WPS6395 Policy Research Working Paper 6395 Monsoon Babies Rainfall Shocks and Child Nutrition in Nepal Sailesh Tiwari Hanan G. Jacoby Emmanuel Skoufias The World Bank Poverty Reduction and Economic Management Network Poverty Reduction and Equity Unit March 2013 Policy Research Working Paper 6395 Abstract Do household consumption-smoothing strategies in poor increase in weight-for-age for children ages 0–36 countries entail significant long-run costs in terms of months. This total impact consists of a negative “disease reduced human capital? This paper exploits the timing environment effect� of no more than 0.02 standard of monsoon rainfall shocks and the seasonal nature of deviations and a positive “income effect� as high as 0.17 agriculture to isolate income effects on early childhood standard deviations. Consistent with this interpretation, anthropometric outcomes in rural Nepal and to provide excess monsoon rainfall also enhances child stature, but evidence on the persistence of these effects into later only if the monsoon rainfall shock is experienced in the childhood. Findings suggest that a 10 percent increase second year of life. Moreover, this effect on child height is in rainfall from historic norms during the most recently transitory, dissipating completely by age five. completed monsoon leads to a 0.15 standard deviation This paper is a product of the Poverty Reduction and Equity Unit, Poverty Reduction and Economic Management Network. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http:// econ.worldbank.org. The author may be contacted at stiwari@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 Monsoon Babies: Rainfall Shocks and Child Nutrition in Nepal Sailesh Tiwari 1 Hanan G. Jacoby Emmanuel Skoufias The World Bank JEL Classification Codes: I12; J16; O15 Key Words: Income insurance; human capital; malnutrition Poverty Reduction and Economic Management Network Poverty Reduction and Equity 1 Corresponding Author: stiwari@worldbank.org. The World Bank Group (Mail Stop: MC4-415), 1818 H Street NW, Washington DC 20433, USA. Jacoby: hjacoby@worldbank.org; Skoufias: eskoufias@worldbank.org. We are grateful to the Department of Hydrology and Meteorology (DHM) of Nepal for providing the rainfall data used in this study. Research support from World Bank’s Development Economics Research Group Research Support Budget is also gratefully acknowledged. We have benefitted greatly from discussions with Andrew Foster, Ana Aizer, Abla Safir and Mariano Rabassa. All remaining errors are our own. The findings, interpretations and conclusions in this paper are entirely those of the authors and do not necessarily reflect the view of the World Bank, its executive directors, or the countries they represent. 1. Introduction Recognition that households in poor countries, especially those in rural areas, face considerable uninsured income risk has led to increasing policy emphasis on expanding social safety nets as well as encouraging crop or weather insurance. One important rationale for these policies is that lack of insurance, even if it does not manifest itself directly in terms of high consumption volatility, can have harmful indirect effects (Chetty and Looney, 2006). In particular, uninsured households may engage in costly ex-post consumption-smoothing strategies, such as liquidating productive assets or drawing down investments in response to bad shocks. Of chief concern is protecting investments in the human capital of children. Evidence from Jacoby and Skoufias (1997) that parents in rural India withdraw children from school in response to negative income shocks induced by poor rainfall indicates a significant downside to self-insurance, one borne largely by the subsequent generation. 2 Yet, without follow-up information on the eventual schooling (or labor market) outcomes of these children it is difficult to evaluate the long-run cost of this particular consumption-smoothing mechanism. More recently, Maccini and Yang (2009) find that rural Indonesian girls exposed to excess rainfall in their first year of life (though not in their second or subsequent years) grew up to become taller and better educated than their otherwise identical peers. While suggestive, this evidence linking uninsured income volatility and eventual human capital is indirect. This paper looks to strengthen the linkage by considering the impact of rainfall shocks on child anthropometrics (weight and height) in rural Nepal. Our analysis shares features of both of the abovementioned papers. Like Maccini and Yang (2009), we exploit the fact that height is a cumulative measure of past health/nutrition shocks and should therefore reflect income variability experienced earlier in life. In our case, however, we examine height-for-age of children no more than five years of age, rather than of adults. The biomedical literature suggests that stature in adulthood is largely predetermined by age three (e.g., Martorell et al., 1994). A large body of evidence from developing countries 2Jensen (2000) provides similar evidence for child health investments in Africa. Alderman et al. (2006), Foster (1995), and Hoddinott and Kinsey (2001) also examine the child health effects of different types of weather shocks. 2 also shows a causal link between malnutrition, as proxied by height, and success in school (see Glewwe and Miguel, 2008, for a review) and in the labor market (Strauss and Thomas, 1998). In short, child height-for-age is a good indicator of adult human capital. 3 But what sets our approach apart is that we analyze a long-term measure of nutritional status (height-for-age) in combination with a short-term measure (weight-for-age) for the same sample of children. Like school attendance in Jacoby and Skoufias (1997), a child’s weight-for-age should be sensitive to the current economic conditions of the household (see also Foster, 1995). This insight along with a key observation about the timing of rainfall shocks allows us to extract compelling evidence of an income effect from anthropometric data. Given its hilly terrain that makes large-scale irrigation facilities infeasible, Nepali agriculture is highly dependent on rainfall, roughly 80% of which falls in the summer monsoon. All the major staples (paddy, maize, and millet) are cultivated during this season. This stark seasonality of agricultural production means that the realization of a shock to income in our setting takes place with a lag relative to the realization of rainfall; that is, income is only obtained once crops have been harvested and, possibly, sold. The influence of rainfall on disease environment, by contrast, is largely contemporaneous. Excessive rainfall in poor countries is linked to higher incidence of diarrheal disease and, indirectly, to an increase in vector-borne diseases. Episodes of acute illness induced by high rainfall can lower a child’s capacity to take in and retain essential nutrients from food. Thus, weight changes observed around the time of a monsoon rainfall shock but before the associated harvest should mainly reflect the change in disease environment whereas weight changes observed post- harvest should reflect the combined effects of income and the lagged effects of the disease environment that prevailed during the most recent monsoon. 