booklets are updated annually; the most recent edition of each is available without charge from the Publications Distribution Unit of the Bank in Washington or from the European Office of the Bank, 66, avenue d'Iena, 75116 Paris, France. Thomas Merrick is director of the Center for Population Research at Georgetown University and a consultant to the World Bank. Library of Congress Cataloging in Publication Data Merrick, Thomas William. Access to piped water and early childhood mortality in urban Brazil, 1970-1976. (World Bank staff working paper ; no. 594) Includes bibliographical references. 1. Children--Brazil--Mortality. 2. Infants--Brazil --Mortality. 3. Municipal water supply--Brazil. I. Title.. II. Series: World Bank staff'working papers ; no. 594. HB1323.C5M47 1983 304.6'4'0880542 83-6681 ISBN 0-8213-0190-X ABSTRACT The Effect of Piped Water on Early Childhood Mortality in Urban Brazil, 1970 to 1976 This paper utilizes data on the ratio of surviving children and other variables for a sub-sample of mothers aged 20 - 29 from the 1970 Brazilian population census and the 1976 Brazilian National Household Survey (PNAD) to' analyze the impact of increased access to piped water in urban households on trends and income class differentials in child mortality. Path analytic regression techniques are used to test a recursive model linking the supply and demand for piped water to selected household and community level variables, and to examine their joint effect on child mortality. Mortality is measured with a standardized ratio proposed by Trussell and Preston (1982). The analysis shows that increased maternal education accounted for a larger share -- 34% -- of the mortality decline between 1970 and 1976 than any other single factor, including access to piped water. Increased access to piped water did also contribute, accounting for about 20 percent of the decline; moreover access to water helped to reduce the mortality differential between lower and higher income and education classes. Obviously, piped water is a costly intervention; an analysis of the relative cost-effectiveness of various interventions -- e.g. pipe water vs. female education -- requires that their costs be estimated and matched to these estimates of relative effectiveness. Table of Contents Introduction . . . . . . . . . . . . . . . . . . .1 Trends and Differentials in Child Mortality . . . 3 Income Class Differentials in Mortality . . . . . 4 Determinants of Mortality Differentials 6 Spec-ification of a Causal Model . . . . . . . . . 9 Measurement of the Dependent Variable . . . . . . 13 Analysis of Differences in CM in 1970 and 1976 . 16 Path Analysis: Total Direct and Indirect Effects .18 Test for Interaction . . . . . . . . . . . . . . . 21 Decomposition of Changes in CM from 1970 to 1976 .... . . . . . . . . . . . . . . . 22 Summary and Conclusions . . . . . . . . . . . . . 23 Footnotes . . . . . . . . . ... . . . . . . .. . 26 Map . . . . . . . . . . . . . . . . . . . . . . . 30 Tables . . . . . . . . . . . . . . . . . . . . . . 33 References ..... ... . 44 High mortality rates among infants and young children are a poignant measure of the human cost of poverty and inequality in developing countries. Though the transfer of advanced medical technology has contributed to declines in the average level of mortality, income class differentials in early childhood mortality rates have persisted and possibly widened (Behm, 1980). For policy makers and agencies concerned the links between child mortality and income inequality, a key question is how to design and implement health and environmental programs that reduce mortality rates among the least privileged groups. This paper examines the link between one such program, Brazil's urban water project, and changes in early childhood mortality patterns in that country during the early 1970s.1 The effectiveness of interventions for lowering mortality depends both on the capacity of implementing agencies to supply the appropriate types of services to low income families, and on those families, capacities to marshall their limited resources in ways that they can utilize such services to improve their children's chances of survival. The theoretical implication of this observation is, as Selowsky (1979) has pointed out, that there is a demand as well as a supply side to the question. "Demand" is understood not in the sense that poor parents would choose between having more or fewer of their children survive, but rather that the success of their efforts to raise healthy children depends on the choices they make in allocating their limited resources. These points are illustrated by Brazilian experience over the last two decades. Brazil has enjoyed sustained growth in aggregated output per capita, but has been plagued by persistent inequality in,the distribution of personal and family income, as well as by large interregional differentials in income (Pfeffermann and Webb, 1979; Knight and Moran, 1981; Lluch, 1981). .1 Significant differentials in mortality also have been observed among regions and income classes, particularly for infants and children (Carvalho and Sawyer, 1971; Carvalho.and Wood, 1978; Brazil, Fundacao IBGE, 1981). Infant mortality rates observed in a number of Brazil's larger cities increased from the late 1960s to the early 1970s, leading some observers to hypothesize that it was the effect of declining living standards among lower income groups *(Yunes and Ronachezel, 1974; Wood, 1977).2 The Brazilian experience is signficant also in that during the 1970s it embarked on an intensive effort to improve urban environmental conditions with a known link to mortality: water supply and sanitation. PLANASA, the national water and sanitation program, provided public subsidies designed to supply 80 percent of urban households with piped water and sewer connections by 1980 (Almeida, 1977). Another consideration is that Brazil made substantial progress in raising the level of educational attainment during the 1960s, including that of women, which is significant in that women educated in the 1960s became mothers in the 1970s, and because maternal education is an important determinant of the survival chances of infants and young children (Behm, 1980; Caldwell, 1980). These developments raise the question of whether interclass differentials in infant and child mortality were-affected by such measures, and of how their effectiveness in reducing mortality may have been conditioned by continued income inequality. This study utilizes data from the one-percent sample of Brazil's 1970 population census (Brazil,, Fundacao IBGE, 1979) and from the 1976 national household survey .(PNAD, Pesquisa Nacional por Amostra de Domicilios: Brazil, Fundacao IBGE, 1982) to measure interclass differentials in child mortality in urban Brazil and to analyze the relation between changes in these * differentials and improved access to piped water, increased educational 2 attairment of mothers, and other household characteristics. The analysis is based on subsamples extracted from both files of a study population consisting of currently married (with husbands present) urban women between the ages of 20 and 29 who had at least one live birth. In addition to data on these women's births, survival of their children, and their education, the subsamples include information on their husbands and on the households in which they resided. One reason for selecting this particular sample is that the PLANASA program was an urban project.3 Another relates to timing. Access to water was observed at the time of the interview, while data on the survival of children relate to a period of years before the interview. Focusing on the experience of women in their 20s reduces the time differential between the two variables (which increases with the interviewee's age), though it does not eliminate the timing problem. Trends and Differentials in Child Mortality Most of the available evidence on trends and differentials in child mortality in Brazil consists of indirect measures derived from survival ratios in the censuses and PNAD surveys.4 Table 1 presents estimates of trends and regional differentials in infant mortality reported in the National Academy Committee's preliminary report on-Brazil (NAS, 1983). The table also presents rates based on advanced tabulations of the 1980 census, which became available after the NAS estimates were prepared. The date of the census or sample on which the estimates are based is shown in the table, though it should be remembered that the date to which the estimates refer is approximately three years prior to the date of the interview. At the national level, infant mortality declined from 152 infant deaths per 1000 live births in 1950 to 107 in 1970, a decrease of 29 percent. The decline was greater during the 1950s 3 (19 percent) than the 1960s (13 percent). Regional differentials (using the percentage difference between infant mortality rates in the Northeast and Sao Paulo as a rough index of such differentials) increased from 54 percent in 1950 to nearly 100 percent in 1960, then dropped to 70 percent in 1970. After 1970 the pace of the national decline accelerated again, with the rate falling by 20 percent between 1970 and 1980. The 1972, 1973, and 1976 PNAD surveys and the 1974-75 ENDEF study present conflicting stories about what may have been occurring during the decade. The PNAD surveys suggest an acceleration of mortality delcine during the first half of the decade, with a slowdown after 1976. The ENDEF results suggest that very little decline occurred during the early 1970s, though caution is required in using 1970 census data as a basis of comparison with ENDEF because of the nature of that survey.5 Comparison of PNAD results with preliminary 1980 census figures suggests either that PNADs may have overstated the decline in the early 1970s or that conditions leading to accelerated decline in the first half of the decade may have changed for the worse in the second half. Income Class Differentials in Mortality Income class differentials are another important feature of Brazilian mortality patterns. Carvalho and Wood (1978) found substantial differences in life expectancy in their tabulations of 1970 data by household income level, with a striking twenty-five year gap between the life expectancy of the lowest income class in the Northeastern region and the highest income class in the Southeast. Published data from the 1976 PNAD survey include tabulations of child survival ratios by minimum wage level, but tests of estimates derived from these tabulations revealed problems of comparability with the categories employed by Carvalho and Wood for.1970. Since raw data files for both the 4 1970 census and 'the 1976 PNAD survey were available, ratios were retabulated using households income decile categories derived by Lluch (1981). The decile categories were then clustered into three groups (deciles 1-3, 4-6, 7-10), following a suggestion by Carvalho and Wood that possible bias arising from interclass mobility in rates based on retrospective reporting can be reduced by grouping the population into fewer and broader categories.6 Trussell's variant of the Brass model life table method of calculating child mortality rates from questions on surviving children was used to derive probabilities of dying between birth and ages two and three (2qo and 3q0) for children born to women grouped according to their household income decile in 1970 and 1976, as shown in Table 2.7 Focusing on the average decrease in the probability of dying between birth and ages two and three, the table shows an overall decline of 26 percent from 1970 to 1976, which is greater than the 17 percent decline in the national level infant mortality rate reported in Table 1, and reflects the restrictions in the study population described above. Declines were greater for higher income groups in both rural and urban areas, but urban declines were greater for all groups than those in rural areas (the upper income groups in rural areas were consolidated because of the limited number of cases in the 7-10 category). Timing may account for some of the income class differentials. Owing to later occurrence of births among women in higher income groups, the reference date for the second mortality rate recorded for them is more recent and would therefore capture more decline than is recorded for lower income categories. Even if the timing effect could account for the apparent widening of interclass differentials, the data in Table 2 raise doubts about the effectiveness of efforts to reduce mortality differentials durng the 1970s. These data 'do not suggest that such differentials narrowed. 