WPS5023 Policy Research Working Paper 5023 How Many More Infants Are Likely to Die in Africa as a Result of the Global Financial Crisis? Jed Friedman Norbert Schady The World Bank Development Research Group Poverty and Inequality Team & Human Development and Public Services Team August 2009 Policy Research Working Paper 5023 Abstract The human consequences of the current global financial then apply this estimated likelihood to expected growth crisis for the developing world are presumed to be shortfalls as a result of the crisis. At current growth severe yet few studies have quantified such impact. The projections, their estimates suggest there will be 30,000­ authors estimate the additional number of infant deaths 50,000 excess infant deaths in sub-Saharan Africa. Most in sub-Saharan Africa likely due to the crisis and discuss of these additional deaths are likely to be poorer children possible mitigation strategies. They pool birth-level data (born to women in rural areas and lower education levels) as reported in female adult retrospective birth histories and are overwhelmingly female. If the crisis continues from all Demographic and Health Surveys collected in to worsen the number of deaths may grow much larger, sub-Saharan Africa nations. This results in a data set of especially those to girls. Policies that protect the income 639,000 births to 264,000 women in 30 countries. The of poor households and that maintain critical health authors use regression models with flexible controls for services during times of economic contraction should be temporal trends to assess an infant's likelihood of death considered. Interventions targeted at female infants and as a function of fluctuations in national income. They young girls may be particularly beneficial. This paper--a product of the Poverty and Inequality Team and Human Development and Public Services Team, Development Research Group--is part of a larger effort in the department to explore the economic and human consequences of economic crises. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at jfriedman@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team How Many More Infants Are Likely to Die in Africa as a Result of the Global Financial Crisis? Jed Friedman and Norbert Schady Development Research Group, The World Bank 1. Introduction The US subprime crisis, and the likelihood that the attendant downturn will spread from developed to developing countries has left policy-makers and academics concerned about the "human" costs of the crisis for poor countries. These human costs could in principle take a variety of forms--lower income and consumption levels, higher unemployment, lower investments in education by families, and worse health and nutrition outcomes. Of particular concern are changes that could have irreversible consequences for the wellbeing of children throughout their lives--for example, school dropout or impaired child development at critical, early ages. No outcome is more dramatic than infant death, and in no region is infant mortality higher than in Sub-Saharan Africa (henceforth Africa, for short).1 In this note we assess the likely impact of the current financial crisis on infant mortality in Africa. To do this, we first use data from all available Demographic and Health Surveys (DHS) for the region to construct retrospective birth and death histories, and assess the impact of past crises on infant mortality in Africa. We then use the most recent (April 2009) IMF growth projections to estimate the likely impact of the current downturn on infant mortality. The basic messages of our analysis are two. First, although severe, the consequences of the current economic crisis for infant mortality in Africa are likely to be much smaller than many of the estimates that have been frequently cited on blogs or in the popular press.2 Specifically, our estimates suggest approximately 30,000 to 50,000 excess infant deaths in Africa in 2009. Second, however, we find that virtually all of these excess deaths take place to girls, not boys. Furthermore, these asymmetries in the impact of the crisis by gender are particularly apparent for sharp (rather than modest) contractions in per capita GDP--a finding that is important if the current growth projections were to be revised further downward. 2. Data and methods The DHS ask women a set of questions about the date of birth, current vital statistics, and date of death (if deceased) of all children ever born. We use the responses to these questions to construct retrospective birth and death histories for every woman in the surveys. Our measure of infant mortality is an indicator that takes on the value of one if a child died at a reported age of 12 1 In 2005, the IMR per 1,000 children born was 90.8 in Sub Saharan Africa. As a comparison, in this same year the rate was 24.2 in Latin America and the Caribbean, 62.2 in South Asia (57.7 in India), and 24.8 in East Asia (21.4 in China). 