4 3 Among others, Rose (1999), Kudamatsu et al (2011) and Baird et al. (2011) look at the impact of shocks on child mortality in developing countries. Given that vastly more children survive into adulthood than die, however, a greater focus on survivors’ human capital seems warranted. 4 Our analysis also addresses the question of whether households prioritize boys’ welfare over that of girls during times of economic hardship (see Dreze and Sen, 1989, for an early review of this literature in the South Asian context). Among many studies, Behrman (1988) and Behrman and Deolalikar (1990) show biases in favor of boys in nutrient allocation among Indian households during lean times. In India again, Rose (1999) shows that the gender bias in infant mortality, which normally favors boys, narrows when districts experience higher rainfall. In rural Pakistan, Alderman and Gertler (1997) find that the demand for girls’ medical care is more income- and price-elastic than the demand for boys’ medical care. 3 Using three rounds of Demographic and Health Surveys (DHS) conducted in Nepal over the last decade linked to data on precipitation from an expansive network of rainfall stations, we find that a 10% increase in rainfall from historic norms during the most recently completed monsoon season is associated with a 0.15 standard deviation increase in weight-for-age across all children ages 0-36 months. However, a similar rainfall shock during the current monsoon leads to a 0.02 standard deviation reduction in weight-for-age. We can thus infer that the total impact of the past monsoon rainfall consists of a negative “disease environment effect� of no more than 0.02 standard deviations and a positive “income effect� of no less than 0.15 standard deviations, but as high as 0.17 standard deviations. Our evidence also indicates that excess monsoon rainfall experienced in the second year of life (though not in the first year, as found by Maccini and Yang, 2009) has a significantly positive impact on both male and female stature (not just on female stature as in Maccini and Yang). This timing pattern suggests heightened vulnerability immediately following the transition to the family diet. Importantly, however, these nutritional shocks appear to be largely transitory. By the time a child reaches age five, the negative height impact of drought experienced at age two completely dissipates. The rest of the paper is organized as follows. Section 2 lays out the conceptual framework. Section 3 describes the anthropometric and rainfall data. Sections 4 and 5 present the empirical specification and estimation results, respectively. The final section recaps and discusses implications for policy. 2. Conceptual Framework A standard human capital production function specifies a child’s nutritional status at the end of period t (Ht) as a function f of her nutritional status at the end of the previous period (Ht-1), all of the relevant nutritional and health inputs she received during period t (It), and her period t disease exposure (Et). Thus, 𝐻𝑡 = 𝑓(𝐻𝑡−1 , 𝐼𝑡 , 𝐸𝑡 ), (1) 4 where lagged nutritional status captures the entire history of inputs and disease prior to period t as well as the child’s genetic endowments. Parental investment in child nutritional status consist principally in allocating caloric intake and other dimensions of food consumption, as well as in responding to illness induced by the disease environment, for instance by administering oral rehydration therapy during spells of diarrhea. Within this framework, there are two main channels through which excess rainfall can influence child nutritional status. The first is through the alteration of the disease environment net of any parental responses to child illness. In particular, flooding is associated with higher contemporaneous incidence of diarrheal disease, most seriously typhoid and cholera (Confalonieri et al., 2007; WHO, 2002). Standing water also indirectly leads to an increase in vector-borne diseases, such as malaria and dengue, through the expansion in the number and range of vector habitats. Such illnesses lower the capacity to take in and retain essential nutrients from food. Insofar as parents cannot entirely prevent or perfectly ameliorate these effects of child illness, excess rainfall shocks will have a negative impact on nutritional status through the disease channel. The second main channel is through food consumption or real income. In a rural economy based largely on rainfed agriculture, household income is of course highly dependent on rainfall realizations. Moreover, credit constraints limit households’ ability to smooth consumption over time, rendering health more vulnerable to economic shocks (Behrman and Deolalikar, 1988 and 1990, Foster, 1995). Finally, insofar as households are spatially dispersed and transport infrastructure is weak, markets in food staples may not be well integrated. Localized rainfall shocks may, consequently, influence food prices. When seasonal rains are plentiful, yields will be high, food supplies robust and prices low. Such general equilibrium effects reinforce the positive association between rainfall and household purchasing power or real income. Given the seasonal nature of agricultural production and limited borrowing opportunities, the effect from income to consumption, and thereby to child nutrition, is likely to take place with some delay; higher rainfall during the current cropping season can increase consumption only after harvesting. A natural question to ask is whether it is possible to distinguish these disease environment and income channels empirically. Consider a child observed (weighed) during the current monsoon season. Excess rainfall around this time (a current monsoon rainfall 5 shock in our parlance) captures the disease environment effect, which, according to the discussion above, is likely to reduce the child’s weight. At this point in the agricultural calendar, however, the household cannot yet avail itself of the increased purchasing power induced by the higher than normal rainfall; this windfall is realized only after the harvest, which has yet to take place. Hence, the income effect of a contemporaneous rainfall shock on the child’s weight is essentially zero. 5 In contrast, excess rainfall during the last completed monsoon season is likely to operate through both the income and disease environment channels. High past monsoon rainfall has a negative impact on previous period nutritional status through Et-1, which is then carried through to the current period by Ht-1 in equation (1). Moreover, plentiful rainfall during the last monsoon results in a better harvest, increasing It. 6 In sum, the reduced form impact of excess rainfall in the last monsoon on child weight consists of two opposing effects: a (negative) lagged disease environment effect and a (positive) income effect; the reduced form impact is thus a lower bound on the income effect of rainfall shocks on child weight. Indeed, in principle, the reduced form effect of excess rainfall in the last monsoon can be negative. 