Drives Child Undernutritionin the Global Tropics

The El Ni˜no Southern Oscillation (ENSO) is a principal component of global climate variability known to inﬂuence a host of social and economic outcomes, but its systematic eﬀects on human health remain poorly understood. We estimate ENSO’s association with child nutrition at global scale by combining variation in ENSO intensity from 1986-2018 with children’s height and weight from 186 surveys conducted in 51 teleconnected countries, containing 48% of the world’s under-5 population. Warmer, drier El Ni˜no conditions predict worse child undernutrition in most of the developing world, but better ones in the small number of areas where precipitation is positively aﬀected by ENSO. This relationship looks similar at both global and regional scale, and has not appreciably weakened over the last four decades. Results imply that the 2015 El Ni˜no pushed over 7 million children into undernutrition, demonstrating the degree to which human well-being remains subject to predictable climatic processes.


Introduction
Climate variability is increasingly recognized as a key determinant of health outcomes (1 ) and a major concern for global climate policy and international public health (2 ), with the IPCC warning that anthropogenic climate change will very likely increase the frequency and intensity of extreme events (3 , 4 ). The El Niño Southern Oscillation (ENSO) is a major source of climate variability known to affect key social, economic and health outcomes (5 -14 ), however, the systematic effects that these correlated shifts in the tropical climate have on global health remain understudied. ENSO's adverse large-scale effects have been documented for hundreds of years (15 ), yet despite our improved predictive capacity over ENSO, much of humanity is still susceptible to its consequences. Given that probabilistic forecasts of ENSO have skill at predicting conditions months in advance, there is an opportunity to decouple food insecurity and human nutrition from this predictable climate process. However, analyses of ENSO's impacts on food security have generally focused on a single country or El Niño episode (16 ) and lack global or regional scope to guide national and international public investments that preempt adverse effects of ENSO.
This paper provides the first estimate of ENSO's impacts on human nutrition in the global tropics, leveraging over one million geolocated child records spanning four decades and all developing country regions. We estimate the systematic effect of ENSO-driven tropical climate variability by examining the association between annual eastern equatorial Pacific ENSO state and measures of children's weight from surveys conducted during that year. Children's anthropometric measures are extremely sensitive to nutritional shocks due to their high caloric needs while growing (17 ), and provide a summary measure of contemporary household food security (18 ). Our interest is in estimating the total effect of ENSO variability across all channels of influence -from agricultural productivity to infectious disease to conflict -that might affect human nutrition, and documenting systematic differences in ENSO response within sample.
We capture ENSO variation using the widely-used Niño 3.4 index of equatorial Pacific sea surface temperature (19 , 20 ), which spans 5 • N-5 • S, 170 • W-120 • W (Fig. 1A). Children's weight-for-age z-scores (WAZ) at time of survey (Fig. 1B) are calculated using the NCHS/CDC/WHO International Reference Standard (21 ) intended to provide a single measure of child nutritional outcomes comparable across ages and sexes.
We first identify all countries with local climates teleconnected to ENSO (Fig. 1C) for which Demographic and Health Surveys (DHS) anthropometric data exist. This yields a sample of 1.3 million children ages 0-4 interviewed in 186 household surveys between 1986 and 2018. The sample includes 51 countries containing 38% of the world's population and 48% of the world's under-5 population as of 2018. We assign treatment (i.e., the ENSO state when the child was surveyed) annually by tropical year, accounting for the spring barrier delay in ENSO state change (8 ), by calculating the maximum Niño 3.4 value between May-Dec of a given year. We assign that to all children interviewed by DHS during that period, as well as all children interviewed during the following year's Jan-Apr months (i.e., before the following year's spring barrier).
While a warmer ENSO leads to higher temperature throughout the tropics, shifts in precipitation pattern lead to some areas getting wetter than normal while others get drier. We account for potential differences in the effects of ENSO by allowing for different effects in subnational regions where precipitation is positively correlated to warmer ENSO (Figs. 1D and S1). Since only 6.4% of our sample lives in regions where warmer ENSO leads to clear wet anomalies, we focus our discussion on results for the majority of the sample. In order to remove potential confounders (22 , 23 ), we purge the estimates of average differences across countries and across rural and urban areas within each country using location fixed effects, and we detrend the data annually and remove monthly seasonality by major world regions. The temporal variation in ENSO anomalies -measured as a deviation from long-run average conditions -is used to statistically isolate the association of ENSO state and child malnutrition.