2 See http://africacan.worldbank.org/a-sub-prime-crisis-in-the-us-and-infant-deaths-in-africa; Harvey Morris, "Forgotten victims of the downturn", Financial Times March 11, 2009, p. 4; Nicholas Kristof, "At Stake Are More Than Banks", New York Times, April 1, 2009. 2 months or younger.3 After removing trends from GDP in various ways, we compare the probability of infant death in years in which GDP was below trend with other years. This approach closely follows Baird Friedman and Schady (2007) (henceforth BFS; see also Paxson and Schady 2005 on a crisis in Peru and Bhalotra 2009 on cyclicality in infant mortality in India).4 Table 1 gives the list of countries (30) and DHS (62) in Africa. These surveys collect information on approximately 260,000 women and 640,000 live births, and are the basis for our calculations. Yearly GDP per capita numbers are taken from the IMF; for every country, Table 2 gives the average growth rate between 1993 and 2008, the growth rate in 2008, and the April 2009 IMF projection of the growth rate for 2009; the last column, finally, gives the "growth shortfall"--the difference between the 2008 growth rate and the 2009 projected growth rate.5 This difference is a (simple) measure of how much lower growth rates in 2009 are expected to be than they would have been in the absence of the financial crisis, and are important for our calculations of "excess deaths". We discuss this in more detail below. The basic approach we take involves pooling data from all of the surveys and running regressions of the following form: (1) D imct c f c (t ) logGDPct imct 3 We use this measure of infant mortality, rather than the standard definition of mortality for children younger than 12 months, because of age heaping in reports of mortality. Further, we discard information for children born within 12 months of the survey when calculating mortality rates to avoid complications with censored data. Results are largely unchanged if we restrict our mortality measure to those infants who died at an age younger than 12 months. 4 Although the DHS are a rich source of data, they also have some limitations for our analysis. We briefly discuss two of these limitations, both of which are related to the use of retrospective information in the DHS to construct birth and death histories. First, recall bias may be a concern if women are less likely to accurately remember more distant births and deaths. To minimize recall errors, we do not use information on births that occurred more than 11 years prior to the date of the survey. However, our results are robust to different cut-offs for the recall period. Second, any given survey is representative of women ages 15-49 at the time of the survey, but is not representative of all births and child deaths in earlier years. To see this, note that a woman aged 49 in a survey carried out in 2000 would have been 39 in 1990. If no surveys were carried out between 1990 and 2000 in this country, no data would be available on births to women aged 40 or older in 1990. Children born to older women may respond to economic fluctuations differently than those born to younger women. To avoid this problem, we discard from the sample births to women age 40 or older. Our analysis therefore provides meaningful estimates of the relationship between income fluctuations and infant mortality for women aged 15 to 39. (We note, however, that a very small fraction, less than 3 percent, of births in our sample of DHS countries occur to women age 40 or older.) This retrospective construction of births and infant deaths to women aged 15-39 results in series of varying lengths and with varying start periods depending on the number and dates of DHS surveys in each country. 5 For four countries in the table ­ Djibouti, Eritrea, Lineria, and Zimbabwe ­ growth information was missing for either 2008 or 2009 (or both). In these cases, we replace the missing value with the Africa-wide average. These four countries represent 2.9% of the SSA population and 2.5% of births and so any mis- measurement as a result of these imputations is highly unlikely to bias the aggregate results for the continent. 