5 Determinants of Mortality Differentials Variables that influence infant and child mortality and that could explain differentials among income groups include those with a potential direct impact on the, survival chances of young children as well as those relating to possible indirect effects associated with broader socioeconomic characteristics of their mothers, their families, and the communities in which they are reared. Relevant causal variables include the nutritional level of both mothers and children, the quality of their environment, and the maternal and child health care to which they have access (Dyson, 1978; Antonovsky, 1980; Gwatkin et al., 1980; Loriaux, 1980). Background variables such as income and education may influence mortality directly or by affecting or conditioning directly causal variables. Empirical investigations of the determinants of infant and child mortality also have suggested that causal relationships are reinforcing rather than distinct and separable. Puffer and Serrano (1973) studied infant and child mortality patterns in several Latin American cities, including three in Brazil (Recife, Riberao Preto, and Sao Paulo). Diarrheal diseases were the leading causes of postneonatal deaths in all three cities, with the proportion being highest in the poorest of the three, Recife. Their investigation revealed a strong synergism between poor nutrition and environmental factors, and concluded that children who were underweight at birth owing to poor maternal nutrition and who were poorly nourished after birth were also more susceptible to infections associated with unsafe water, poor sanitation, and .8 inadequate housing. Their study also showed that a high proportion of the households reporting infant deaths lacked adequate water supply, though they did not analyze systematically the relation between the two variables. Access to internal piped water in the household is likely to be of most 6 direct benefit in lowering child mortality by reducing exposure to water-borne diseases, particularly diarrheal disorders. Although evidence of the specific links between water supply and such disorders is limited, the literature suggests that increases in the amount of water used contributes to better hygiene and the elimination of bacteriological contamination reduces the risk of infection through intake. Schneider, Shiffman, and Faigenblum (1978) present clinical data from a Guateinalan study that document the loss of nutrients during diarrheal episodes, and a recent World Bank study reports that gastrointestinal diseases reduce absorption of nutrients by as much-as 30 percent in acute cases (World Bank, 1980). An added dimension in the Brazilian situation relates to the very low iverage duration of breastfeeding (Anderson, 1980). Thus infants are exposed to the risk of infection from contaminated water and through poor hygiene in prepared food at a very early age. This synergism extends to other socioeconomic variables, particularly maternal education (Behm et al., 1978; Arriaga, 1980; Cochran et al., 1980). Because the causal influences may be both direct and indirect, explanations of the way in which maternal education affects mortality differentials vary. In interpreting Nigerian data, Caldwell (1979: 13-14) suggests three links: (1) mothers with more. education tend to be less fatalistic about illness and therefore prone to seek outside medical assistance for an ill child; (2) educated mothers are more likely to adopt improved child care practices such as boiling water used in-the preparation of infant fornulas; and (3) education may change intrafamily family relationships, leading to a more "child- centered" orientation that would have a positive impact on children's health. Schultz (1979), drawing on household production-theory, hypothesized that more educated women earn more in the labor market and marry more educated men, 7 and consequently have higher family income that would enable them to purchase goods and services that improve child health. Education could also increase the effectiveness of women's non-market child care activities, though, as Schultz cautions, the fact that market work requires women to be absent from the home could have an offsetting negative impact on the quality of child care. Brazilian data illustrate the strong link between maternal education and children's survival chances. Table 3 shows the proportion dying among children of urban mother grouped by education. Education categories range from women with no formal education to those with ten or more years of schooling as reported in the 1970 census and the 1976 PNAD survey. Current age of the mother is used as a control variable. There is a steady decline in the proportion of children dying as educational attainment increases. The most striking difference in mortality ratios, however, is the contrast between mothers with no formal education and other groups: the rate of child mortality among women with ten or more years of schooling is less than a fourth of the rate for mothers with no education. Comparing 1970 with 1976, there appears to have been improvement among the upper three education groups, but the picture is less clear for the group with no formal education. There is a substantial decline in the relative number of women with no formal education, and proportional increases in the two highest classes, suggesting that between 1970 and 1976 shifts occurred from the "none" group into the "1-4" category, as well as shifts from the latter into the higher categories. These shifts are consistent with the fact that a significantly increased proportion (41 percent) of women who were 20-24 years old in 1976 were in school as 14-18 year olds in 1970. The 1970 figure is nearly double what was reported a decade earlier in 1960 (21 percent). The 8 last two columns of Table 3 are also of interest in that they show a decline in the proportion of children. dying for mothers of all ages, which is consistent with the improved survival probabilities shown in Table 2. Specification of a Causal Model The census and PNAD provide information on the education, the earnings and employment status of household members, and on a number of household characteristics, including access to piped water, thereby permitting a more detailed examination of at least some aspects of the link between increased access to piped water, recent social and economic changes in Brazil, and trends and differentials in child mortality. The measure of child mortality analyzed below, CM(i), is based on the ratio of surviving children to births for the mothers in the study population, and will be discussed in detail in the next section. For each mother, a dummy variable, WAT(i), indicates whether the household in which she resided had access to piped water at the time of the interview. An adequate causal model of the links between improved access to piped water and child mortality needs to take account of two sets of relation- ships; One set pertains to variables associated with the consumption of this particular service by urban households. The other is concerned with the impact that this consumption, along with other relevant variables, has on the survival chances of children. Empirical testing of such a model requires specification of structural equations for both sets of relationships, as well as the links between them. Consumption has both a supply and a demand dimension. As Selowsky (1979) has illustrated in his study of the consumption of services in Colombia, a household may or may not consume a service such as water (1) because the 9 supply network is not geographically accessible or (2) because, in spite of accessibility, that household chooses not to do so for reasons of price, an income limitation, or as a matter of preference. The former is a supply constraint, the latter demand. A specification that accounts for supply variables requires information on availability of water at a comparatively low level of aggregation. Selowsky worked with neighborhoods in his study of Colombia. The level of aggregation is much greater in the Brazilian data. The public use sample of the 1970 census permitted tabulations for 117 geographical areas that consisted of clusters of municipalities. A water supply variable, WS(j), was calculated as the proportion of households in each of the 117 areas reported as having internal piped water in 1970. Since the 117 areas were not identifiable in the PNAD survey, WS(j) was calculated at the state level for 1976. A second variable, PWS(j), measuring the percentage point increase in the number of households served in each state between 1970 and 1976, was added as a measure of the impact of the PLANASA program on the supply of water. Water supply was assumed to be exogenous at the level of individual households, with the added restriction that the effect of supply on mortality at the household level was indirect (via WAT). Data limitations also hampered specification of demand variables. While the data files included information on a number of household level variables (e.g. income and education) that influence demand, no information was available at any level of aggregation on price (hookup and use fees) differentials. Further, those household level variables that determine the demand for water are also likely to have a causal relation with child mortality in their own right, and one (income) is endogenous at the household level. Since the study population consists of mothers, with husbands present in single family households, and with only a small proportion (about 8 10 percent) of those women as working, the father's earnings, FY(i), is used as a measure of the endogenous family income variable. Education of fathers, FED(i), and mothers, MED(i), are assumed to be exogenous at the household level. Identification of the state of residence in the two data files also made it possible to add state level information from other sources: a variable measuring per capita value added taxes, TAX(j), in each state serves as a proxy for regional differences in labor market conditi'ons. The causal relations between mortality differentials, access to pipe water, and the household and community level determinants of those variables that this paper seeks to test can be summarized in the following set of structural equations: [1] FY(i) = f[FED(i), TAX(j)] [2] WAT(i) = w[MED(i), FED(i), FY(i), WS(j)] [3] CM(i) = c[MED(i), FED(i), FY(i), WAT(i), TAX(i)] Equation [1] specifies the first of the three endogenous variables in the model, father's earnings (FY), in terms of his education, FED, and the value added tax per capita in the state in which he resides, TAX. While FED reflects the influence of individual characteristics on earning, the value added tax is as an argument in the earnings function because it reflects regional differences in the productivity of manufacturing industry in Brazil and their impact on earnings. The second endogenous variable, WAT, is represented in equation [2] as a function of household variables related to demand for water (FED, MED, FY) as well as the community level water supply variable, WS. The variable measuring increases in water supply between 1970 and 1976, PWS(j) was added to the 1976 equation: 11 [2a] WAT = w[FED(i), MED(i), FY(i), WS(j), PWS(j)] Finally, equation [3] represents child mortality, CM, as a function of exogenous household variables (FED and MED), endogenous household variables (FY and WAT), and the exogenous community level variable, TAX. The latter is included in [3) because there is good reason to expect that the regional differences in economic conditions that TAX measures not only affect CM indirectly through FY and WAT, via equations [1] and [2], but also directly through its effects on other aspects of the standard of. living in the canmunity. Since the arguments on the right hand side of equation [3] also appear in equations [1) and [2], ordinary least squares regression will not yield estimates of structural coefficients without further restrictions (Marsden, 1981: 118). Two such restrictions are invoked, one on the basis of reasoning about the nature of the causal relations expressed in the system, and the other based on an empirical check on the error terms in the estimating equations. The first restriction is that the model is fully recursive in terms of the relations among endogenous variables: FY is causally prior to WAT, and both causally prior to CM. The alternative would be to argue that there is a simultaneous relationship between the three variables, which does not appear to be warranted in this case. The second restriction is that the error terms of the three estimating equations be uncorrelated with each other and with the right hand side variables. Empirical tests showing that this restriction is justified will be reported below. With these restrictions, ordinary least squares (OLS) estimates of the reduced form coefficients of the exogenous variables can be used to estimate their total effect on CM, and path analytical techniques can be utilized to decompose their total effect into direct effects on CM and their indirect 12 effects via FY and WAT, as well as the total and direct effects of the latter (Land, 1973; Alwin and Hauser, 1975).9 Measurement of the Dependent Variable Several methodological problems arise in the use, of ratios of children surviving to children ever born in analyses of differences in mortality at the level of individual mothers. Indirect measurement techniques that are used to translate survival ratios into mortality rate for groups of women cannot be used to assign a mortality rate to an individual woman. Differences in survival ratios among individual mothers reflect not only the particular mortality regime to which their children are subject, but also differences in these children's exposure to the risk of dying, which is related to the number and timing of the births that the woman has had. This, in turn, is related to her age and to the duration of her marriage or other type of union. Finally, the relation of survivors to births is binomially rather than normally distributed, making OLS an inappropriate estimating technique unless the relation can somehow be normalized. Trussell and Preston (1982) have suggested a method for standardizing survival ratios in individual level data that (1) controls for duration of risk and (2) normalizes the distribution of these ratios. Their method employs a model schedule of survival ratios for women in different marital duration (or age) categories. Observed survival ratios are then divided by the ratio that is expected given the woman's age or marital duration and the appropriate model schedule, which yields an index that relates the individual's experience to the standard.10 The authors have demonstrated that this index can be used in multiple regression analysis of individual level data on child survivorship. Nonetheless, tradeoffs are involved in using OLS 13 rather than estimating techniques that take account of the fact that the distribution of the index is bounded by zero.- Table 4 illustrates how the index of observed/expected deaths can be used to measure differences in the child mortality ratio, CM, by educational attainment and income for various types of water supply. In the table, CM is cross-tabulated by categories of mothers' years of schooling, MED, as well as by the husband's income decile and the various types of water to which households in the sample had access, including internal piped water, the water variable used in the analysis. The first column for each of the two dates shows CM for women in households with internal piped water, followed by columns for women in households with water from various external sources. Differentials in CM relate both to water access and to education. For example, among women in households with piped water, the index for women in the "10+" educational attainment category is less than a third of that for women in the "none" group. For women in households with water, from external sources, the educational differential is less, but stil substantial. For all education groups, women with access to internal piped water, CM is 0.58 compared to 0.95 for those whose water was carried from a city tap or spring, and 1.24 for those reporting "other" sources. Within education classes, the percentage differential between those with internal piped water and external sources increases with the level of education, indicating a possible interaction among the effects of the two variables. The possibility of an interaction suggests that the specification of equation [3] be modified to test for interactions: [3a] CM = c[MED(i), FED(i), TAX(j), FY(i), WAT(i), D(i)] where D(i) = MED(i) * WAT(i). 14 Between 1970 and 1976, the proportion of women with no education declined from 35 percent to 12 percent-of the sample population, which consists of women aged 20-29 rather than women aged 20-49 as reported in Table 3. The proportion of women with 10 and more years of schooling increased from 5 to 16 percent. The proportion of households with piped' water also increased sharply, from 37 percent to 65 percent. The pattern of differentials in child mortality by education and access to water'also changed: while the ration of CM for the "10+" group to that of the "none" category fell by about 10 percent for women in households with internal piped water, it increased for women whose water came from other sources, suggesting that increased access to piped water may have helped to offset a tendency for differentials in child mortality associated with educational attainment to increase during the period under study. In the second panel of Table 4, differences in the child mortality ration by income deciles are more pronounced in 1970 among hQuseholds that have internal piped water than among those with outside sources of water supply. In 1976, larger differentials between the lowest and highest income deciles are observed among households with external sources. This narrowing suggests that income class differentials may have been reduced for households with internal piped water. It also reflects a change in the distribution of households relying on outside sources, reflecting a substantial decline in the percent of household with well-spring and other external sources. There may be interaction between income and water supply and/or a change in the nature of the relation between income and' child mortality, suggesting a test for income class/water supply interaction along the lines suggested above for water and education. For all groups, the decline in child mortality is greatest for women in 15 households with internal piped water, while there is practically no change for women in the "other" sources category and intermediate rates of decline in the groups with water from city taps and springs. Variability in child mortality among households without internal piped water is of interest because city taps and springs represent an intermediate (and less expensive) means of improving access to water. However, since the main effort in the PLANASA program was to increase access to internal piped water, and since most of the increased water access during the period under study occurred in that category, the analysis focuses on the contribution of that type of water access to differentials and changes in CM beween 1970 and 1976. Analysis of Differences in CM in 1970 and 1976 This section presents OLS estimates of the parameters in the causal model of determinants of child mortality presented in equations [1) to [3) using a path analytic- approach. Given the recursive structure of the causal model, the analysis begins by estimating the reduced forms (including exogenous variables only) of equations for the three endogenous variables, and then proceeds to estimation of the structural equations in steps in which each of the causally prior endogenous variables is added to the regression equation of the final dependent variable, CM. The total, direct, and indirect effects of each variable on CM are then measured using standardized regression coefficients and conventional decomposition techniques (see Alwin and Hauser, 1975). A check on correlations among the error terms provides a test of the validity of the recursive specification. Table 5 states the units in which the variables used in the regressions are measured, and provides sample means and standard deviations. Following 16 the suggestion of Trussell and Preston, observations are weighted by the number of children ever born in order to make the results representative of births rather than women. There is generally greater variability (as measured by the ratio of standard deviations to means) in the 1970 sample than in 1976, suggesting either increasing homogeneity in the population with respect to these variables between 1970 and 1976, or that the 1976 sample may be underrepresenting socioeconomic groups at the extremes of the range. An appendix table presents zero order correlation coefficients between each variable for both dates. The high intercorrelation between MED and FED (0.6 to 0.7) reflects the tendency for the educated to seek mates with similar levels of education. Regression results are presented in Tables 6 (for 1970) and 7 (for 1976), with both following the sequence just described: columns (1), (2), and (4) are the reduced forms of equations [1], [2], and [3] of the causal model, while columns (3) and (6) are OLS estimates of structural equations [2] and [3] (there are no causally prior endogenous variables in equation [1] for fathers' earnings). Column (5) includes the endogenous variable, FY, which is causally prior to WAT but omits WAT itself as a step in estimating indirect effects: To avoid repetition, the results in Tables 6 and 7 will not be described in detail since most of the conclusions that can be drawn from them are revealed when effects are decomposed. It is worth noting that addition of FY to the WAT regression had a greater effect in 1976 than in 1970, while the reverse occurred with the addition of FY to the CM equation. In fact, the FY coefficient was not statistically significant in 1976, suggesting a change in nature of the causal linkage between these FY, WAT and CM. Another observation is that the coefficient for WS in the CM equations was not 17 significant for either date, which is consistant with the point made earlier that the supply effect may be more indirect than direct. A final point is that analysis of error terms revealed no major problems with the recursive specification. Details are reported in a footnote.12 Path Analysis: Total, Direct, and Indirect Effects Decomposition of effects using standardized regression coefficients is useful first in revealing both the level and, since we have estimates for two points in time, changes in the relative contributions of variables to differences in child mortality, and second in providing a measure of the extent to which the effects of exogenous variables have been mediated by intervening endogenous variables. This is particularly appropriate for a study, which, in addition to seeking to determine the effect of access to pipe water on child mortality, also tries to clarify the roles that both supply and demand variables affecting that access play in the relationship between water and child mortality. Table 8 presents the decomposition of the effects on child mortality of variables in the regressions in Tables 6 and 7. Total effects are shown in the first column. For the exogenous variables, total effects are the standardized coefficients from the reduced form regressions (column 4 in Tables 6-7). For the endogenous variables, they are the coefficients from the first equation in which the variable enters (column 5 in Tables 6-7 for FY and column 6 of Tables 6-7). In the case of WAT, the total effect equals the direct effect since that variable is the last in terms of causal ordering in the model. Indirect effects are shown in columns 2 (for FY) and 3 (for WAT), and were derived using the technique described by Alwin and Hauser. 