3. Data Anthropometrics Child and household-level data are drawn from three rounds of the nationally representative Demographic and Health Surveys conducted in Nepal in 2001, 2006 and 2011. The DHS samples were selected using a stratified, two-stage cluster design. The 2001 round consists of 8602 households in 257 clusters of which 84% are rural; 2006 and 2011 rounds are larger with, respectively, 9036 households in 260 clusters (68% rural) and 10,826 households in 289 clusters (71% rural). 5 Excess precipitation could have a contemporaneous effect on household income via changes in the local agricultural labor market. In particular, if the rainfall shock is realized prior to planting and there is significant scope to increase area planted or the use of inputs that require a lot of labor to apply, then wages paid to hired farm workers may rise. We suspect, however, that such effects are of second order importance at best in the context of rural Nepal. 6 To reiterate, it is not necessary to be a food producer in order to reap the benefits of higher yields if food markets are not perfectly integrated. In addition, rural households engaged in non-farm activities may experience an increase in demand, and hence income, when local agriculture is doing well. 6 Given our focus, we restrict attention to children age 0-35 months born and residing in rural areas for our analysis of weight-for-age. For height-for-age on the other hand, we make use of the expanded sample of children 0-59 months as it allows us to assess the persistence of impact of rainfall shocks experienced in the early years over a longer period of time. Height (or length) and weight are standardized based on WHO (2006) to create height-for-age (HAZ) and weight-for-age (WAZ) z-scores. 7 After eliminating biologically implausible z-score values, we are left with 6,941 children pooled across DHS rounds for the analysis of weight-for-age and 11,491 for height-for-age. Tables 1 and 2 present summary z-scores for the sample children broken down by gender, age group and the three survey years. The averages can be compared with the international reference group which has an expected mean z-score of zero. Z-scores below -2 indicate severe growth retardation or “stunting� in the case of HAZ and acute malnutrition or underweight in the case of WAZ. Average WAZ for children in our sample is -1.70 with an underweight prevalence of 49%. Underweight prevalence among boys is slightly higher than for girls and most serious for children older than 0-11 months of age. The latter fact is consistent with the weight-loss children typically experience once they have transitioned from breast milk to the family diet. Comparing across the three survey years, there appears to have been a steady improvement, with underweight prevalence decreasing from 41% in 2001 to 31% in 2011. Summary statistics for height-for-age exhibit similar patterns but are somewhat worse relative to the reference population. Average HAZ is -2.23 with a corresponding stunting prevalence of around 58%. Stunting rates are fairly similar for boys and girls, but starkly higher for children 12-59 months-old compared to infants, which again corresponds to the point of full transition to the family diet. 8 Finally, between 2001 and 2011, the rate of stunting fell from 62% to 48%. 7 Another commonly used anthropometric indicator is weight-for-height, a measure of how “lean� a child is for her height. Weight-for-height, like weight-for-age, is considered a short-run indicator of nutritional status, especially useful when age cannot be determined precisely. Results of our empirical analyses replacing weight-for-age by weight-for-height are very similar (available upon request). 8 The median duration of exclusive breastfeeding in Nepal is 4.2 months, whereas roughly 70% of children between the ages of 6 months and 9 months combine breast milk with complementary food. Global standards of infant and child feeding practices recommend that children who have fully transitioned to the family diet be fed a minimum of four food groups and be fed a minimum number of times per day. Only 25% of the children less than 23 months of age in our 2011 sample met that requirement. 7 Rainfall Monthly precipitation data for a network of 171 rainfall stations and their respective geo-coordinates were obtained from the Department of Hydrology and Meteorology in Nepal. Since each of the DHS sample clusters is geo-coded as well, we are able to match each child to a precise time series of monthly precipitation. The location of rainfall stations and DHS clusters are shown in Figure 1. Geographic coverage of the rainfall stations is fairly expansive and well-balanced throughout Nepal. Though there are fewer stations in the northern, mountainous, parts of the country, these are also areas with sparser population. A key task is to match precipitation records from the weather stations to each of the DHS clusters. The usual practice in the literature is to match each locality – a cluster or the geographic center of an administrative unit – to the geographically closest weather station (Yang and Choi, 2007; Maccini and Yang, 2009; and Menon, 2009). However, this approach may give rise to inaccuracies if there is high prevalence of missing records in the data, stations have been affected by site relocation, or if cluster and station differ considerably in terms of topographic characteristics (e.g., elevation). Instead, therefore, we predict climate or weather based on a station’s observable characteristics (latitude, longitude, elevation, and interactions among these variables) and impute values for each of the DHS clusters based on their corresponding characteristics (Nordhaus et al., 1994, use a similar procedure). We calculate normal rainfall for every month using the average precipitation in that month over the 26 year period 1972-97. Month-to-month precipitation patterns are broadly similar across the four major regions and topographic zones of Nepal as seen in Figure 2. 8 4. Empirical Specification Rainfall shock variables Let 𝑅𝑗𝑡𝑚 be the amount of precipitation in cluster j in month m of year t, and denote �𝑗𝑚 corresponding historical monthly averages. 