3 where Dimct is an indicator that takes on the value of one if child i born to mother m in country c and year t died in her first year of life, zero otherwise; c is a set of country fixed effects; fc(t) are country-specific time trends--in practice, we consider linear, quadratic, and cubic trends. These trends are important to ensure that we do not confound the effect of downturns in GDP with long- term trends in mortality (which may vary by country); logGDPct is the natural logarithm of per capita GDP; and imct is the regression error term. Standard errors are clustered by country which corrects for any autocorrelation in shocks to infant mortality across years within country. It is also possible to restrict the sample to women who have had more than one child--in the African DHS this is approximately 85 percent of women--and include a set of mother fixed effects. This compares the probability of death for two children, born to the same woman, in different economic times. This specification has the advantage that it controls for all time- invariant characteristics of mothers. The characteristics of mothers giving birth have been shown to vary with economic circumstances in some settings, and it is important to parse out differences in mortality to comparable children and families from compositional changes.6 The impact of a downturn on mortality may be more or less severe for different kinds of children. For example, more educated (and wealthier) mothers may be better able to smooth consumption of critical inputs into child health. In practice, because we construct retrospective birth and death histories for children born in the past to a given woman, we can only analyze heterogeneity in outcomes for mother or household characteristics that are arguably time- invariant. We focus on two: urban-rural location, and education. This approach implicitly assumes that place of residence at the time of the survey is correlated with place of residence at the time of reported child birth, and that schooling has been completed by age 15; these should be reasonable approximations for most of the countries and years in our sample. (When we restrict our sample to women age 18 or older we get very similar results.) We also consider the possibility of differential impacts for boys and girls. Finally, it is possible that households respond differently to positive and negative shocks to GDP. It is also likely that households are better able to protect investments in child health and nutrition during relatively modest negative shocks than when these shocks are deep or sustained--as would be the case, for example, if credit constraints are important. In terms of estimation, this implies that the coefficient in (1) above will vary with the sign and magnitude 6 For example, Dehejia and Lleras-Muney (2004) find that, among blacks in the United States, the fraction of births to high-risk mothers goes up during recessions; while Bhalotra (2009) estimates that higher-risk women are more likely to defer fertility in India. On the other hand, BFS and Paxson and Schady (2005) do not find important changes in the composition of women giving birth in crisis and non-crisis years. 4 of the shock. There are different ways of exploring heterogeneity of this kind. In practice, we run a series of spline regressions. The spline regressions we run are motivated by the fact that the one-step specification in (1) is very closely related to a two-step approach. This involves first calculating Dct , the average mortality rate in country c in year t, then regressing this mean on country fixed effects and a flexible, country-specific formulation of time, as before: (2) D ct c f c (t ) ct The residual from this regression gives de-trended measures of the infant mortality rate, D* . ct Similarly, a regression of logGDPct on country fixed effects and a flexible, country-specific formulation of time gives de-trended measures of logGDPct. In the second step, D * is regressed ct on detrended logGDPct, with each country-year cell being weighted by the number of live births. Dct logGDPct ct * * (3) In practice (and as expected) the coefficient in (3) is very close in value to the coefficient in (1). However, the two-step process lends itself naturally to a spline specification, as follows: Dct 1logGDPct 2 [logGDPct * I(K 1)] ct * * * (4) Where I(K=1) is an indicator value that takes on the value of one if the de-trended value of * logGDPct is below a given value (the "knot" in the spline). We consider spline specifications in which the knot is at zero, -1, and -1.5 standard deviations of logGDPct, respectively. (For our sample of African countries, a one standard deviation corresponds to a 3.4 percentage point change from trend GDP.) In all of these specifications, the impact of a negative shock of a given magnitude is given by the sum of the 1 and 2 coefficients. For example, when the knot is fixed at zero, 1 gives the change in infant * mortality for a one-unit, positive deviation in logGDPct from expected trends, while the sum of * 1 and 2 gives the change in infant mortality for a one-unit, negative deviation in logGDPct . 