18 What does Table 8 tell us about: (a) the comparative importance of exogenous household and community level variables on differences in child mortality; (b) the extent to which their effect is direct, or mediated through endogenous household variables, WAT in particular; (c) the degree to which the mediation of WAT is influenced more strongly by household or community variables; and (d) changes, if any, in these relationships between 1970 and 1976? In considering (c) it is important to recall the dual roles that were assigned to the exogenous variables in the causal model that is being analyzed. They have a direct causal relation to child mortality, and indirect effects via FY and WAT, which have been specified as supply side effects on WAT for the community level variables and demand side effects for the household variables. Mothers' and fathers' education, which are exogenous household level variables, have the largest total effect on child mortality in both 1970 and 1976. Most of the effect of MED is direct (the ratio of the direct effect to the total effect is .85 in both instances). The ratio of direct to total effect for FED is somewhat lower (.73) in 1970, but increases in 1976 ( to .84) at the same time that its total effect increases. The exogenous community level variables, TAX and WS, have a lower but still significant total effect on child mortality . They differ in the way in which their effects are mediated, since a larger share of the effect of TAX is direct, while less than half of the effect of WS is direct. In fact, the indirect effect of TAX via WAT is positive in sign, indicating that once demand is controlled there is a lower level of individual household access to piped water in areas with higher taxes per capita. Possible adverse effects of this on CM appear to be offset by other unspecified community level influences reflected in the negative direct effect of TAX. In the case of WS, a large 19 share of its effect is mediated by WAT both in 1970 and in 1976, which is to be expected since it was introduced to measure the supply side impact of water availability at the community level on access at the household level. When community level increase in water supply (PWS) is introduced in 1976, a higher proportion of its effect is direct. The total effects of the endogenous household level variables, FY and WAT, differ in comparative importance vis-a-vis the other two sets of variables. The total effect of WAT is greater than for either of the two community level exogenous variables, but less than MED and FED. The effect of WAT increases from 1970 to 1976. FY's total effect is about the same as the community level variables in 1970 and slips well below them (and is, in fact, insignificant) in 1976. In sum, the main changes from 1970 to 1976 are as follows: (1) The total effect of FED increases, as does the direct component of its effect; at the same time, the total effect of FY drops, while the indirect component of its effect (via WAT) increases. (2) The total effect of water supply variables, WS in 1970 and WS plus PWS in 1976, increases, with most of the effect of that increase being mediated by WAT, the household level water variable. This increase in the indirect effect of exogenous supply via WAT suggests (3) that supply played a greater role in 1976 than in 1970, which is what one would expect the outcome of a successful public sector intervention to be. Interestingly, the total effect of the other community level exogenous variable, TAX, also increases, but none of that increase was mediated by the endogenous household level variables. As noted above, the indirect effect of TAX via WAT worked in the opposite direction, more so in 1976 than in 1970, which, given its interpretation as an indication that households in high TAX areas had lower access to WAT, also lends support to the conclusion that the 20 importance of water supply at the community level increased from 1970 to 1976. Test for Interaction An additional issue that is raised in relation to the causal model specified in equations [1] - [3) was interaction: the extent to which the effect of access to piped water on child mortality may have been greater or less for women within different education classes or in households with higher or lower income levels. The question of interactions is of particular interest because it would be useful to know whether a public sector intervention such as increased water supply has a redistributive effect on a sensitive measure of welfare, thereby compensating for some of the adverse effects of inequality in the distribution of human resources and income on the demand side. As suggested earlier, equation [3) can be modified to test for interactions by adding interaction terms. Two such terms have been tested in both the 1970 and 1976 data, one for WAT and MED and the other for WAT and FY. If interaction is operating in the compensatory manner described above, then one would expect that differences in CM by MED and FY would be less in households with access to piped water than those without it, which would show up in the regressions in the form of flatter slopes for the MED and FY regresson lines when WAT is present. Thus we would expect positive signs for both interaction terms. Table 9 summarizes the tests for interaction. The first and fourth columns show the 1970 and 1976 regression results for equation [3) as reported in Tables 6 and 7. The next two columns show the results when an interaction term is added, first for MED and then for FY. While the signs of all four interaction terms are positive, as expected, only the last (WAT and FY in 21 1976) is statistically significant (according to the "t" test, the probability that its coefficient is, in fact, zero is less than .02). This result suggests that while the slope of the overall FY regression line was zero in 1976, the slope for non-WAT households was negative, so that WAT did indeed contribute to a narrowing of income related differences in child mortality in 1976. Decomposition of Changes in CM from 1970 to 1976 A final question relates to the relative contribution of increased access to piped water to the 27 percent decline in child mortality between 1970 and 1976. Both the 1970 and 1976 regression equations can be evaluated usng sample means for both dates to show the contribution of changes in those means to the change in the predicted value of child mortality between the two dates. This has been done in Table 10. To simplify the discussion of the role of water, regressions were rerun without the exogenous community supply measures, so that only WAT appears in the regression coefficients shown in the first column of the table. The second and third rows are the products of these coefficients and the sample means of variables in 1970 and 1976, respectively. Column four shows the differences between columns two and three, which sum to the total change in CM. Column five expresses the ratio of the change contributed by each variable as a percent of the total change in CM. Changes in the mean values of exogenous household characteristics account for the largest share of the total decline in child mortality . With 1970 regression coefficients, MED and FED account for 54 percent of the decline. FY, an endogenous household variable, accounts for another 18 percent. With 1976 coefficients, FY's contribution disappears (it shows up as a positive 22 value, though it is statistically insignificant), and the combined contributions of MED and FED are 65 percent. Increased access to piped water plays a secondary but still significant role, contributing 18.7 percent to the decline using 1970 coefficients, and 22.3 percent with the 1976 results. While the 3.6 percentage point increase in the contribution of water to the decline in child mortality with 1976 coefficients does not signal a major structural change, the direction of the change lends further support to the view that the water program contributed to the amelioration of differences in child mortality that might have otherwise persisted owing to inequality in the distribution of income and human resources at' the household level. Summary and Conclusions The objective of this paper has been to analyze the determinants of trends and differentials in early childhood mortality in urban Brazil during the 1970s. The mortality measure used in the study was the ratio, CM, of the actual proportion of children surviving for women in the study population to the proportion that would be expected to survive given a standard mortality schedule, those women's current age, and the ages at which they began to have children. The study population consisted of samples of currently married urban mothers extracted from the Brazil's 1970 population census and its 1976 national samply survey (PNAD). Particular attention was focused on the contribution of increased access to piped water in reducing both the level of child mortality and differences between socioeconomic groups during the period covered by the two data sources. In assessing this contribution, the paper attempted first to determine the relative importance of household and community level variables as determinants of differences in child mortality, and second to identify the 23 dual roles that the two sets of variables played. In the underlying conceptual model, both household and community variables relate directly to child mortality as determinants in their own right as well as indirectly through variables that are endogenous at the household level, water being the case in point, either as demand variables (at the household level) or supply variables (at the community level). A recursive model consisting of three equations was estimated, and the results used to assess the contribution of increases in access to piped water in urban Brazil to declines in the level of child mortality by means of path analysis. The main conclusions are as follows: (1) That exogenous household variables (educaton of mothers and fathers) had the greatest total effect on differences in child mortality and that most of this effect was direct. (2) That access to piped water had a significant but secondary impact on differences in child mortality, accounting for about one-fifth of such differences. (3) That community level variables played a minor, but still significant role in child mortality differentials, with most of the effect being indirect via the household water access variable (WAT), which was expected since they entered the model via the water supply equation in the specification. Examination of changes from 1970 to 1976 indicated that most of the decline in child mortality could be accounted for in terms of changes in the composition of the study population rather than changes in the parameters of 24 the causal model, with increased education of both mothers and fathers playing a prime role, and increased access to piped water a secondary role. Some changes were observed in the parameter estimates, ones consistent with the expectation that increased water supply could contribute to attenuation of differentials in child mortality associated with education and income. The relative weight of supply variables in the decomposition of child mortality differentials increased form 1970 to 1976, though most of this increase continued to be indirect, via WAT, the household level water access variable. Interaction terms were introduced as a test of the strength of this attenuation effect, and indicated that in 1976 the effect of water was somewhat greater among lower income groups. The ameliorating effect of increased access to piped water on income class differentials in early childhood mortality was thus significant but limited in comparison with other household level characteristics, particularly education. This does not imply that community level interventions were ineffective; though the level of education was specified as exogenous at the household level in this study, it was also the product of community level action. The main difference is that most of the interventions that affected child mortality through educational attainment were made earlier than those operating via increased access to water. Increased education had a greater impact, but it also took longer. 