9 Now, since the monsoon season (June- by 𝑅 September) consists of m = 6, 7, 8, and 9, the rainfall shock in the last completed monsoon, in terms of log deviations, is ∆𝑗𝑡−1 = log(∑9 9 � 𝑘=6 𝑅𝑗𝑡−1𝑚 ) − log( ∑𝑘=6 𝑅𝑗𝑚 ) (2) The current monsoon rainfall shock is slightly more complicated because it depends upon the month 𝑀𝑖𝑗𝑡 in which child i was interviewed and weighed. Thus, define 𝑀 𝑀 𝑖𝑗𝑡 𝑖𝑗𝑡 � 𝑖𝑗𝑡 = �log(∑𝑘=6 𝑅𝑗𝑡𝑚 ) − log( ∑𝑘=6 𝑅 �𝑗𝑚 ) 𝑖𝑓 𝑀𝑖𝑗𝑡 = 6,7,8, 𝑜𝑟 9 ∆ (3) 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 where the additional i subscript reminds us that ∆ � 𝑖𝑗𝑡 can vary across children in the same survey round and cluster. Weight-for-age The basic regression for weight-for-age is then � 𝑖𝑗𝑡 + 𝛿′𝑋𝑖𝑗𝑡 + 𝜑𝑟 + 𝜇𝑡 + 𝜀𝑖𝑗𝑡 , 𝑊𝐴�𝑖𝑗𝑡 = 𝛼 + 𝛽∆𝑗𝑡−1 + 𝛾∆ (4) where 𝑊𝐴�𝑖𝑗𝑡 is the weight-for-age z-score of child i in cluster j in survey round t. Controls consist of normal monsoon rainfall in the cluster (∑9 � 𝑘=6 𝑅𝑗𝑚 ) along with gender, age, year of birth, month of birth, birth order of the child, fixed effects (𝜑𝑟 ) for the 15 regions and fixed effects (𝜇𝑡 ) for the three survey rounds. An augmented set of controls includes characteristics of the mother (age, education level, occupation) and her partner (education level and occupation), and the ethnicity of the household head. 9More precisely, each of these variables is the exponential of the log rainfall prediction using DHS cluster- level explanatory variables and coefficients from the regression of the corresponding log rainfall variable on the same set of explanatory variables (see section 3) at the weather station-level. 9 Identification of 𝛽 and 𝛾 is secured through the considerable cross-sectional and temporal variation in cluster-level rainfall deviations, which should be uncorrelated with any unobserved determinants of child nutritional status (𝜀𝑖𝑗𝑡 ). It should be noted with regard to 𝛾 (impact of current monsoon rainfall), that roughly a quarter of household interviews were carried out during the four monsoon months of June-September; i.e., somewhat less than the expected one-third if interviews were uniformly distributed across the year. The reason for this appears to be that survey work was suspended in the three most mountainous regions during the monsoon. In any case, to account for this we also include a dummy for whether the child was interviewed during the monsoon in all of our specifications. Recall from our discussion in Section 2 that 𝛾 captures the disease environment effect, which we expect to be negative, whereas 𝛽 captures the income effect plus any residual of the disease environment effect from the last monsoon. If there is no weight recovery from past illness, or “catch-up� growth, then the residual disease environment effect should equal 𝛾, the disease environment effect from the current monsoon; if there is complete catch-up, then the residual effect should be zero. In general, the residual effect is 𝜆𝛾, where 𝜆 ∈ [0,1] is a parameter reflecting the persistence of illness shocks on body weight. It follows, then, that the income effect is equal to 𝛽 − 𝜆𝛾 , and since 𝛾 < 0, we see that the income effect of a rainfall shock is bounded from below by 𝛽 and from above by 𝛽 − 𝛾. Height-for-age: A cohort analysis There are two key differences between our analyses of weight-for-age and height- for-age. First, since height-for-age is a cumulative measure nutritional status, and as such unresponsive to current conditions, contemporaneous rainfall shocks are irrelevant. We, therefore, only consider rainfall shocks from prior completed monsoons (i.e., we exclude � 𝑖𝑗𝑡 ). Second, children of different ages will have experienced different numbers of ∆ completed monsoons. For this reason, we divide our sample of children age 0-59 months into five cohorts defined by the number of past monsoons they have lived through. Cohort 1 (the oldest) is comprised of those children who have experienced five completed 10 monsoons since birth, cohort 2 are those with four completed monsoons, and so on. Cohort 5, the youngest, has experienced only one monsoon. Now, let ∆𝑗𝑘� be the excess monsoon rainfall (as defined in equation 2) experienced by children of cohort c in their kth year of life in DHS cluster j (Appendix A lays out the details of cohort construction and the associated rainfall variables). Our regression for the height-for-age z-score of child i in cluster j and survey round t is 𝐻𝐴�𝑖𝑗𝑡 = 𝛼 + ∑3 5−𝑘+1 � 𝑘=1 ∑�=1 𝛽−𝑘 ∆𝑗𝑘� +𝛿 ′ 𝑋𝑖𝑗𝑡 + 𝜑𝑟 + 𝜇𝑡 + 𝜀𝑖𝑗𝑡 , (5) with the parameters 𝛿 , 𝜑𝑟 , and 𝜇𝑡 having the same interpretation as in equation (4). In keeping with our weight-for-age results, we focus on shocks experienced in the first three � years of life. This implies twelve 𝛽−𝑘 parameters to be estimated: five for monsoon rainfall shocks in the first year of life, relevant to cohorts 1 through 5; four for shocks in the second year of life (cohorts 1-4); and three for shocks in the third year of life (cohorts 1-3). Identification of multiple past-monsoon rainfall effects for a given cohort is facilitated by the lack of correlation in cluster-level rainfall deviations across calendar � years. We have no particular priors on the sign or magnitude of the 𝛽−𝑘 , however. Even if we find large disease environment or income effects on weight-for-age, there is still the possibility for growth recovery during these formative early years of life resulting in no � discernible impact on height-for-age. Conversely, a positive (or negative) estimate of 𝛽−𝑘 without a corresponding positive (negative) estimate of 𝛽 in equation (4) would be inconsistent with the framework laid out in Section 2. For instance, it would be surprising if past monsoon rainfall had no effect at all on short-run nutritional status (weight-for-age), yet a substantial effect on longer run status (height-for-age). 5. Results Weight-for-age results Table 3 reports the OLS estimates for weight-for-age, based on the sample discussed in Section 3 and the specification developed in Section 4. The first column includes the current and past rainfall shocks as explanatory variables along with the normal monsoon 11 rainfall in the cluster, a dummy for whether the child was interviewed during the monsoon and region and survey year fixed effects. The second column adds child-level characteristics, and the third column adds characteristics of the mother, her partner, and household. We take this last set of estimates as the basis for distinguishing impacts of excess rainfall through the income and disease environment channels. Returning to equation (4), the implied estimate of 𝛽 is thus 0.15 and that of 𝛾 is - 0.02. The former magnitude, for example, implies that a 10% deviation in the past monsoon rainfall from its historic norm leads to a 0.15 standard deviation change in weight-for-age. 10 Disease environment effects induced by excessive current monsoon rainfall, captured by 𝛾, are relatively small in magnitude, but statistically significant and negative, as the epidemiology suggests. Based on our discussion in Section 4, we infer that the income effect of past monsoon rainfall on weight-for-age lies between 0.15 (if there is complete catch-up from illness induced weight-loss) and 0.17 (zero catch-up growth). Fortunately, this is a fairly narrow window. The last two columns of Table 3 report experiments with controls for proximity to a drinking water source and health posts and the interaction of these variables with current and past monsoon rainfall shocks. 