3. Results Main results: The basic set of results for the full sample is reported in Table 3. The first row shows that a 1 percent reduction in per capita GDP is associated with an increase in infant mortality of between 0.34 per thousand (in the specification that accounts for linear trends) and 0.62 per thousand (in the specification that accounts for cubic trends). 5 Should we think of these as "large" or "small" effects? We consider two ways of putting the magnitude in context. First, the average infant mortality rate for countries and years in our sample is 106.7 per 1,000 children born; this implies that a 1 percent change in per capita GDP is associated with a 0.32 to 0.58 percent change in mortality, implying an elasticity of mortality with respect to GDP of between -0.32 and -0.58. Second, the average annual decline in infant mortality for countries and years in our data is 2.56 per thousand. A 1 percent decline in GDP from expected trends is therefore associated with an increase in infant mortality that is equivalent to between 0.13 (0.34/2.56) and 0.24 (0.62/2.56) of the average annual decline in infant mortality in the sample. More meaningfully, perhaps, in countries in which per capita GDP growth was below trend, the average deviation from trend was -2.6 percentage points. The average event of below- trend growth in our data was therefore associated with an increase in mortality of between 0.34 and 0.62 percent of the average annual decline in IMR. Changes in the composition of women giving birth: The second and third rows of Table 3 focus on possible changes in the composition of women giving birth. We first re-estimate our basic model, but limit the sample to women who have had at least two live births. The coefficients on log per capita GDP for this somewhat smaller sample are very similar to those for the full sample of women--a little bit larger in the linear specification, and a little bit smaller in the quadratic and cubic specifications. In the third row of the table we then present the results from regressions that include mother fixed effects. These results are reasonably similar to those in the second row, which suggests that changes in the composition of women giving birth explain at most a small fraction of the increases in infant mortality we observe during negative shocks to GDP in Africa. Heterogeneity by gender, place of residence, education: In Table 4, we consider the possibility that shocks to per capita GDP have different effects on the mortality of boys and girls (upper panel), on the mortality of children born to women living in urban and in rural areas (middle panel), and on the mortality of children born to women with more or less education (lower panel). The table suggests that, in Africa, the mortality of girls is substantially more sensitive to income shocks than is the mortality of boys: A 1 percent deviation in GDP results in approximately 0.33 more boy deaths per thousand, but 0.62 more girl deaths per thousand; note, moreover, that the differences in percentage terms are even larger because of the higher average mortality of boys than girls--a well-known finding in the demographic literature.7 The middle and lower panels of the table suggest that the mortality of children in rural areas, and of children 7 For example, the World Health Organization (2006) estimates that the male-to-female ratio in neonatal mortality and in early neonatal mortality in developing countries is 1.3. 6 born to mothers with lower levels of education is also more sensitive to deviations from GDP trend (although base infant mortality rates for women in rural areas and those with less education tend to be higher; as a result, the proportional increases in mortality tends to be more closely similar to each other). Heterogeneity by size of shock: Table 5 presents the results from a variety of spline specifications, as described above, separately for boys (left-hand side of the table) and girls (right-hand side). In each case, the right-hand panel, corresponding to positive shocks (or negative shocks of smaller magnitude) reports the coefficient 1, while the left-hand panel, corresponding to negative shocks of increasingly large magnitude, reports the sum of the 1 and 2 coefficients. The table makes clear that deep, negative shocks to GDP have particularly dire consequences for the infant mortality of girls. Thus, in the specification that fixes the knot at zero, a 1 percent negative deviation in GDP results in an increase in infant mortality of 0.25 per thousand boys born, but 0.90 per thousand girls born; for negative deviations of 1 standard deviations or larger, the comparable coefficients imply increases in mortality