25 Footnotes 1. This paper is a partial output of the World Bank's Research Project #672- 21, and was produced with partial financial support from the World Bank. The author alone is responsible for views presented in it. Jaana Myllyluoma and Mahesh Sharma assisted in the preparation of data files and their analysis, and Pam Nall assisted in preparation of the tables. Beth Soldo, Nancy Birdsall, T. Paul Schultz, and participants in a World Bank sponsored workshop on the project held in Washington in May, 1982 provided valuable comments on an earlier version of the paper. 2. Though some of the "increase" in infant mortality rates could have resulted from improved reporting, the trends were associated with deteriorating conditions of housing and sanitation in the rapidly growing peripheral areas of larger cities (Lesser, 1974; Monteiro, 1982), and with poor nutrition resulting from inflationary pressures on the limited purchasing capacity of low income families (Iunes et al., 1975). Data from the 1975 Estudo Nacional de Despesa Familiar (ECNDEFT revealed that 37 percent of children under the age of 18 suffered from first-degree malnutrition, i.e. 76- 90 percent of nonmal weight for age (Knight and Moran, 1981:27-28). 3. The analysis is limited to the effectiveness of water projects even though the PLANASA program included both water and sanitation, because preliminary analysis indicated that in urban areas (where the program was being implemented) almost all households that had piped water also had a link to the city sanitation system, or to another "modern" sanitation system (e.g., septic tanks). Most of the cells required for testing their separate effects (water but no sanitation, or vice-versa) were empty. Care should be taken in interpreting the results of this study to avoid any implication that the contribution of safe water to increased child survival is more or less than improved sanitation, since that trade-off was not tested. 4. Brazil's national system of vital statistics was established in 1974, and data for the municipalities of state capitals are available for earlier dates. While vital statistics for Sao Paulo and a few other states are reasonably complete and consistent with estimates from the census, they still fall short of the reliability and completeness of coverage needed for analysis of national trends and differentials (Altmann and Ferreira 1979: 55). A thorough assessment of the available data on Brazilian fertility and mortality during the period from 1950 to 1976 has been carried out by the U. S. National Academy of Sciences (NAS), Committee on Population and Demography (NAS, 1979). The report on Brazil provides indirect mortality estimates derived from questions on children ever born and children surviving as reported in the 1950, 1960, and 1970 censuses and in PNAD surveys taken after 1970. Trussell's variant of the Brass model life table approach was used to generate indirect estimates (NAS, 1981) for Brazil as a whole and for each of the seven PNAD sampling regions (See Figure 1). The rural areas of frontier states (the North and Central-West macroregions, which accounted for about 5 percent of Brazil's population in 1970) were not included in the PNAD samples and are excluded from the NAS estimates. Preliminary results of the 1980 census have since been released, and it was also possible to calculate a national level estimate of the infant mortality rate from 1980 data. 26 5. A puzzling feature of the estimates derived from surveys during the 1970s is the divergence of rates for 1974-75 from those for 1972, 1973, and 1976. The 1974-75 estimates were calculated from special tabulations of the special nutrition survey (ENDEF) conducted in those years. While the ENDEF results suggest a pattern of regional differentials that is similar to the ones in census and PNAD data, the level of infant mortality is significantly higher than in adjacent PNAD surveys. When ENDEF results are compared to the 1970 census, it appears that there was no decline at the national level, and that the Northeast/Sao Paulo differential increased even more than the PNAD indicates. The ENDEF results need to be viewed with caution, especially in comparison with estimates from the census and PNAD surveys. No assessment of the comparability of ENDEF to census and PNAD data is available. It is quite possible that ENDEF is closer to the Brazilian reality in terms of the level of mortality than tfh census or PNADs. However, this does not imply that mortality levels were constant after 1970; rather, it suggests that both the census and the PNADs may have understated the true level of mortality and that the decline occurred from a higher base. 6. Household income was calculated by adding up the monetary earnings of family members (after imputing earnings to members whose earnings were reported as unknown--see Fox 1981). The 1976 survey included several types of income not explicitly covered in the 1970 census. To insure comparability, 1970 concepts have been applied in both instances. Boarders and household employees were excluded, while other related individuals are included, so that "household" income measures extended family income. 7. Estimates are included for both urban and rural women, though the latter are excluded from further analysis. Also excluded are women not related to the head of the household in which they resided, a step taken to eliminate bias arising from the inclusion of live-in domestic servants in a substantial proportion of middle and upper income Brazilian households. 8. Data on the relation between child mortality, nutrition, and access to health care in Brazil are limited to local area studies such as the ones reported by Puffer and Serrano. The one exception is the 1974-75 ENDEF survey; however, the results of that survey are available only in the form of published summary tables for large regional aggregates, and the tabulations that have been published to not include any that link survival rates to nutrition levels or other aspects of household consumption that were included in the scope of the survey. Since a pubic use file of ENDEF data has not been released, special tabulations were limited to variables included in the 1970 census and 1976 PNAD survey, which are quite limited in comparison with ENDEF. 9. Two additional problems remain with the use of OLS with these equations. One relates to measurement of the dependent variables, which as measured has statistical properties that make OLS an inappropriate estimating technique. A transformation of CM that yields the desired properties is discussed in the next section. The second concerns equation [2], whose dependent variables is WAT, a dummy variable taking values of zero and one depending on whether a household has water or not. With a dummy dependent variable, OLS can lead to predictions of values less than zero and greater than one, in addition to yielding inefficient estimates, so that logistic maximum-likelihood estimating techniques are preferred. However, the error term restriction described in the previous paragraph becomes much more complex when OLS is not used; 27 further, OLS is least likely to manifest these properties when the sample mean of the variable to be estimated falls between 0.3 and 0.7, which is the case with these data (Amemiya, 1981:1488). Therefore the decision was made to estimate equation [2] with OLS. 10. The life table values reported in the table were derived from child survival ratios in the census and PAND data, which were adjusted using multipliers based on ratios of parity at ages 15-19 to parity at ages 20-24. Owing to the limited number of births to higher income women at these ages, the parity ratios could be distorting the adjustment. The observed survival ratios and parity ratios are as follows: Tncome PI/P2 Survival PI/P2 Survival Deciles 1970 Rate 1976 Rate 1 - 3 .087 .86 .109 .88 4 - 6 .129 .86 .142 .89 7 - 9 .157 .88 .147 .93 10 .118 .93 .076 .97 The survival ratios suggest that the widening differential between the upper and lower groups does not derive from distortions in the adjustment factors, though the pattern of average parity differentials does suggest caution about the appropriateness of the assumptions of indirect measurement of life table values when populations are classified by income. 11. Two assumptions required for application of the Trussell-Preston method are violated by the Brazilian data. These relate to geographic and interclass mobility within the study population and to changes in mortality levels in the years prior to the time that retrospective questions are asked. Since the index is based on the cumlative experience of women's births and surviving children, the mortality level observed at a particular point in time will not be representative of women in a particular reference group (for example, and income class) if there has recently been significant movement of women between classes, or if mortality rates have been declining rapidly. One way to reduce the time bias is to restrict the analysis to women in their twenties, whose experience is more recent. Table 2 suggested that the experience of younger women reflected mortality rates 2-3 years before the 1970 census and 1976 PNAD survey, whereas the time reference of rates for older women was 3-5 years. In dealing with the mobility problem, Carvalho and Wood have suggested that comparisons as well as the selection of standard schedules based on groups that are broad enough to minimize mobility effects. Following this suggestion, separate standards were used in calculating the index depending on whether women were members of families that were above or below the median level of family income. 12. Zero order correlations between error terms for the endogenous variables are as follows. Coefficients are shown above the diagonal and their significance (the likelihood that the correlation is random) below it. 28 1970 1976 FY WAT CM FY WAT CM FY 1.0 -.00 .00 1.0 .00 .07 WAT .99 1.0 .00 .99 1.0 .00 CM .99 .99 1.0 .47 .99 1.0 In all but one instance the zero order correlations between the error terms and exogenous variables are not significantly different from zero. The exception is MED and the error term for FY, owing to the high correlation between MED and FED. MED is not included in the FY equation on the ground that this correlation is more related to the selectivity of the marriage process than to spouse contributions to earning capacity. 29 IBRD-17297 t\7io h4 JUNE 1983 V E N E Z U E L A 1) J~} H { F< -Kilometers 0 1,000 o | .\ a4.l | 10 PNADINAMEIGNS \M{ I oiles of PNADregio600 < j A R G E N T I N A I _ Rio de Joneiro~~~~~~~~A t n i -00 -~~..#.