11 Of course, endogenous placement of such facilities may render the estimates of the direct effects of these covariates spurious. Our interest, however, is in the interaction terms; i.e., whether better access to drinking water sources and health facilities mitigates the disease environment effect. Evidently, none of these interactions are significant. Next we probe heterogeneity of impact across gender and age of children. Evidence from south Asia (see fn. 4) suggests that girls are particularly disadvantaged in intra- household reallocations as households respond to economic distress. As for age, while it is generally agreed that the first three years of life are critical for a child’s long term human development, less clear is the particular age “window� of greatest vulnerability to transitory shocks. 10 Reported coefficients have been scaled up by a factor of 100 to conform to this interpretation. 11 Rainfall has direct bearing on the availability of potable water in rural Nepal. Access to piped water is limited and households often rely on community taps, wells and sometimes even river water for drinking purposes. Among several basic health services, one of the most important delivered by health posts in rural Nepal is oral rehydration therapy for diarrheal disease. 12 The results in Table 4 suggest that disease environment effects (the current monsoon rainfall shock) are more salient for boys in the first two years of life. This apparent disadvantage of boys, however, could just as well reflect the interaction of biological and environmental factors than any differences in behavioral responses to child illness on the part of parents. At any rate, based on the coefficients on last completed monsoon rainfall deviations, we do not find much evidence of “discrimination� against girls; the point estimates across columns 2 and 3 are statistically indistinguishable. Indeed, once we net out the larger disease environment effects for boys – using our arguments from Section 4 – the pure income effects are extremely close for boys and girls. Robustness As support for our argument that excess rainfall in the current or ongoing monsoon is associated with the disease environment, we examine the determinants of diarrhea episodes in our sample children. DHS surveys record the incidence of diarrhea in the two weeks preceding the day of the interview; 20% of children in our sample reported a bout of diarrhea. Results in Table 5 indicate a significant positive relationship between excess rainfall and contemporaneous diarrhea among children. Insofar as diarrheal morbidity inhibits nutrient uptake and thereby decreases body weight in the short-run, we take this evidence as corroboration that the current monsoon rainfall shock operates through a disease environment channel. Turning next to the rainfall shock in the last completed monsoon, we perform two robustness checks. First, if our interpretation is correct, then we should see that the implied income effect is much smaller, if not nonexistent, for children residing in urban areas (see also Maccini and Yang, 2009). After all, urban households’ purchasing power is likely to be much less sensitive to the vagaries of the harvest than that of rural households. Indeed, the results reported in Table 6 suggest as much. Using the coefficients in the last column, we can bound the income effect of monsoon rainfall shocks for urban children at no greater than -0.014, which is actually negative but statistically indistinct from zero. This finding bolsters our argument that much of the positive rainfall effect on the body weight of rural children is mediated through fluctuations in agricultural income. Second, we perform a “placebo� test of whether a future completed monsoon rainfall shock – that is, excess 13 rainfall in the monsoon that begins only after the date of interview – has any effect on weight-for-age. Unless correlations between rainfall shocks and child nutritional status are entirely spurious, future shocks should be irrelevant and the results reported in Table 7 show that indeed they are. Height-for-age results Results for height-for-age, our longer-term measure of nutritional status, appear in Table 8. To reiterate, cohorts are defined according to how many past monsoons were experienced. The impact of excess monsoon rainfall in the first year of life is statistically indistinguishable from zero for all five cohorts. Recall that the identical shock had a significant effect on body weight for children 0-11 months. Arguably, for cohort 5, the youngest, not enough time has elapsed following the shock (and associated harvest) to observe an impact on body length. For the older cohorts, however, the finding that shocks affecting body weight in the first of year of life have no lasting impact on stature suggest that, at these very young ages, catch-up growth is possible. Further support for the catch- up hypothesis emerges in the effects of monsoon rainfall shocks experienced in the second year of life: For children in cohorts 3 and 4, excess rainfall in their second monsoon is associated with a significant increase in stature, while the effects of the identical shocks for cohorts 1 and 2 (i.e., children age 4-5) are statistically indistinguishable from zero. 12 These findings stand in stark contrast to the evidence from Indonesian adults presented by Maccini and Yang (2009). First, although the effect-size implied by our point estimate of 0.13 standard deviations (for cohort 4) for a 10% excess rainfall is roughly twice as large as that of Maccini and Yang (based on the standard deviation of adult heights in their sample), the impact disappears by age five. Second, whereas in Maccini and Yang’s study, the shock experienced in the first year of life was most salient for the stature of adult women, our results highlight shocks in the second year of a child’s life (for both girls and boys). The average child in the sample was 17 months old at the close of their second monsoon season. Thus, we place the critical age window closer to the end of the child’s transition to 12 Here, as with weight-for-age, there is no significant difference between boys and girls in the cohort-specific effects of rainfall shocks. Joint F-tests for shock × cohort × female are insignificant for all cohorts taken together or for each individually (results of unrestricted specification available upon request). 