of 0.14 per thousand for boys, and 1.73 per thousand for girls; finally, for the deepest shocks of 1.5 standard deviations or larger, every percentage point reduction in GDP results in an increase in mortality of 0.046 per thousand boys born, and 2.14 per thousand girls. All of the coefficients for girls are highly significant, but none of those for boys. Thus, a negative shock that would lower the growth rate of per capita GDP by 10 percentage points from its underlying (cubic) trend would result in approximately 22 more deaths per thousand girls born, but have no impact on boy mortality. Put differently, for the largest negative shocks to GDP, the elasticity of infant mortality with respect to GDP is -2.1 for girls, but not significantly different from zero for boys.8 The very stark differences in the impact of aggregate shocks on the mortality of boys and girls are surprising, as Africa is not a region with clear patterns of "son preference" (unlike South Asia, for example). One possible explanation is that households protect sons over daughters when their incomes fall (even if they do not necessarily do so at other times). An alternative explanation is that there is selection in births. In comparison to females, male fetuses have a relatively difficult time surviving pregnancy (see Perls and Fretts, 1998; Mizuno, 2000) and this difference may be exacerbated during hardship periods. Almond et al. (2007) argue that, during the economic chaos associated with the Great Leap Forward in China, the ratio of male to female 8 This finding of differential responses in male and female infant mortality is not limited to the cubic functional form. The same patterns are found with a quadratic polynomial in time, or if the relationship between infant mortality and GDP is estimated with first differenced data. 7 births fell because of an increase in spontaneous abortions of male fetuses. Given this, the health endowment of girls born would on average then be lower than that of boys, and this could result in higher mortality of girls after birth. To investigate this possibility, we calculated the ratio of male to female births when deviations in per capita GDP was negative, and compared it to cases where it was positive; when per capita GDP was below -1 standard deviations, and compared it to cases where it was above this value; and when per capita GDP was below -1.5 standard deviations, and compared it to cases where it was above this value. (In each case, a cubic trend is first removed from GDP, as before.) In each of these cases, the ratios of female to male births were very similar, and we can never reject the null that they are the same. For example, the ratio of male to female births was 1.031 when deviations from GDP were negative, and 1.019 when positive. Similarly, the ratio of male to female births was 1.022 when deviations from trend GDP were less than -1.5 standard deviations, and 1.026 when above. In neither comparison is the difference in the ratios statistically significant at conventional levels. By contrast, and consistent with the results we present above, the ratio of male to female deaths in the first year is always higher when detrended GDP is negative, especially when the negative shocks are large. We conclude that the data are not consistent with a selection story along the lines described above. Rather, they are likely to reflect differences in behavior towards boys and girls by parents or health providers at times of economic distress. Calculation of "excess deaths": In Table 6, we present some basic calculations of the number of excess infant deaths that are likely to take place in Africa in 2009. To do this, we first calculate the number of children born in each country (using the total population and crude birth rates for 2008 taken from the World Bank DDP database). We then calculate excess deaths corresponding to three scenarios. In the first scenario, we compare the April 2009 IMF projected growth rates for 2009 with actual growth rates in 2008--see Table 2. The difference between these two values then gives the decline in per capita GDP that can plausibly be attributed to the global financial crisis. This difference, multiplied by the coefficient in Table 3 (or Table 5, for the spline specification), and scaled by the number of children born in a given country, gives a country-specific estimate for excess deaths. (In the table, we report the sum for all of Africa.) These calculations, summarized in the top row in the table, suggest that, using the most recent IMF projections, we could expect there to be somewhere between 28,000 and 49,000 excess deaths in Africa in 2009.9,10 9 All of these calculations make a number of important assumptions. First, that