*see. ~ Fortaleza 2~~~~~~~~~~~~~~~~~~~~~~~~~~~*" i 9 o SdoPoul e #PoreNinaa, Ro /O NdoSl *wOq**O* ~~Pessoa $ ) 1 twt} Mmos Geruis~~~~~~~~~~~~~~~~~~0 , EsirtSa U: :-.V-. ~ 'Mocei6 s tE \ / : ~~~~~~~~~~~~~~~... ................PrbaPrabc,lgsSeipBoa 12* e , , Sal2ador 0 c a a n 0~ ~" 1~ :2 /f'RT SANTO n tetoudn hw A R AGUAY~ , C L \ Part ot The WortdRioade aanieho h( I I augnnea s -Florian6poibs PNAD REGIONS II) 72° 60°uchBoundaries of PNAD regions URUGU PNAD regions: ARGENTINA ~ ~ ~ ~ ~~n~ Rio de Janeiro ~.3i~S8o Paulo *Paran , Santa Catarina, Rio Grande do Sul Jv Minas Gerais, Espirito Santo 36" Maranhao, Piaui, Ceara, Rio Grande doaNorte, :..:1: Pa raiba, Perna mbuco, Alagoas, Sergipe, Bahia This map feas been prepared by J Distrito Federal The World Bank' saffbe exclusively NI.4 Rond6nia, Acre, Amazonas, Roraima, Pari5, for ihe convenhence ot ithe readers and is exciusiveely for tie Amap6, Mato Grosso, Goidis, MatoGroisso internai use of The Wofld Bank and the interna tonal Finance do Sul Corporahon The denom,nai,ons * Ntoa aia used and the boundarnes s"hownNtinlcaia on this map dfo not imply. on the State or territory capitals part of The World Bank and mhe lnlernatonel Fnance Corporation, State or territory boundaries any judgmenion" the legal status -International boundaries of aIny territory or any endforsement or acceptance of 600such boundaries, 480 360 Statistical Appendix Table 1: Estimates of Infant Mortality Derived from Census and Survey Data Using Model Life Table Techniques A/, Brazil and PNAD Regions, 1950-80. Region 1950 1960 1970 1972 1973 1974-5 1976 1980 1. Rio de Janeiro 128.0 92.2 92.2 6.9.8 69.1 84.7 60.9 69.8 2. Sao Paulo 125.5 92.3 87.5 66.1 64.6 79.4 66.2 64.3 3. Southern States 112.3 88.8 82.8 67.9 66.4 74.1 68.7 76.4 4. Minas Gerais, Espirito Santo 138.2 111.5 99.5 79.2 79.6 93.0 74.4 79.4 w tA 5. Northeastern States 193.7 184.4 139.6 141.1 140.5 148.4 121.9 116.4 6. Brasilia - - 129.0 100.9 80.0 78.6 84.3 65.4 66.9 7. Amazon States 129.2 105.0 94.6 - - - - - - - - 71.6 Brazil b/ 151.7 122.7 107.1 99.9 97.8 107.6 89.2 86.2 Source: NAS (1979) and Preliminary Tabulations of 1980 Census. Notes: a) Coale-Demeny West for 1950, South thereafter. b) Excludes rural areas of Amazon states. Table 2: Estimated Probabilities of Death By Ages Two (2qA) and Three (3qO) by Household Income Deciles, Rural and Urban Areas, 1970 and 1976, Approximate Reference Point (t) in Years Prior to Interview. Household 1970 1976 Income ercent Percent Percent Deciles 2qO (t) 3q) (t) of 2qO (t) (t) of Decline Women Women Urban 1 - 3 .191 (2.13) .217 (3.77) 10.8 .137 (2.53) .161 (4.19) 12.2 27% 4 - 6 .143 (2.19) .144 (4.08) 16.5 .104 (2.11) .108 (4.02) 18.6 26% 7 - 10 .097 (1.93) .103 (3.29) 35.8 .066 (1.86) .061 (3.25) 37.7 37% Rural 1 - 3 .154 (2.22) .166 (4.13) 17.8 .138 (2.43) .139 (4.32) 13.4 13% 4 - 10 .131 (1.90) .126 (3.39) 19.1 .106 (1.96) .104 (3.41) 18.1 18% TOTAL .133 (2.06) .138 (3.78) 100.0 .099 (2.16) .100 (3.83) 100.0 26% Source: Tabulations of 1970 and 1976 sample files. Table 3: Proportion of Children Who Have Died by Mother's Age and Years of Schooling, Urban Brazil, 1970 and 1976. Educ. None 1 - 4 5 - 9 10+ All Groups Age 1970 1976 1970 1976 1970 1976 1970 1976 1970 1976 20-24 .201 .193 .106 .098 .092 .053 .034 .029 .131 .094 25-29 .192 .134 .107 .093 .083 .071 .063 .029 .127 .089 30-34 .220 .211 .127 .096 .114 .085 .048 .019 .150 .110 35-39 .219 .193 .170 .110 .127 .067 .066 .029 .176 .121 40-44 .262 .222 .168 .119 .076 .076 .055 .068 .184 .143 45-49 .301 .232 .180 .145 .127 .082 .094 .052 .224 .163 Percent 26.8 16.6 47.3 46.6 19.0 23.4 6.9 13.4 100. 100. Source: Observations from 1970 and 1976 sample files. Table 4: Ratio of Observed Deaths to Expected Deaths for Mothers Aged 20-29, by - Years of Schooling/Income Decile and Type of Access to Water, 1970 and 1976. 1970 1976 Percent Distribution Water Supply Internal External Internal External Piped City Well/ Other Piped City Well/ Other 1970 1976 System Spring System Spring Education: None 0.99 1.22 1.17 1.40 0.73 1.43 1.10 1.54 34.8 12.3 1-4 0.59 0.83 0.83 1.09 0.46 0.70 0.83 1.18 46.1 43.1 5-9 0.42 0.80 0.93 0.78 0.31 0.43 0.69 0.69 14.2 28.4 10+ 0.30 0.84 * 0.78 0.20 * 0.32 * 4.9 16.2 Income Decile: 1-3 0.84 1.01 1.04 1.30 0.57 0.90 1.07 1.46 -- 30 -- 4-6 0.63 0.80 0.99 1.21 0.37 0.53 1.12 1.39 -- 30 -- 7-10 0.53 1.07 0.81 1.10 0.29 0.40 0.58 0.67 -- 40 -- All Groups 0.58 0.95 0.95 1.24 0.37 0.77 0.85 1.22 100.0 100.0 N 3722 825 2086 3455 6312 1118 1325 987 10088 9742 Percent 36.9 8.2 20.7 34.2 64.8 11.5 13.6 10.1 100.0 100.0 Source: Observations from 1970 and 1976 sample files. Note: A "*" indicates that cell had fewer than 25 cases. Table 5: Weighted Means and Standard Deviations of Variables in Analysis of Child Mortality in Urban Brazil, 1970-1976. Variables Unit 1970 1976 mean st. dev. mean st. dev. (1) (2) (3) (4) Endogenous: CM Standardized Index 1.264 2.263 0.852 1.639 WAT Dummy variable 0.306 0.523 0.544 0.499 (=1 when household had piped water) FY Log of 1970 Cruzeiros 5.104 1.300 6.091 0.885 Exogenous: MED Log of Number of Years 0.842 0.920 1.356 0.780 of School Completed (+1)* FED 0.931 0.937 1.416 0.785 TAX Log of 1970 Cruzeiros 4.583 0.991 4.849 0.844 WS Ratio of households in region with piped water 0.419 0.240 0.467 0.191 to total households in region. PWS Change in above between 1970 and 1976. - 0.144 0.057 Note: To avoid taking the logaritum of zero, one was added to MED and FED in the log transformation of these variables. 37 Table 6: Ordinary Least Squares Regression Estimates, Reduced Form and Structural Equations [1-3], 1970 Data. INDEPENDENT DEPENDENT VARIABLE!/ VARIABLE FY WAT CM (1) (2) (3) (4) (5) (6) CONSTANT 3.318 -.070 -.184 2.288 2.542 2.493 MED 0.088 0.081 -0.270 -0.252 -0.231 (13.4) (12.1) (8.4) (7.8) (7.1) FED 0.579 0.095 0.082 -0.184 -0.156 -0.134 (46.9) (14.1) (12.0) (5.7) (4.7) (4.0) TAX 0.272 -0.013 -0.018 -0104 -0.093 -0.098 (23.3) (1.6*) (2.1*) (2.6) (2.3*) (2.4*) WS 0.653 0.637 -0.350 -0.314 -0.145 (18.1) (17.7) (2.0*) (1.8*) (0.8*) FY 0.032 -0.071 -0.063 (7.8) (3.6) (3.1) WAT -0.264 (5.6) R2 .277 .237 .241 .046 .047 .050 F-Y 1977.9 804.1 659.0 124.9 102.6 90.9 Notes: a) The t statistic is shown in parentheses under the coefficients for independent variables. All coefficients are significant at the .01 level unless an (*) is noted. b) Degrees of freedom for the F test are (NV,10356 - NV -1), where NV = the number of independent variables. 38 Table 7: Ordinary Least Squares Regression Estimates, Reduced Form and Structural Equations [1-3], 1976 Data. INDEPENDENT DEPENDENT VARIABLEAI VARIABLE FY WAT CM (1) (2) (3) (4) (5) (6) CONST 4.121 -0.131 -0.694 2.493 2.626 2.424 MED 0.127 0.082 -0.234 -0.223 -0.200 (18.5) (12.0) (9.2) (8.5) (7.5) FED 0.588 0.158 0.092 -0.261 -0.245 -0.219 (63.6) (23.0) (13.0) (10.2) (8.9) (7.9) TAX 0.234 -0.061 -0.015 -0.131 -0.120 -0.151 (27.1) (3.9) (7.0) (2.3*) (2.1*) (2.6) WS 0.962 1.010 -0.411 -0.263 -0.129 (13.6) (14.8) (1.6*) (1.6*) (0.5*) PWS 0.865 0.727 -0.890 -0.857 -0.646 (10.3) (8.9) (2.8) (2.7) (2.1*) FY 0.153 -0.036 -0.008 (24.6) (1.5*) (0.3*) WAT -0.290 (7.5) R 2 .375 .271 .314 .067 .067 .073 Fk/ 2920.3 725.6 743.4 140.0 117.1 109.0 Notes: a) The t statistic is shown in parentheses under the coefficients for independent variables. All coefficients are significant at the .01 level unless an (*) is noted. b) Degrees of freedom for the F test are (NV,9767 - NV - 1), where NV = the number of independent variables. 39 Table 8: Total, Direct and Indirect Effects of Variables in Equation [3) on CM, 1970 and 1976. Variable Total Indirect Effect Direct Effect FY WAT Effect (1) (2) (3) (4) 1970: MED -.1099 -.0073 -.0087 -.0939 FED -.0762 -.0117 -.0090 -.0555 TAX -.0457 -.0049 +.0021 -.0429 WS -.0371 -.0038 -.0178 -.0155 FY -.0408 a/ -.0048 -.0360 WAT -.0612 a/ a/ -.0612 1976: MED -.1113 -.0050 -.0114 -.0949 FED -.1250 -.0073 -.0130 -.1047 TAX -.0675 -.0055 +.0158 -.0778 WS -.0479 +.0013 -.0342 -.0150 PWS -.0311 -.0012 -.0073 -.0226 FY -.0195 a/ -.0150 -.0045 WAT -.0882 a/ a/ -.0882 Note: a There is no indirect effect for the variable. 40 Table 9: Tests for Interaction Effects in Equation [3a], Dependent Variabl e is CM, 1970 and 1976 Data. INDEPENDENT 1970 1976 VARIABLE (1) (2) (3) (4) (5) (6) CONST 2.493 2.521 2.560 2.424 2.477 2.812 MED -0.231 -0.254 -0.234 -0.200 -0.227 -0.199 (7.1) (6.8) (7.2) (7.5) (6.8) (7.5) FED -0.134 -0.137 -0.136 -0.219 -0.221 -0.217 (4.0) (4.1) (4.1) (7.9) (8.0) (7.8) TAX -0.098 -0.099 -0.100 -0.151 -0.149 -0.146 (2.4) (2.4) (2 5) (2.6) (2.6) (2.5) WS -0.145 -0.127 -0.123 -0.129 -0.128 -0.122 (0.8*) (0.7*) (0.5*) (0.5*) (0.5*) (0.5*) PWS - - - -0.646 -0.644 -0.617 (2.1*) (2.0*) (2.0*) FY -0.063 -0.065 -0.082 -0.008 -0.003 -0.067 (3.1) (3.3) (3.4) (0.3*) (0.1*) (1.6*) WAT -0.264 -0.338 -.554 -0.290 -.0372 -0.894 (5.6) (4.6) (2.8) (7.5) (5.3) (3.3) INT - 0.069 0.054 - 0.063 0.102 (1.3*) (1.5*) (1.4*) (2.2*) R2 .050 .050 .050 .073 .073 .073 Fk/ 90.9 78.2 78.2 109.0 95.6 96.0 Notes: a) The t statistic is shown in parentheses under the coefficients for independent variables. All coefficients are significant at the .01 level unless an (*) is noted. b) Degrees of freedom for the F test are as follows: 1970 data: (NV,10356 - NV - 1); 1976 data: (NV,9767 - NV - 1), where NV = the number of independent variables. 41 Table 10: Evaluation of OLS Estimate of Equation [3] to Decompose Change in CM from 1970 to 1976. Coefficient x 1970 x 1976 Change Percent means means (3) - (2) (4)/Total (1) (2) (3) (4) (5) 1970 CONSTANT 2.5654 2.5654 2.5654 - - MED -.2316 -.1949 -.3140 -.1191 34.4 FED -.1377 -.1282 -.1949 -.0667 19.3 TAX -.1249 -.5724 -.6054 -.0332 9.6 FY -.0633 -.3231 -.3856 -.0624 18.0 WAT -.2714 -.0830 -.1478 -.0648 18.7 Total (CM) - 1.2637 0.9175 -.3462 100.0 1976 CONSTANT 2.3738 2.3738 2.3738 - - MED -.1972 -.1660 -.2674 -.1014 32.0 FED -.2165 -.2016 -.3064 -.1048 33.1 TAX -.1688 -.7736 -.8185 -.0449 14.2 FY -.0051 -.0260 -.0311 .0050 (1.6) WAT -.2954 -.0903 -.1609 -.0705 22.3 Total (CM) - 1.1683 0.8517 -.3166 100.0 42 APPENDIX TABLE: Zero Order Correlation Coefficients, Variables in Table 5. 1970a/ MED FED TAX WS FY WAT CM 1976 MED 1.00 .589 .188 .189 .514 .389 - .205 FED .673 1.00 .221 .227 .572 .421 - .215 TAX .292 .341 1.00 .936 .339 .317 - .155 WS .350 .419 .843 1.00 .312 .336 - .150 FY .441 .488 .350 .378 1.00 .481 - .178 WAT .368 .392 .331 .404 .312 1.00 - .196 CM -.187 -.181 -.135 -.146 -.144 -.149 1.00 PWSb/ -.058 -.060 -.186 -.317 -.009 .025 .010 Notes: a) Coefficients for 1970 are shown above the diagonal; those for 1976 below it. b) PWS measured in 1976 only, so no diagonal shown. References Almeida, Wanderly J. M. 1977. Abastecimento de agua a populacao urbana: uma avaliacao do PLANASA. Rio de Janeiro: IPEA/INPES. Altmann, Ana Maria G., and Ferreira, Carlos E. 1979. "Evolucao do Censo Demografico e Registro Civil como Fontes de Dados para a Analise da Fecundidade e Mortalidade no Brasil," Boletim Demografico 10 (Out./Dec.): 1-85. Amemiya, Takeshi. 1982. "Qualitative Response Models: A Survey," The Journal of Economic Literature 19:1483-1535. Anderson, John E. 1981. "Survey of Maternal and Child Health and Family Planning in Northeastern Brazil:, Measurement of the Duration of Breastfeeding and Postpartum Amenorrhea," U. S. Department of Health and Human Services, Centers for Disease Control, Working Paper. Antonovsky, Aaron. 1980. "Implications of Socio-Economic Differentials in Mortality for the Health System." In Proceedings of the Meeting on Socioeconomic Determinants and Consequences of Mortality, Mexico City, June 19-25, 1979. Geneva: World Health Organization. Arriaga, Eduardo. 1980. "La Mortalite des Infants dans quelques Pays D'Amerique Latine." In P.M. Boulanger and D. Tabutin (eds.) La Mortalite des Infants dans le Monde et dans L'Histoire. Liege: Ordina Editions, pp. 223-257. Behm, Hugo. 1980. "Socioeconomic Determinants of Mortality in Latin America." In Proceedings of the Meeting on Socioeconomic Determinants and Consequences of Mortality, Mexico City, June 19-25, 1979. Geneva: World Health Organization. Brazil, Fundacao IBGE. 1971. Anuario Estatistuo do Brasil - 1971. Rio de Janeiro: Fundacao IBGE. Brazil, Fundacao IBGE. 1979. Amostra de 1% dos Registros do Censo Demografico de 1970--Manual do Usuario. Rio de Janeiro: Fundacao IBGE. Brazil, Fundacao IBGE. 1981a. 0 Quadro da Mortalidade por Classes de Renda: Um Estudo de Diferenciais nas Regioes Metropolitanas. Rio de Janeiro: Fundacao IBGE. Brazil, Fundacao IBGE. 1981b. Metodologia da Pesquisa Nacional por Amostra de Domicilios na Decada de 70. Rio de Janeiro: Fundacao IBGE Caldwell, J.C. 1980. "Education as a Factor in Mortality Decline. An Examination of Nigerian Data." In Proceedings of the Meeting on Socioeconomic Determinants and Consequences of Mortality, Mexico City, June 19-25, 1979. Geneva: World Health Organization. 44 Carvalho, Jose Alberto M., and Sawyer, Diana 0. 1977. "Diferenciais de Mortalidade no Brasil." Campos do Jordao, Brazil: Anais, Primeiro Encontro National, Associacao Brasileiro de Estudos Populacionais, 231-259. Carvalho, Jose Alberto M. and Wood, Charles. 1978.. "Mortality Income Distribution and Rural-Urban Residence in Brazil." Population and Development Review 4: 405-20. Cochrane, Susan H., O'Hara, Donald J. and Leslie, Joanne. 1980. The Effect of Education on Health. Washington D.C.: The World Bank, Staff Working Paper No. 405. Dyson, Tim. 1978. "Levels, Trends, Differentials and Causes of Child Mortality--A Survey." World Health Statistics Reports 30: 282-311. Fox, M. Louise. 1981. "The Adjusted Household Income Distribution in Brazil, 1970." World Bank, Development Economics Department, Working Paper for Project, 672-21. Gwatkin, Davidson R. et al. 1980. Can Health and Nutrition Interventions Make a Diference? Washington, D.C.: Overseas Development Council. Iunes, M. et al. 1975. Estudo Natricional de Criancas de 6 a 60 Meses no Municiplo de Sao Paulo. Sao Paulo: Instituto de Medicina Preventiva de Escola Paulista de Medicina e Instituto de Pesquisas Economicas, Universidade de Sao Paulo. Knight, Peter T. and Moran, Ricardo. 