14 the family diet, a time when caloric intake is likely to be more sensitive to fluctuations in household purchasing power relative to the breastfeeding period; it is a pattern consistent as well with the timing of growth-faltering that we observe on average in the sample. To home in on the dissipation effect, we now drop cohort 5 (the youngest) from the estimation sample and eliminate all shock × cohort interactions from the regression. Instead, we count the number of months elapsed for each child since their second monsoon ended and interact this variable with excess rainfall in that monsoon. The average number of months elapsed since the event in question is 43, 31, 19, and 7, for cohorts 1 through 4 respectively. In the first column of Table 9, we see that the (positive) impact of 10% higher monsoon rainfall in the second year of life on child stature falls by 0.003 standard deviations for every month removed from that monsoon (the time pattern is not significantly different for girls; col. 2). Thus, according to this linear-in-months specification, by 37 months out, the shock impact goes to zero. A quadratic specification (col. 3) leads to essentially the same conclusion. In other words, three years after the fact, a deficit in the first post-weaning monsoon leaves no detectable trace on child nutritional status. 6. Conclusions We have made considerable progress in understanding the long-run costs of uninsured income risk in poor countries. Exploiting the seasonal nature of agriculture in Nepal and the fact that body weight (as opposed to length) is sensitive to current economic conditions, our empirical analysis has uncovered compelling evidence of income effects on child nutritional status. Although we do find that excessive rainfall reduces the weight of young children by altering the disease environment, the income channel working in the opposite direction (i.e., wet years leading to good harvests) dominates in terms of magnitudes. Our evidence also indicates that rainfall shocks experienced early in life contribute to child growth-faltering, but that the damage is transitory, with no discernible scars by age five. Similar findings from sub-Saharan Africa (Rabassa et al., 2012) suggest 15 that catch-up growth may be sufficient to restore a child to his or her development trajectory after such temporary perturbations. 13 Child malnutrition is a complex development challenge, to be sure, calling for coordinated policy response on several fronts, ranging from improving the intake of essential nutrients, expanding basic health care facilities, water quality, and sanitation. Our evidence, however, does not support an emphasis on income insurance as a means to protect this form of human capital accumulation. While providing access to insurance may be valuable for a host of other reasons, enhancing long-run nutritional status is not likely to be one of them. 13Whether there might be long-term impacts of early childhood income shocks on such forms of human capital as cognitive skills is an interesting topic for future research. 16 References Alderman, H. and P. Gertler (1997). “Family Resources and Gender Differences in Human Capital Investments: The Demand for Children’s Medical Care in Pakistan�. In Intrahousehold Resource Allocation in Developing Countries: Models, Methods and Policy, ed. L. Haddad, J. Hoddinott and H. Alderman, 231-48. Baltimore, MD: Johns Hopkins University Press for the International Food Policy Research Institute. Alderman, H., J. Hoddinott and B. Kinsey (2006), “Long term consequences of early childhood malnutrition�, Oxford Economic Papers, 58: 450-474 Baird, S., J. Friedman and N. Schady (2011), “Aggregate Income Shocks and Infant Mortality in the Developing World�, Review of Economics and Statistics, 99(3), 847-856 Behrman, J. (1988). “Intrahousehold Allocations of Nutrients in Rural India: Are Boys Favored? Do Parents Exhibit Inequality Aversion?� Oxford Economic Papers, 40(1): 32-54. Behrman, J., and A. Deolalikar (1988). “Health and Nutrition.� Handbook of Development Economics, edited by H. Chenery and T. N. Srinivasan. Amsterdam: Elsevier. Behrman, J. and A. Deolalikar (1990). “The Intrahousehold Demand for Nutrients in Rural South India: Individual Estimates, Fixed Effects, and Permanent Income.� Journal of Human Resources, 25(4): 665-96. Confalonieri, U., B. Menne, R. Akhtar, K.L. Ebi, M. Hauengue, R.S. Kovats, B. Revich and A. Woodward (2007). Human health. Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, M.L. Parry, O.F. Canziani, J.P. Palutikof, P.J. van der Linden and C.E. Hanson, Eds., Cambridge University Press, Cambridge, UK, 391-431. Currie, J. (2009). “Healthy, Wealthy and Wise: Socioeconomic Status, Poor Health in Childhood and Human Capital Development�. Journal of Economic Literature. 47(1): 87-122. Dreze, J. and A. Sen (1989). “Hunger and Public Action�. WIDER Studies in Development Economics. Oxford: Oxford University Press. Glewwe, P. W. and Miguel, E. A. (2008). “The Impact of Child Health and Nutrition on Education in Less Developed Countries�. In Schultz, T. P. and Strauss, J. A., editors, The Handbook of Development Economics, volume 4, chapter 56, pages 3561–3606. Elsevier. Hoddinott, J. and B. Kinsey (2001). “Child Growth in the Time of Drought.� Oxford Bulletin of Economics and Statistics, 63(4): 409-436. Jacoby, H. and E. Skoufias (1997). "Risk, Financial Markets, and Human Capital in a Developing Country," Review of Economic Studies, 64(3): 311-35. Jensen R. (2000). “Agricultural Volatility and Investment in Children.� American Economic Review, 90(2): 399-404. Kudamatsu, M., T. Persson, and D. Strömberg (2011). “Weather and Infant Mortality in Africa.� Unpublished manuscript, Stockholm University, Sweden. Maccini, S. and D. Yang (2009). “Under the Weather: Health, Schooling, and Economic Consequences of Early-life Rainfall.� American Economic Review, 99(3): 1006-1026. 17 Martorell, R., L. Kettel Khan and D. Schroeder (1994). “Reversibility of Stunting: Epidemiological Findings in Children from Developing Countries." European Journal of Clinical Nutrition, 48: 45- 57. Menon, N. (2009), “Rainfall Uncertainty and Occupational Choice in Agricultural Households of Rural Nepal�, Journal of Development Studies, 45(6), 864-888. Nordhaus, W., R. Mendelsohn and D. Shaw (1994). “The Impact of Global Warming on Agriculture: A Ricardian Approach.� American Economic Review, 84(4): 753-771. Rabassa, M., E. Skoufias, and H. Jacoby (2012). “Weather and Child Health in Rural Nigeria,� Policy Research Working Paper no. 6214, World Bank. Washington, DC. Rose, E. (1999). “Consumption Smoothing and Excess Female Mortality in Rural India.� The Review of Economics and Statistics, 81(1): 41-49. Strauss, J. and D. Thomas (1998). “Health, Nutrition, and Economic Development.