negative shocks to GDP do not have an effect on aggregate fertility rates. This is because the number of deaths depends not only on the 8 Suppose, however, that the IMF growth projections were too conservative, so that the "true" contractions in GDP were larger. In the middle and bottom row we assume that every country in Africa were to have a growth rate that was 6 percent (middle row) or 10 percent (bottom row) below trend. Under these much more dire scenarios, excess deaths would be approximately 61,000 to 166,000 for the 6 percent reduction, and 101,000 to 322,000 for the 10 percent reduction--a very large number by almost any measure, but still substantially below many of the numbers discussed in the popular press. We note, however, that Table 5 suggests that essentially all of these excess deaths would take place to girls, rather than boys. In Africa, there appears to be a very important gender dimension to the likely impact of negative shocks to aggregate GDP. 4. Discussion and conclusion Aggregate economic shocks can have serious, deleterious consequences for the wellbeing of children. In practice, these consequences depend on the nature and depth of the shock; on the extent to which households have access to credit or other mechanisms that allow them to smooth out the effects of an income shortfall; on public and other investments in education and health; and on the outcome being considered. There may also be important differences in the impact of a systemic shock by place of residence, socioeconomic status, and gender. In practice, aggregate economic shocks, whether from macroeconomic crises, collapses in the price of key exports, or weather shocks, appear to have had negative implications for child health and education in Africa (Ferreira and Schady 2009). mortality rate, but also on the number of children born. However, in regressions of the crude birth rate on country-specific intercepts, country-specific time trends and a variable for logGDPct, the coefficient on GDP is small and not significant at conventional levels. Second, we assume that the coefficients estimated above can be "applied" to the full list of African countries--including those that have never fielded a DHS, and were therefore not part of our sample. This assumption seems reasonable given the large number of countries that have carried out DHS in Africa--see Table 1. Third, we assume that a given change in per capita GDP will on average have the same effect on infant mortality in 2009 as it has in the past. There is obviously no way to test this assumption. Conceivably, the fact that the crisis is global (unlike the trend deviations in per capita GDP that form the basis for the calculations in the tables in the paper) could mean that the impacts of contractions in GDP will be larger--for example, if developed countries are less able or willing to provide aid--or smaller--if widespread awareness of the potential consequences of such a crisis on mortality prompts African countries or donors to take measures that counteract its effects. 10 An alternative approach to estimating the growth impacts of the financial crisis compares the World Bank's latest growth projections for 2009 (as of mid-April, 2009) with earlier projections (from 2007) for the same calendar year. This approach, as adopted in Chen and Ravallion (2009), yields a smaller growth shortfall for Sub-Saharan Africa than the IMF projections used here ­ a cross-country average contraction of 1.7% of GDP in comparison with a 2.5% contraction ­ and consequently somewhat fewer excess deaths, ranging from 16,000 to 23,000 depending on the specification. Because the IMF growth projections are more current (at the time of writing) we report the excess deaths based on these figures. 9 In this note, we use household level data on mortality, national accounts data on past shocks, and the most recent IMF projections of future growth to estimate the impact of the current global slowdown on infant mortality. Our estimates suggest that there will be on the order of 30,000 to 50,000 excess deaths in Africa in 2009--deaths that would not have taken place had the subprime crisis which began in the United States not spread to African countries. This is a large number. Nevertheless, it is substantially smaller than alternative estimates, generally based on aggregate (rather than household) data, and with cruder if any controls for household characteristics. The increase in infant mortality that can be attributed to the global slowdown is also a small fraction of overall infant deaths in Africa: In Africa, approximately 30 million children are born, and more than 3 million of them die before they reach their first year of life. The bigger tragedy is perhaps not