1982. Brazil: Poverty and Basic Needs Series. Washington, D.C.: World Bank. Leser, Walter. 1974. "Crescimento da Populacao e Nivel de Saude na Cidade de Sao Paulo." Problemas Brasileiros 16 (October): 17-36. Lluch, Constantino. 1981. "Income Distribution and Family Characteristics in Brazil." World Bank, Development Economics Department, Working Paper for Project 672-21. Loriaux, Michael. 1980. "La Mortalite des Infants et les Indicateurs Socio- Economiques du Developpement: Une Vision Mondiale." In P.M. Boulanger and D. Tabutin (eds.) La Mortalite des Infants dans le Monde et dans L'Historie. Liege: Ordina Editions, pp. 287-386. Marsden, Parker V. (ed.), 1981. Linear Models in Social Relations. Beverly Hills, California: Sage Publications. Monteiro, C. A. 1982. "Contribuicao para o Estudo do Significado da Evolucao da Coeficiente de Mortalidade Infantil no Municipio de Sao Paulo, SP (Brasil) nas Tres Ultimas Decadas (1950-1979)," Revista da Saude Publica. 16: 7-18. NAS - Committee on Population and Demography, Panel on Brazil, National Academy of Sciences. 1983. Levels and Recent Trends in Fertility and Mortality in Brazil. Washington, D.C.: National Academy Press. 45 NAS - Committee on Population and Demography, National Academy of Sciences. 1981. Demographic Estimation: A Manual of Indirect Techniques, Washington, D.C.: National Academy Press. Pfeffermann, Guy and Webb, Richard. 1979. The Distribution of Income in Brazil, Washington, D.C.: The World Bank, Staff Working Paper No. 356. Puffer, Ruth R. and Carlos V. Serrano. 1973. Patterns of Mortality in Childhood. Washington, D.C.: Pan American Health Organization. Schneider, R. E., Shiffman, M., and Faigenblum, J. 1978. "The Potential Effect of Water on Gastrointestinal Infections Prevalent in Developing Countries," The American Journal of Clinical Nutrition. 31 (November): 2089-2099. Schultz, T. Paul. 1980. "Interpretation of Relations Among Mortality, Economics of the Household, and the Health Environment." In Proceedings of the Meeting on Socioeconomic Determinants and Consequences of Mortality, Mexico City, June 19-25, 1979. Geneva: World Health Organization. Trussell, James and Preston, Samuel. 1982. "Estimating the Covariates of Childhood Mortality from Retrospective Reports of Mothers." Health Policy and Education 3: 1-36. Vallin, Jacques. 1976. "World Trends in Infant Mortality Since 1950". World Health Statistics Reports 29: 646-674. Wood, Charles. 1977. "Infant Mortality and Capitalist Development in Brazil," Latin American Perspectives 4 (Fall): 56-65. The World Bank. 1980. Water Supply and Waste Disposal. Washington: The World Bank. Yunes, Joao and Ronchezel, Vera S.C. 1974. "Evolucao da Mortalidade Geral, Infantil e Proportional no Brasil." Revista de Saude Publica 80 (Suplemento): 3-48. 46 A World Bank Paper. March 1976. 98 Number 1: Appropriate World Bank pages (including 4 annexes). English, Sanitation Alternatives: Publications French, and Spanish. A Technical and Economic Stock Nos. PP-7602-E, PP-7602-F, Appraisal Df Related PP-7602-S. $5.00. John M. KRalbermatten, Interest DeAnne S. Julius, and Village Water Supply: Charles G. Gunnerson Economics and Policy in the This volume summarizes the techni- Developing World cal, economic, environmental, health, and sociocultural Findings of Robert J. Saunders the World Bank's research program and Jeremy J. Warford on appropriate sanitation alternatives Addresses the problem of potable and discusses the aspects of pro- water supply and waste disposal in gram planning that are necessary to rural areas of developing countries implement these findings. It is where the majority of the poor tend directed primarily toward planning to be found. Emphasizes the officials and sector policy advisers economic, social, financial, and for developing countries. administrative issues that character- The Johns Hopkins Uniuersity Press, ize village water supply and sanita- 1982. 172 pages (including bibliogra- tion programs. phy, index). The Johns Hopkins University Press, LC 80-8963. ISBN 0-8018-2578-4 1976. 292 pages (including 4 appen- $12.95 (i9.10) paperback. dixes, bibliography, index). LC 76-U"758. ISBN 0-8018-1876-1, $21.00 (412.70) hardcover. Nuntber 2: Appropriate ,aboratory Testing, Field Sanitation Alternatives: dlals, and Technological French: L'alimentation en eau des A Planning and Design communaut6s rrls cnmee lnigadDsg levelopment politique s rurales: lconomie et Manual ontains results of laboratory tests gdeloppement. Economica, 1978. John M. KNalbermatten, 3rried out on twelve hand pumps for DPv-Anneent.SconJucai1978 ie World Bank and presents recom- ISBN 2-7178-0022-0, 45 francs. Charles G. Gunnerson, and iendations for improvements in per- Spanish: Agua para zonas rurales y D. CDuncan Mara )rmance, safety, and durability. poblados: economfa y polDtica en el ural Water Supply Handpumps Project mundo en de5arrollo. Editorial This manual presents the latest fleld mundo n desrroll. Ediorialresults of the research, summarizes eport Number 1. March 1982. 122 Tecnos, 1977. selected portions of other publica- Rges (Including 3 appendixes). ISBN 84-309-0708-4, 575 pesetas. tions on sanitation program plan- tock No. WS-8202. $5.00. ning, and describes the engineering details of alternative sanitation World Bank Studies in Water technologies and how they can be lunicipal Water Supply Supply and Sanitation upgraded. 'roject Analysis: Case The United States has designated the The Johns Hopkins Uniuersity Press, itudies 1980s as the International Drink- 1982.172 pages (including bibliogra- rank 11. Lamson-Scribner, Jr., ing Water Supply and Sanitation phy, index). nd John iHuang, editors Decade. Its goal is to provide two of LC 80-8963. ISBN 0-8018-2584-9 ight case studies and fourteen exer- needs-safe water and sanitary $15.00 (S.10.50) paperback. sastealin with the water and disposal of human wastes-to all people. To help usher in this impor- (orld Bank (EDI), 1977. ix + 520 tant period of international research Appropriate Technology for ages. (Available from ILS,1715 Con- and cooperation, the World Bank Water Supply and Sanitation ecticutAvenue, N.W., Washington, published two volumes on appro- ,.C. 20009, U.S.A.) priate technology for water supply Volume 1: Technical and and waste disposal systems in Economic Options 8.50 paperback. developing countries. Since the John M. Kalbermatten, technology for supplying water Is John S. Julbes,and Wage Water Supply ~better understood, the emphasis in DeneSJuisad Illage Water Supply btthese volumes is on sanitation and Charles G. Gunnerson escribes technical aspects, costs, waste reclamation technologies, their Reports technical, economic, health, nd institutional problems related to contributions to better health, and and social findings of the research upplying water for domestic use in how they are affected by water ser- project on "appropriate technology" ural areas and proposes guidelines vice levels and the ability and willing- and discusses the program planning )r future World Bank lending in this ness of communities to pay for the ector. systems. necessary to implement technologies Volume 3: Health Aspects of Volume 10: Night-Soil available to provide socially and Excreta and Sullage Manage- Composting environmentally acceptable low-cost water supply and waste disposal. ment-A StatCf-the-Art Halles 1. Shuval, Decmbe 190.12 pges(inludng Review Charles G. Gunnerson, December 1980. 122 pages (including Richard G. Feachem, and DeAnne S. Julius bibliography). David J. Bradley, Describes a safe, inexpensive treat- Stock No. WS-8002. free of charge. Hemda Garelick ment method for night-soil compost- and D. Duncan Mara ing that is ideally suited for developing countries because of its Volume la: A Summary of Provides information on the ways in simplicity in operation, limited need Technical and Economkc which particular excreta disposal and for mechanical equipment, low cost, Options reuse technologies affect the survival and its effectiveness in inactivating John M. Kalbermatten, and dissemination of pathogens. It is pathogens. intended for planners, engineers, DeAnne S. Julius, and economists, and health workers. December 1981. 81 pages (including Charles G. Gunnerson December 1980. 303 pages (including bibliography, 2 appendixes). A summary of the final report on 14 appeindixes, references), Stock [1o. WS-8101. $3.00. appropriate technology for water sup- Stock I. WS-8005 $15.00 Volume . A Sanitation ply and waste disposal In developing SokN.W-05 1.0 outl:ASntto countries, a World Bank research pro- Field Manual ject undertaken by the Energy, Water, Volume 4 Low-Cost John M. Kalbermatten, and Telecommunications Department Technology Options for DeAnne S. Julius, and Sanitation-a State-of-the- Charles G. Gunnerson December 1980. 38 pages. English, Art Review and Annotated December 1980. 87 pages. Bibliography Stock No. WS-8007. $3.00. Stock No. WS-8003. $3.00. Witold Rybczynski, Chongrak Polprasert, and A Model for t Volume Ib: Sanitation Michael McGarry of a Self-Help Water Supply Alternative for Low-Income A comprehensive bibliography that Program Communities-A Brief describes alternative approaches to Colin Glennie Introduction the collection, treatment, reuse, and D. Duncan Mara disposal of wastes. Presents one version of a practical Describes, in non-technical language, A joint World Bank/lInternational munity participation, water supply the various low-cost sanitation Development Research Centre publica- programs in developing countries. technologies that are currently avail- . tion. 1978. Available from Intemational Consideration is also given to sanita- able for low-income communities in Development Research Centre (IDRC), tion program development and prac- developing countries and presents a P.O. Box 8500, Ottawa KIG 3119, tical guidelines for program general methodology for low-cost Ontario (Canada). development are included. One of a sanitation program planning. series of informal Working Papers pre- pared by the Technology Advisory February 1982. 48 pages. Volume 5: Sociocultural Group, established under UNDPs Stock No. WS-8201. $3.00. Aspects of Water Supply and Global Project, executed by the World Excreta Disposal Bank. Volume 2: A Planner's Guide Mary Elmendorf and Technology Advisory Group Working John M. K'albermatten, - Patricia Buckles Paper [umber 1. October 1982. 45 DeAnne S. Julius, This report was prepared as part of pages (including 2 annexes, D. Duncan Mara, the World Bank research project con- references). and Charles G. Gunnerson cerning appropriate technology for ISB[Y 0-8213-0077-6. $3.00. water supply and waste disposal. Provides information and instructions Social and cultural factors influencing World Bank Research in on how to design and implement responses to water supply and excreta appropriate technology projects disposal technologies are investi- Water Supply and Sanita- based on the findings reported in gated in seven case studies of com- tion-Summary of Selected Volume 1: Technical and Economic munities in the rural and urban fringe Publications Options. areas of Latin America. A bibliography summarizing the December 1980.194 pages (Includitg December 1980. 67 pages (including 3 papers in the Water Supply and references). annexes, references). English and Sanitation Series, as well as the World Stock [1o. WS-8004. $5.00. Spanish. Bank studies in Water Supply and Sanitation published for the Stock [1o. WS-8006. $3.00. World Bank by The Johns Hopkins University Press. NIovember 1980. Stock [o. WS-8000. Free of charge. Confronting Urban these studies is presented In Malnutrition: The Design of Migration in WestAfrica: N-utrition Programs Demographic Aspects. James E. Austin World Bank Staff Working Paper No. Describes a framework for 415. September 1980. vl + 385 pages systematically carrying out urban (including statistical annexes, nutrition programs that examines bibliography). several key considerations in nutri- Stock No. WP-0415. $15.00. tion education, on-site feeding, take- home feeding, nutrient-dense foods, ration shops, food coupons, fortiflca- Economic Motivation versus tion, direct nutrient dosage, and food City Lights: Testing processing and distribution. Hypotheses about Inter- The Johns Hopkins University Press, Changwat Migration in 1980.136 pages. ThaiFland LC 79-3705. ISBN 0-8018-2261-0, Fred Arnold and $6.50 (24.50) paperback. Susan H. Cochrane nzw World Bank Staff Working Paper No. The Costs and Beneflts of 416. September 1980. 