� Journal of Economic Literature, 36(2): 766-817. World Health Organization (2002). “Health and Environment in Sustainable Development." WHO, Geneva. World Health Organization Multicentre Growth Reference Study Group (2006). WHO Child Growth Standards: Length/height-for-age, weight-for-age, weight-for-length, weight-for-height and body mass index-for-age: Methods and development. Geneva: World Health Organization; pp 312. (available on the web site: http://www.who.int/childgrowth/publications/en/). Yang, D. and H. Choi (2007). “Are Remittances Insurance? Evidence from Rainfall Shocks in the Philippines.� World Bank Economic Review, 21(2): 219-248. 18 Table 1: Descriptive statistics for weight-for-age z-scores Gender Age in months All children Boys Girls 0-11 12-23 Mean -1.77 -1.79 -1.75 -1.44 -1.95 St. Dev. 1.17 1.19 1.15 1.25 1.13 2001 Underweight (%) 0.41 0.43 0.39 0.31 0.47 Obs 3399 1655 1744 1143 1166 Mean -1.73 -1.74 -1.71 -1.33 -1.91 St. Dev. 1.17 1.19 1.15 1.31 1.07 2006 Underweight (%) 0.39 0.39 0.39 0.28 0.45 Obs 2421 1215 1206 750 811 Mean -1.45 -1.55 -1.34 -1.22 -1.49 St. Dev. 1.19 1.21 1.15 1.31 1.11 2011 Underweight (%) 0.31 0.35 0.26 0.25 0.31 Obs 1121 563 558 371 359 Mean -1.70 -1.73 -1.67 -1.37 -1.87 St. Dev. 1.18 1.19 1.16 1.28 1.12 Total Underweight (%) 0.49 0.49 0.48 0.29 0.44 Obs 6941 3433 3508 2264 2336 Note: Based on children aged 0-35. Underweight refers to children whose weight-for-age z-scores were below 2 standard deviation of the international reference group of children the same age. 19 Table 2: Descriptive statistics for height-for-age z-scores Gender Age in months All children Boys Girls 0-11 12-23 24-35 36-47 48-59 Mean -2.35 -2.36 -2.34 -1.35 -2.24 -2.58 -2.59 -2.43 St. Dev. 1.29 1.28 1.29 1.31 1.29 1.23 1.24 1.16 2001 Stunted 0.62 0.62 0.62 0.29 0.57 0.69 0.71 0.66 Obs 4966 2459 2507 473 1159 1083 1135 1116 Mean -2.19 -2.17 -2.22 -1.18 -2.08 -2.32 -2.38 -2.29 St. Dev. 1.24 1.23 1.26 1.23 1.29 1.23 1.15 1.14 2006 Stunted 0.58 0.57 0.58 0.25 0.53 0.62 0.65 0.61 Obs 3477 1748 1729 233 811 858 778 797 Mean -1.94 -1.99 -1.88 -1.02 -1.67 -2.18 -2.23 -2.05 St. Dev. 1.34 1.36 1.32 1.52 1.38 1.28 1.18 1.20 2011 Stunted 0.48 0.50 0.46 0.19 0.41 0.56 0.57 0.51 Obs 1666 859 807 168 355 389 400 354 Mean -2.23 -2.23 -2.23 -1.24 -2.10 -2.41 -2.45 -2.32 St. Dev. 1.29 1.28 1.30 1.34 1.32 1.25 1.21 1.17 Total Stunted 0.58 0.58 0.58 0.26 0.53 0.64 0.67 0.62 Obs 10109 5066 5043 874 2325 2330 2313 2267 Note: Based on children aged 0-59. Stunted refers to children whose weight-for-height z-scores were below 2 standard deviation of the international reference group of children the same age. 20 Table 3: Impact of rainfall shocks on weight-for-age z-scores (1) (2) (3) (4) (5) VARIABLES Weight-for-age z-scores Shock in current period monsoon rain (RC) -0.001 -0.003* -0.002** -0.002** -0.003*** (0.001) (0.001) (0.001) (0.001) (0.001) Shock in the most recent completed monsoon rain (RP) 0.023*** 0.022*** 0.015*** 0.015*** 0.016*** (0.006) (0.007) (0.005) (0.005) (0.005) Normal monsoon rainfall (mm) 0.000 -0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) Interviewed during the monsoon ( = 1) -0.170*** 0.110 0.159** 0.158** 0.154** (0.048) (0.074) (0.054) (0.053) (0.054) Drinking water source far? (DWS) -0.077** -0.077* (0.035) (0.039) Health post difficult to access? (HPA) -0.029 -0.020 (0.029) (0.041) DWS x RC 0.001 (0.002) DWS x RP -0.003 (0.003) HPA x RC 0.001 (0.002) HPA x RP 0.001 (0.003) Year of survey effects Yes Yes Yes Yes Yes Region fixed effects Yes Yes Yes Yes Yes Child characteristics No Yes Yes Yes Yes Mother's characteristics No No Yes Yes Yes Observations 6,853 6,853 6,853 6,853 6,853 R-squared 0.013 0.106 0.250 0.251 0.251 Notes: Robust standard errors clustered by region in parentheses (*** p<0.01, ** p<0.05, * p<0.10). Current period monsoon shock and the most recent completed monsoon shock variables represent the log deviation of the current period monsoon and the most recent completed monsoon from the historic norm for each DHS cluster. DWS is an indicator for whether the nearest drinking water source is at least a 5 minute walk from the dwelling; HPA is an indicator for whether the household reports getting to the nearest health post is difficult due to distance. Normal monsoon rainfall is the average monsoon rainfall for the cluster for the period 1972-1996. Region fixed effects are a combination of three topographic zones (mountains, hills and terai) and five regions (eastern, central, western, mid-western and far-western). Child characteristics include age in months, birth order, month of birth, year of birth, year of birth interacted with month of birth and gender. Mother's characteristics are education level, occupation, partner's education, and his occupation, as well as ethnicity of household head. 21 Table 4: Impact of rainfall shocks on weight-for-age by gender and age group (1) (2) (3) VARIABLES Weight-for-age z-scores All children Boys Girls 0-11 months x Shock in the current period monsoon -0.006** -0.010** -0.002 (0.002) (0.004) (0.002) 12-23 months x Shock in the current period monsoon -0.001 -0.004* 0.001 (0.002) (0.002) (0.003) 24-35 x Shock in the most recent completed monsoon -0.000 -0.000 -0.001 (0.002) (0.003) (0.003) 0-11 months x Shock in the most recent completed monsoon 0.019*** 0.017** 0.021** (0.006) (0.007) (0.008) 12-23 months x Shock in the most recent completed monsoon 0.012** 0.010* 0.016* (0.005) (0.005) (0.009) 24-35 x Past rainfall shock 0.012** 0.010* 0.017** (0.004) (0.005) (0.007) Female 0.074*** (0.022) Normal monsoon rainfall (mm) 0.000 -0.000 0.000 (0.000) (0.001) (0.001) Interviewed during the monsoon ( = 1) 0.160*** 0.156* 0.162*** (0.051) (0.080) (0.050) Year of survey effects Yes Yes Yes Region fixed effects Yes Yes Yes Child characteristics Yes Yes Yes Mother's characteristics Yes Yes Yes Observations 6,853 3,383 3,470 R-squared 0.264 0.338 0.346 Notes: Robust standard errors clustered by region in parentheses (*** p<0.01, ** p<0.05, * p<0.10). Current period monsoon shock and the most recent completed monsoon shock variables represent the log deviation of the current period monsoon and the most recent completed monsoon from the historic norm for each DHS cluster. DWS is an indicator for whether the nearest drinking water source is at least a 5 minute walk from the dwelling; HPA is an indicator for whether the household reports getting to the nearest health post is difficult due to distance. For other variables, see notes to Table 3. 22 Table 5: Impact of rainfall shocks on incidence of diarrhea (1) (2) (3) (4) VARIABLES Dependent variable: Diarrhea ( =1 if reported in the last two weeks) Shock in the current period monsoon (RC) 0.0014** 0.0016** 0.0015* 0.0015* (0.0006) (0.0006) (0.0007) (0.0007) Female x RC 0.0003 0.0004 (0.0006) (0.0006) Female -0.0436** -0.0434** -0.0342 -0.0330 (0.0147) (0.0178) (0.0291) (0.0255) Drinking water source far? (DWS) -0.0012 -0.0064 (0.0268) (0.0301) Health post difficult to access? (HPA) 0.0108 0.0331 (0.0157) (0.0365) DWS x RC -0.0005 (0.0009) HPA x RC 0.0008 (0.0010) Normal monsoon rainfall (mm) 0.0001 0.0001 0.0001 0.0001 (0.0004) (0.0004) (0.0004) (0.0004) Year of survey effects Yes Yes Yes Yes Region fixed effects Yes Yes Yes Yes Child characteristics Yes Yes Yes Yes Mother's characteristics No Yes Yes Yes Observations 1,908 1,908 1,908 1,908 R-squared 0.086 0.132 0.132 0.155 Notes: Robust standard errors clustered by region in parentheses (*** p<0.01, ** p<0.05, * p<0.10). See notes to Table 3. 