that 30,000 to 50,000 additional children are likely to die in Africa in 2009, but rather that so many children die in any given year. Nevertheless, the likely increase in infant mortality presents some important policy challenges, as does the fact that the bulk of the additional children who will die is likely to be found among poorer households (in rural areas, and those with lower education levels) and is concentrated among girls. This note cannot discuss the benefits and limitations of individual policies in great detail. However, we briefly mention a handful of policy responses that are perhaps important for policy-makers to consider: Policies that limit the extent to which the aggregate GDP shortfalls translate into income shortfalls for households: Such policies could include a well-designed public works scheme, perhaps with guaranteed employment at low wages (the latter, to ensure self targeting), or cash transfers to households in poor areas, areas that appear to have been particularly hard-hit by the contraction, or to households with characteristics that are correlated with low incomes (for example, low education levels). Policies that limit the extent to which the contraction in aggregate GDP leads to a decline in the quantity or the quality of critical health services--such as prenatal and post-natal check-ups for women, attended birth deliveries, and growth monitoring for children. Paxson and Schady (2005) show that, in Peru, public health expenditures fell by more than 50 percent during the crisis of the late 1980s, and that the utilization of critical health services by households also declined. Ferreira and Schady (2009) show that, by contrast, in Indonesia there was a much smaller contraction in public expenditures in 1997-98, and that this contraction was made up by increases donor assistance; they argue 10 that this may explain, in part, why infant mortality appears to have spiked much more sharply in Peru than in Indonesia. Interventions targeted at girls: The fact that, in crises past, increases in infant mortality have been almost exclusively concentrated among girls presents a serious challenge to policy-makers in the region. We do not have a good understanding of the mechanisms whereby aggregate economic contractions translate into more deaths for girls, but not boys in Africa--a critical element to designing effective policies (although selection into birth does not appear to be an important part of the story). Nonetheless, policies that ensure that girls receive sufficient food and critical services are worth considering; conditional cash transfers that focus on conditions for girls may also be effective. More broadly, it is important to track inputs and outcomes at the household and provider level, and to disaggregate the findings by gender. 11 References Almond, Douglas, Lena Edlund, Hongbin Li, and Junsen Zhang. 2007. "Long-Term Effects of the 1959-61 China Famine: Mainland China and Hong-Kong." NBER Working Paper 13384, Boston, MA. Baird, Sarah, Jed Friedman, and Norbert Schady. 2007. "Aggregate Income Shocks and Infant Mortality in the Developing World." World Bank Policy Research Working Paper 4346. Bhalotra, Sonia. 2009. "Fatal Fluctuations? Cyclicality in Infant Mortality in India." Forthcoming, Journal of Development Economics. Chen, Shaohua and Martin Ravallion. 2009. "The Impact of the Global Financial Crisis on the World's Poorest." VoxEU.org Available online at http://www.voxeu.org/index.php?q=node/3520 Dehejia, Rajeev, and Adriana Lleras-Muney. 2004. "Booms, Busts, and Babies' Health." Quarterly Journal of Economics 119(3): 1091-1130. Ferreira, Francisco, and Norbert Schady. 2009. "Aggregate Economic Shocks, Child Schooling and Child Health." Forthcoming, World Bank Research Observer. Mizuno, Reiko. 2000. "The Male/Female Ration of Fetal Deaths and Births in Japan". Lancet, 356(9231): 738-739. Paxson, Christina, and Norbert Schady. 2005. "Child Health and Economic Crisis in Peru." World Bank Economic Review 19(2): 203-223. Perls, Thomas T. and Ruth C. Fretts. 1998. "Why Women Live Longer than Men". Scientific American: Women's Health: 100-103. World Health Organization. 