41 pages Analyzing the Impact of Family Planning Programs (including footnotes, references). Health Services: Project George C. Zaidan Stock No. WP-0416. $3.00. Experiences from India, A technique for measuring the Ghana, and Thailand economic returns from Investing In Economics of Supplemental Rashid Faruqee population control, with an appraisal Feeding of Malnourished of Inherent assumptions and Reviews four categories of health liniltations. Children: Leakages, Costs, Indicators (environment, services and Beneflts offered, services received, and The Johns Hopkins University P Odin K. Knudsen changes In mortality, morbidity, and. 1971. 62 pages (Including nutridonal status) In order to evaluate. bibliography). Analyzes some of the economic the Impact of health projects In India, LC 70-155166. ISBN 0-8018-1317-4 Issues Involved In the supplemental Ghana, and Thailand. S4.00 (f2.40) paperback. feeding of malnourished children. Demonstrates that supplemental World Bank Staff Working Paper feeding programs are economically No. 546.1982. 44 pages. Demographic Aspects of Justifled If minimum Improvements ISBNY 0-821.3-011 7-9. $5.00. In mortality rates and more substan- Migration In West Africa tial Increases In productivity take -Volume 1 place. K. C. Zachariah World Bank Staff Working Paper No. nEW and others 451. April 1981. iu + 76 pages. Background data on migration In Stock No. WP-0451. $3.00. Benefits and Costs of Food four English-speaking countries: Benefltsand Cost of Food Ghana, Sierra Leone, Liberia, and Distribution Policies: The The Gambia. A regional analysis Experiments in Family India Case based on these studies is presented Planning: Lessons from Fasquale L. Scandizzo and In Migration in West Africa: the g Qurushri Swamy Demographic Aspects. Developing World Word ankStff oringPaer o. Roberto Cuca and Analyzes some of the characteristics World Bank Staff Working Paper No. Catherine S. Pierce and the main consequences of the 414. September 1980. vl + 363 pages food distribution polilcies followed by (including statistical annexes, A comprehensive review of experi- the Indian govemment and provides bibliography), mental efforts in the developing a quantiflcation and a cost-beneflt world to determine more effective analysis of their effects on con- Stock No. WP-0414. $15.00. ways of providing family planning sumers, producers, and the govern- services. ment budget. Demographic Aspects of V7- Johns Hopkins University Press, World Bank Staff Working Paper Migration in West Africa 19/8. 276 pages (including bibliogra- No. 509. August 1982. 54 pages. -Volume 2 phy, index of experiments). ISBN 0-8213-0011-3. $3.00. K. C. Zachariah LC 77-16596. ISBN 0-8018-2013-8, and others $19.50 (£71.50) hardcover; ISBN 0-8018-2014-6, $8.95 (£4.00) Background data on migration Inpaebc four French-speaking countries: Ivory paperback. Coast, Upper Volta, Senegal, and lbgo. A regional analysis based on Family Planning Programs: Health NEW An Evaluation of Experience Fredrick Golladay, Roberto Cuca coordinating author Integrating Family Planning World Bank Staff Working Paper No. Draws on experience gained from with Health Services: Does 345. July 1979. xii + 134 pages health components of seventy World It Help? (including 2 annexes, references). Bank projects in forty-four countries Rashid Faruqee between 1975 and 1978. Stock No. WP-0345. $5.00. Emphasizes the disproportionately Analyzes the findings of an experi- high expenditures incurred on cura- ment carried out in Narangwal, a tive medicine, maintenance of expen- village in Punjab, India, between 196E Fertility and Education: sive hospitals, and sophisticated and 1974 related to health care and What Do We Really Know? training of medical personnel at the family planning. The World Bank col- Susan 11. Cochrane cost of preventive care for the laborated with The Johns Hopkins A modl idetifyig themany han- majority of the people. Points out University in analyzing this data from A model identifying the many chan- that low-cost health care systems are one of the best known and well-docu- nels through which education might feasible and recommends that the mented fleld experiments in health act to dene fertion Bank begin regular and direct lend- care and family planning in the world. of the evidence of the relationinfohel,inadtntoavgI between education and the interven- ing for health, in addition to having World Bank Staff Working Paper ing variables in the model that affect projects in other sectors. No. 515. September 1982. 47 pages. fertility. ISBN 0-8213-0003-2. $3.00. The Johns Hopkins University Press, Sector Policy Paper. February 1980. 1979. 188 pages (including bibliogra- 90 pages (including 8 annexes, phy, index). 4 figures, map). English, French, Kenya: Population LC 78-260 70. ISBN 0..8018-2140.1, Japanese, Spanish, andArabic. and Development $6.95 (f4.75) paperback. Stock Nos. PP-8001-E, PP-8001-F, (See description under Country PP-8001-J, PP-8001 -S, PP-8001 -A. Studies listing.) $5.00. Fertility and Its Regulation Malnourished People: In Bangladesh Health Issues and Policies A Policy View R. Amin and in the Developing Countries (See description under Development Rashid Faruqee Fredrick Golladay listing.) World Bank Staff Working Paper No. World Bank Staff Working Paper No. 383. April 1980. iu + 50 pages 412. August 1980. ii + 53 pages. Malnutrition and Poverty: (including references). Stock No. WP-0412. $3.00. Magnitude and Policy Stock No. WP-0383. $3.00. Options Shlomo Reutlinger and NEW NEW Marcelo Selowsky Health, Nutrition, and The first large research effort in the Food Distribution and Family Planning in India: A World Bank to determine the global Nutrition Intervention: The Survey of Experiments and dimension of malnutrition. Case of Chile Special Projects The Johns Hopkins University Press, Lloyd Harbert and Rashid Faruqee and 1976; 2nd printing, 1978. 94 pages Pasquale L. Scandizzo Ethna Johnson (including 5 appendixes). The impact of Chile's Complementary Surveys fourteen experiments and LC 76-17240. ISBN 0-8018-1868-0, Feeding Program (CFP), both on the special projects in health, nutrition, $4.75 (S2.85) paperback. direct and indirect beneficiaries, is and family planning in India and pro- Spanish: Desnutricion y pobreza: analyzed. Describes Chile's major poses guidelines for future Bank magnitudes y opciones de polftica. nutrition Intervention programs and projects on the basis of the survey. Editorial Tecnos, 1977. establishes the relative importance of the CFP in terms of budgetary expen- World Bank Staff Working Paper ISBN 84-309-0726-2, 380 pesetas. ditures and number of beneficiaries No. 507. February 1982. xi + 97 pages reached. Reviews briefly the pro- (including references). Measuring Urban Malnutri- grams past limitations, recent Stock No. WP-0507. $5.00. tion and Poverty: A Case reforms, and potential effectiveness. Study of Bogota and Call, World Bank Staff Working Paper No. Colombia 512. May 1982. v + 45 pages (including Rakesh Mohan, bibliography, annex). M. Wilhelm Wagner, ISBN 0-8213-0001-6. $3.00. and Jorge Garcia Attempts to measure the extent of malnutrition and poverty in the cities of Bogota and Cali, Colombia. One of flve papers resulting from a Population Policies and REPRINTS research program entitled "City Economic Development Health Care In the Developing World: Study," a study of the workings of Timothy King and others Problems of Scarcity and Choice five major urban sectors in (Shattuck Lecture) Colombia. The English-language edition is John R. Evans, Karen Lashman Hall, and World Bank Staff Working Paper No. out of print. Jeremy J. Warford World Bank Reprint Series: NIumber 209. 447. April 1981. 80 pages (including Spanish: Polifticas de poblaci6n y Reprinted from New Englaind Journal of Medicine, bibliography, appendixes). desarrollo econ6mico. Editorial vol. 305 (November 1981):1117-27. Stock No. WP-0447. $3.00. Tecnos, 1975. Stock /to. RP-0209. Free of charge. ISBN 84-309-0605-3, 440 pesetas. Measurement of Deprivation and Migration in West Africa: Poverty Based on the Proportion Demographic Aspects Population Policy and Spent on Food: An Exploratory K. C. Zachariah and Exercise Juie C.Zchai Family Planning Programs: v.v. Bhanoji Rao Julien Conde Trends in Policy and World Bank Reprint Series: Nlumber 193. The first study of the large-scale Administration Reprinted from World Development. vol. 9, no. 4 movement of people in nine West Kandiah Kanagaratnam and Stock 1 3o. RP-01953. Free of charge. African countries. Discusses the 5okN.R-13 reo hre volume and direction of internal and Catherine S. Pierce external flows and the economic and World Bank Staff Working Paper No. lNutrition, Health, and Education: The social characteristics of migrants. 411. August 1980. iii + 22 pages Economic Significance of Complementarlties at Early Age 4 joint World Bank-OECD study. Oxford (including footnotes). Marcelo Selowsky University Press, 1981. 166 pages Stock No. WP-0411. $3.00. World Bank Reprint Series: Nyumber 218. (including 22 maps, bibliography, Reprinted from Journal of Development Index). Economics. vol. 9 (1981):331-46. LC 80-21352. ISBN 0-19-520186-8, Regional Aspects of Family Stock /Yo. RP-0218. Free of charge. $19.95 (flO.50) hardcover; Planning and Fertility ISBN 0-19-520187-6, $8.95 (£ 4.50) Behavior in Indonesia paperback. Dov Chernichovsky and Oey Astra Meesook Discusses the recent decline in IYutrition and Food Needs Indonesia's population growth rate In Developing Countries despite that country's relatively low Odin K. Knudsen and level of income and socioeconomic Fasquale L. Scandizzo development. Reviews the history ,and organization of the family plan- World Bank Staff Working Paper No. ning program and attempts to iden- 528. May 1979. 73 pages (including tify those factors that have been t appendixes). responsible for its success and stock No. WP-0328. $3.00. assesses its prospects for the future. World Bank Staff Working Paper No. Population and Poverty in 462. M'ay 1981. 62 pages (including theoDevelopioang Woverld iappendix, references). the Developing World Stock No. WP-0462. $3.00. 4Vorld Bank Staff Working Paper No. p04. July 1980. 96 pages (including ' appendixes, bibliography). 5tock No. WP-0404. $3.00. WORLD BANK PUBLICATIONS ORDER FORM SEND TO: WORLD BANK PUBLICATIONS WORLD BANK PUBLICATIONS P.O. BOX 37525 or 66, AVENUE D'IENA WASHINGTON, D.C. 20013 75116 PARIS, FRANCE U.S.A. Name: Address: Stock or ISBN # Author, Title Qty. Price Total Sub-Total Cost: Postage & handling fee for more than two free items ($1.00 each): Total copies: Air mail surcharge ($2.00 each): TOTAL PAYMENT ENCLOSED: _ Make checks payable: WORLD BANK PUBLICATIONS Prepayment on orders from individuals is requested. Purchase orders are accepted from booksellers, library suppliers, libraries, and institutions. All prices include cost of postage by the least expensive means. The prices and publication dates quoted in this Catalog are subject to change without notice. No refunds will be given for items that cannot be filled. Credit will be applied towards future orders. No more than two free publications will be provided without charge. Requests for additional copies will be filled at a charge of US $1.00 per copy to cover handling and postage costs. Airmail delivery will require a prepayment of US $2.00 per copy. Mail-order payment to the World Bank need not be in U.S. dollars, but the amount remitted must be at the rate of exchange on the day the order is placed. The World Bank will also accept Unesco coupons. HG 3881.5 .W57 W67 NO.594 C.3 MERRICK, THOMAS WILLIAM. THE EFFECT OF PIPED WATER ON EARLY CHILDHOOD MORTALITY The World Bank Headquarters: 1818 H Street, N.W. Z Washington, D.C. 20433, U.S.A. Telephone: (202) 477-1234 Telex: WUI 64145 WORLDBANK RCA 248423 WORLDBK Cable address: INTBAFRAD WASHINGTONDC X European Office: 0 66, avenue d'lena 75116 Paris, France Telephone: (1) 723-54.21 Telex: 842-620628 Tokyo Office: 0 Kokusai Building 1-1, Marunouchi 3-chome Chiyoda-ku, Tokyo 100, Japan Telephone: (03) 214-5001 Telex: 781-26838 0 C' :2 S ISSN 0253-2115/ISBN 0-8213-0190-X -