23 Table 6: Impact of rainfall shocks on weight-for-age z-scores for urban children (1) (2) (3) VARIABLES Weight-for-age z-scores Shock in current period monsoon rain (RC) 0.000 -0.004 0.000 (0.004) (0.003) (0.004) Shock in the most recent completed monsoon rain (RP) -0.003 -0.005 -0.014 (0.009) (0.006) (0.012) Normal monsoon rainfall (mm) -0.001 -0.002*** -0.001 (0.001) (0.001) (0.001) Year of survey effects Yes Yes Yes Region fixed effects Yes Yes Yes Child characteristics No Yes Yes Mother's characteristics No No Yes Observations 1,257 1,257 1,257 R-squared 0.006 0.141 0.446 Notes: Robust standard errors clustered by region in parentheses (*** p<0.01, ** p<0.05, * p<0.10). See notes to Table 3. 24 Table 7: ‘Placebo test’ for future rainfall shocks on weight-for-height z-scores (1) (2) (3) VARIABLES Weight-for-age z-scores Shock in current period monsoon rain (RC) -0.001 -0.003* -0.002*** (0.001) (0.001) (0.001) Shock in the "future" monsoon rain (t+1) (RF) -0.001 0.001 0.000 (0.002) (0.002) (0.002) Normal monsoon rainfall (mm) -0.002 -0.002* -0.001 (0.001) (0.001) (0.001) Year of survey effects Yes Yes Yes Region fixed effects Yes Yes Yes Child characteristics Yes Yes No Mother's characteristics No Yes No Observations 5,820 5,820 5,820 R-squared 0.006 0.106 0.261 Notes: Robust standard errors clustered by region in parentheses (*** p<0.01, ** p<0.05, * p<0.10). See notes to Table 3. 25 Table 8: Impact of past rainfall shocks on height-for-age (1) (2) (3) VARIABLES Height-for-age Rainfall shock in first year of life (%)[R1] x Cohort 1 -0.014 -0.014 -0.010 (0.010) (0.009) (0.009) R1 x Cohort 2 0.005 0.002 0.002 (0.005) (0.005) (0.004) R1 x Cohort 3 0.000 -0.001 -0.001 (0.006) (0.006) (0.005) R1 x Cohort 4 -0.007*** 0.002 -0.002 (0.002) (0.003) (0.003) R1 x Cohort 5 (youngest) 0.001 -0.001 -0.002 (0.004) (0.009) (0.009) Rainfall shock in the second year of life (%)[R2] x Cohort 1 0.004 0.006 0.004 (0.004) (0.005) (0.005) R2 x Cohort 2 -0.001 -0.005 -0.006 (0.007) (0.008) (0.008) R2 x Cohort 3 -0.004 0.009** 0.008* (0.002) (0.004) (0.004) R2 x Cohort 4 0.012** 0.012** 0.013** (0.005) (0.005) (0.004) Rainfall shock in the third year of life (%)[R3] x Cohort 1 0.009 0.005 0.004 (0.011) (0.012) (0.012) R3 x Cohort 2 -0.005 0.003 0.000 (0.003) (0.004) (0.005) R3 x Cohort 3 -0.002 -0.003 -0.004 (0.003) (0.006) (0.005) Year of survey effects Yes Yes Yes Region fixed effects Yes Yes Yes Child characteristics No Yes Yes Mother's characteristics No No Yes Observations 10,102 10,102 10,102 R-squared 0.064 0.125 0.146 Notes: Robust standard errors clustered by region in parentheses (*** p<0.01, ** p<0.05, * p<0.10). R1 is the monsoon rainfall shock experienced by the child in the first year of life, R2 is the monsoon shock experienced in the second year of life and R3 the monsoon shock experienced in the third year of life. Monsoon shocks for each child are computed as the log deviation from the normal rainfall in the cluster in which he/she resides. Cohort 1 corresponds to the oldest children in the sample while Cohort 5 is comprised of the youngest ones. See notes to Table 3. 26 Table 9: Impact of rainfall shock experienced in the second year of life (1) (2) (3) VARIABLES Height-for-age Rainfall shock in the second year of life (%) [R2] 0.011*** 0.011*** 0.016*** (0.00216) (0.00254) (0.00400) R2 x No. of months since the shock -0.00030*** -0.00021*** -0.00069 (0.00008) (0.00007) (0.00041) R2 x No. of months since the shock x Female -0.00010 (0.00018) R2 x Female 0.00114 (0.00371) R2 x (No. of months since the shock)2 0.0000082 (0.0000078) Year of survey effects Yes Yes Yes Region fixed effects Yes Yes Yes Child characteristics Yes Yes Yes Mother's characteristics Yes Yes Yes Observations 7,767 7,767 7,767 R-squared 0.084 0.082 0.084 Notes: Robust standard errors clustered by region in parentheses (*** p<0.01, ** p<0.05, * p<0.10. See notes to Table 8. 27 Figure 1: Location of DHS clusters and rainfall stations 28 Figure 2: Normal rainfall in mountains, hills and the terai, across four regions of Nepal Eastern Region Central Region 500 500 Normal Rainfall (mm) Normal Rainfall (mm) 400 400 300 300 200 200 100 100 0 0 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 Month Month Western Region Mid-Western Region Normal Rainfall (mm) 100 200 300 400 500 500 Normal Rainfall (mm) 400 300 200 0 1 2 3 4 5 6 7 8 9 10 11 12 100 Month 0 1 2 3 4 5 6 7 8 9 10 11 12 Mountains Hills Month Terai 29 APPENDIX A Construction of cohorts and rainfall variables for height-for-age analysis DHS 2001 Survey period in 2001 Jan Feb March April May June July Aug Sept Oct Nov Dec Monsoon Cohort Year of Month of Rainfall shock experienced in the…. Birth Birth ..1st year ..2nd year ..3rd year ..4th ..5th year year 5 2000 Jan. – Rain Sep. 2000 5 1999 Sept. – Rain Dec. 2000 4 1999 Jan. – Rain Rain 2000 Sep. 1999 4 1998 Sept. – Rain Rain 2000 Dec. 1999 3 1998 Jan. – Rain Rain 1999 Rain Sep. 1998 2000 3 1997 Sept. – Rain Rain 1999 Rain Dec. 1998 2000 2 1997 Jan. – Rain Rain 1998 Rain Rain Sep. 1997 1999 2000 2 1996 Sept. – Rain Rain 1998 Rain Rain Dec. 1997 1999 2000 1 1996 Jan. – Rain Rain 1997 Rain Rain Rain Sep. 1996 1998 1999 2000 30 DHS 2006 Survey period in 2006 Jan Feb March April May June July Aug Sept Oct Nov Dec Monsoon Cohort Year of Month of Rainfall shock experienced in the…. Birth Birth ..1st year ..2nd year ..3rd year ..4th ..5th year year 5 2005 Jan. – Rain Sep. 2005 5 2004 Sept. – Rain Dec. 2005 4 2004 Jan. – Rain Rain 2005 Sep. 2004 4 2003 Sept. – Rain Rain 2005 Dec. 2004 3 2003 Jan. – Rain Rain 2004 Rain Sep. 2003 2005 3 2002 Sept. – Rain Rain 2004 Rain Dec. 2003 2005 2 2002 Jan. – Rain Rain 2003 Rain Rain Sep. 2002 2004 2005 2 2001 Sept. – Rain Rain 2003 Rain Rain Dec. 2002 2004 2005 1 2001 Jan. – Rain Rain 2002 Rain Rain Rain Sep. 2001 2003 2004 2005 31 DHS 2011 Survey period in 2011 Jan Feb March April May June July Aug Sept Oct Nov Dec Monsoon Cohort Year of Month of Rainfall shock experienced in the…. Birth Birth ..1st year ..2nd year ..3rd year ..4th ..5th year year 5 2010 Jan. – Rain Sep. 2010 5 2009 Sept. – Rain Dec. 2010 4 2009 Jan. – Rain Rain 2010 Sep. 2009 4 2008 Sept. – Rain Rain 2010 Dec. 2009 3 2008 Jan. – Rain Rain 2009 Rain Sep. 2008 2010 3 2007 Sept. – Rain Rain 2009 Rain Dec. 2008 2010 2 2007 Jan. – Rain Rain 2008 Rain Rain Sep. 2007 2009 2010 2 2006 Sept. – Rain Rain 2008 Rain Rain Dec. 2007 2009 2010 1 2006 Jan. – Rain Rain 2007 Rain Rain Rain Sep. 2006 2008 2009 2010 32