2006. "Neo-natal and Peri-natal Mortality: Country, Regional, and Global Estimates". WHO Press, Geneva. World Bank. 2007. World Development Indicators. Available online at http://ddp.worldbank.org/ddp/home.do. 12 Table 1. List of DHS datasets used in the analysis, including information on country, year of survey, number of mothers, and number of births Total Country Survey Years Total Mothers Births Benin 1996, 2001 7515 18891 Burkina Faso 1993, 1999, 2003 16362 39410 Burundi 1987 2416 6464 Central African Republic 1995 3373 7962 Cameroon 1991, 1998, 2004 11444 27350 Chad 1997 4655 11829 Comoros 1996 1405 3838 Cote d'Ivoire 1994, 1999 6660 15993 Ethiopia 2000 8436 20484 Gabon 2001 3371 7084 Ghana 1988, 1994, 1999, 2003 11841 25675 Guinea 1999 4549 11224 Kenya 1989, 1993, 1998, 2003 18457 44289 Liberia 1986 3419 8669 Madagascar 1992, 1997 7592 19195 Malawi 1992, 2000 11368 27292 Mali 1987, 1996, 2001 17915 47710 Mozambique 1997 5535 12468 Namibia 1992, 2000 6674 13550 Niger 1992, 1998 9468 26714 Nigeria 1990, 1999, 2003 14333 36543 Rwanda 1992, 2000 9317 23607 Senegal 1986, 1993, 1997 11881 30636 South Africa 1998 6017 9970 Sudan 1990 4242 11314 Tanzania 1992, 1996, 1999 12826 29743 Togo 1988, 1998 7611 18582 Uganda 1989, 1995, 2001 11883 30062 Zambia 1992, 1997, 2002 13776 32044 Zimbabwe 1989, 1994, 1999 9346 19913 Total: 30 Countries 62 Surveys 263687 638505 13 Table 2. Past growth and 2009 projected growth in Sub-Saharan African nations Average Growth IMF 2009 Country 1993-2008 2008 Growth projections Growth shortfall SubSaharan Africa 3.9 4.9 2.5 -2.4 Angola 8.1 14.3 -3.6 -17.9 Benin 4.5 4.9 3.8 -1.1 Botswana 5.6 3.0 -10.4 -13.4 Burkina Faso 4.8 4.1 3.5 -0.6 Burundi 0.4 4.4 3.5 -0.9 Cameroon 3.1 3.9 2.4 -1.5 Cape Verde 6.1 6.7 2.5 -4.2 Central African Republic 2.0 3.4 2.4 -1.0 Chad 5.5 1.6 2.8 1.2 Comoros 1.5 0.6 0.8 0.2 Congo, Dem. Rep. 0.3 8.7 2.7 -6.0 Congo, Rep. 2.8 9.1 9.5 0.4 Cote d Ivoire 1.8 2.7 3.7 1.0 Djibouti* 3.9 4.9 2.5 -2.4 Equatorial Guinea 30.0 8.1 -5.4 -13.5 Eritrea 4.2 1.2 1.1 -0.1 Ethiopia 7.0 10.8 6.5 -4.3 Gabon 2.1 3.0 0.7 -2.3 Gambia, The 4.1 5.3 4.0 -1.3 Ghana 4.9 6.4 4.5 -1.9 Guinea 3.6 3.0 2.6 -0.4 Guinea-Bissau 1.3 2.9 1.9 -1.0 Kenya 3.2 2.4 3.0 0.6 Lesotho 3.6 4.1 0.6 -3.5 Liberia* 3.9 4.9 4.9 0.0 Madagascar 3.4 6.8 -0.2 -7.0 Malawi 3.9 7.0 6.9 -0.1 Mali 4.8 5.0 3.9 -1.1 Mauritania 3.9 2.1 2.3 0.2 Mauritius 4.7 5.0 2.1 -2.9 Mozambique 7.5 6.5 4.3 -2.2 Namibia 3.9 2.7 -0.7 -3.4 Niger 3.6 4.9 3.0 -1.9 Nigeria 4.4 6.1 2.9 -3.2 Rwanda 5.0 8.4 5.6 -2.8 Senegal 3.9 4.5 3.1 -1.4 Sierra Leone 4.0 5.8 4.5 -1.3 South Africa 3.4 3.1 -0.3 -3.4 Sudan 6.4 6.1 4.0 -2.1 Swaziland 2.3 2.2 0.5 -1.7 Tanzania 5.1 7.2 5.0 -2.2 Togo 3.0 0.8 1.7 0.9 Uganda 6.6 6.9 6.2 -0.7 Zambia 3.2 5.5 4.0 -1.5 Zimbabwe -1.7 -4.9 2.5 7.4 Note: where missing (Djibouti, Liberia, 3rd column Eritrea, 4th column Zimbabwe) missing value replaced with SSA average 14 Table 3. Estimates of relation between IMR and GDP for Sub-Saharan Africa, by various trend accountings Dependent Linear Quadratic Cubic variable Unadjusted IMR -33.73 -46.07 -47.67 [16.70]* [15.95]*** [22.05]** Unadjusted, restricted to mothers with multiple births IMR -38.68 -41.91 -46.38 [17.74]** [15.17]*** [25.77]* Mothers' fixed effects IMR -42.93 -39.54 -34.76 [18.79]** [18.07]** [25.87] Note: Number of observed births equals 613,468 in panel, 532,419 in bottom two panels. Robust standard errors are clustered at the country level - there are30 countries. GDP is measured in year 2000 international (PPP) dollars.* p<.10, ** p<.05, *** p<.01 Table 4. Estimates of relation between IMR and GDP by child or mother characteristics, quadratic trend Dependent Characteristics variable Boy Girl IMR -32.91 -61.94 [25.03] [25.72]** Mean IMR 113.68 100.35 N 310265 303203 Urban Rural IMR -29.70 -53.45 [25.77] [23.32]** Mean IMR 83.24 116.36 N 171649 441819 Primary or greater Less than primary IMR -28.67 -51.24 [21.44] [26.95]* Mean IMR 71.58 119.93 N 162881 450587 Note: Robust standard errors are clustered at the country level - there are30 countries. GDP is measured in year 2000 international (PPP) dollars. * p<.10, ** p<.05, *** p<.01 15 Table 5. Heterogeneity in IMR and GDP relation by size of GDP deviation from cubic trends, by gender Boys Girls <= 0 >0 <= 0 >0 Knot at zero -24.47 -39.80 -90.25 -24.80 [26.39] [40.75] [29.32]*** [36.96] <= -1 sd > -1 sd <= -1 sd > -1 sd Knot at -1 s.d. -14.21 -36.47 -172.68 -28.30 [29.11] [29.74]* [46.64]*** [28.32] <= -1.5 sd > -1.5 sd <= -1.5 sd > -1.5 sd Knot at -1.5 s.d. -4.16 -35.46 -214.12 -38.14 [26.87] [27.73] [56.93]*** [25.69] Note: Slope coefficients are estimated from a continuous spline specification. Robust standard errors are clustered at the country level. GDP is measured in year 2000 international (PPP) dollars.* p<.10, ** p<.05, *** p<.01 Table 6. Summary estimates of excess infant deaths as a result of the economic crisis, by various growth forecasts and impact specifications Cubic, Cubic, restricted mother Spline, Independent variable Linear Quadratic Cubic to FE fixed cubic sample effects Difference between 2009 IMF growth 28472 38888 40229 39140 29334 49370 projections and 2008 growth Posited 6% decline in projected growth 60719 82933 85913 83491 62573 165901 Posited 10% decline in projected growth 101198 138221 143022 143022 139151 321746 16