Results
We find that warmer, more El Niño-like ENSO conditions increase short-term undernutrition in children across the tropics, with the opposite occurring in places where precipitation tends to increase during El Niño. Even in the absence of controls, the empirical distribution of detrended WAZ is significantly and substantially different (p < 0.001) between El Niño and La Niña years ( Fig. 2A). After detrending the data and controlling for location-specific unobservable confounders and mother characteristics (Table 1), a 1 • C increase in the ENSO index is associated with 0.03σ (p = 0.02) average decrease in WAZ. We allow the relationship between ENSO and WAZ to vary flexibly (Fig. 2B) and find that the negative association remains across the distribution of ENSO values. The result is substantively similar across a broad range of model specifications (Table S2), and across other outcomes reflecting recent nutrition, including weightfor-height and body mass index (-0.04σ/ • C and p<0.01 for both measures). Height-for-age or stunting do not decrease with warmer ENSO conditions, consistent with height being slower to respond to health shocks than weight (24 ). Using WHO z-score classification thresholds, warmer ENSO increases the prevalence of undernourishment (below -2σ in weight for age) significantly by 0.6 percentage points per 1 • C (p < 0.05).
We find that the risk of wasting (below -2σ in weight for height) is positive but not significant (0.3pp/ • C, p = 0.21), consistent with higher noise in height measurements, especially for very sick children (25 ). ENSO conditions do not have a contemporaneous effect on stunting (below -2σ in height for age). All of these patterns are reversed in the minority of places in our sample (6.4%) where warmer ENSO is correlated to wet anomalies. The heterogeneity in results across regions of wet and dry anomalies point towards the importance 3 Notes: Different anthropometric effects of ENSO are concentrated on short run measures (1-3) weight-for-age, weight-for-height, and body mass index z-scores, which all measure shorter-run effects of scarce nutrition, show evidence of contemporaneous ENSO effects measured in • C, while (4) height-for-age, a slower-varying measure, shows only weak evidence in the minority (∼6%) of the sample where NINO3.4 is positively correlated to precipitation. Estimates are from OLS regressions with controls consisting of: fixed effects (indicators) for each country; country-specific mother's age at child's birth, total years of mother's education, and rural vs. urban indicator; and UNICEF world region-specific linear trends in survey year and fixed effects for month of interview. Standard errors are two-way clustered at the level of tropical year and subnational administrative unit, and observations are reweighted using DHS sample weights and country size weights in order for estimates to be representative for an average country. (5-7) WHO threshold outcomes show ENSO increases the likelihood of being undernourished (below -2σ in weight for age), but shows a weaker, statistically insignificant effect on wasting (below -2σ in weight for height) and no relationship with stunting (below -2σ in height for age). Asterisks indicate statistical significance at the 1% (***), 5% (**) and 10% (*) levels.
of agriculture in mediating the ENSO-nutrition link, though others (e.g., conflict) cannot be ruled out.
The several degree variation in ENSO cycle implies that it is a meaningful source of variation in population nutrition in the tropics. According to these estimates, the 2.25 • C increase in the Niño 3.4 index during the 2015 El Niño event, one of the largest on record, likely caused average WAZ in the representative child of our sample countries to decrease by 0.1σ. The human scale of this impact is large given that the under-5 population in our sample countries was 311 million in 2015. To give context to the size of these effects, we look at the scale of public health interventions needed to offset an event of similar magnitude to the 2015 El Niño, using published effect sizes of nutritional interventions (26 ).  (Tables S1-S3). Positive deviations from the ENSO index mean state decrease z-scores, and negative deviations of the ENSO index mean (La Niña events) reduce undernutrition compared to baseline state (Table S3). For the purposes of population-wide attribution statistics, we also calculate the average effect of warmer ENSO in the average country, without separating the sample by whether warm ENSO leads to dry or wet anomalies (Table S4). The average effect across the sample suggests that warmer ENSO leads to a 0.04σ/ • C reduction in weight-for-age (p = 0.02), and a 1 percentage point increase in prevalence of undernourishment (p < 0.01). We also test the use of alternative detrending of the data with decade fixed effects ( Table S5). The lagged effects of ENSO (Table S6) indicates no persistent effect of ENSO on child nutrition, except in the subsample with positively correlated rainfall. Finally, we show that timing of DHS survey does not vary as a result of ENSO state ( Figure S2 and Table S10) and that results hold under a placebo randomization test ( Figure S3).

Discussion
The negative relationship between child nutrition and warm ENSO state does not appear to vary appreciably across space and time, with the effects for major world regions and different decades in the sample being statistically indistinguishable from our main effect (Fig. 2C). That there has been little progress in attenuating the food security effects of ENSO despite increasing incomes and trade connectivity implies that the limits to adaptation may be strict, at least over the income range of countries in the sample. The international community has set the target of eliminating all forms of malnutrition worldwide by 2030 as part of the Sustainable Development Goals (SDG) agenda, and is making efforts to establish metrics, monitor, and implement policies to achieve this goal (27 ). Developing countries deemed to be making insufficient progress 7 are being pressured to do more (28 ). During 2015-2018, 34% of the children in our sample countries were undernourished, implying that in order to meet the hunger SDG the percentage of undernourished children  (29 ). We assemble our treatment variable by taking the maximum monthly anomaly value of the index during the May-Dec period of a given calendar year, assigning it to all children interviewed during that tropical year (May-April). We note that alternative constructions of ENSO3.4 state perform similarly (see Table S3). DHS Children's Anthropometric Data are from all DHS surveys containing children's anthropometric data (N=186, 51 countries, 1986-2018). We standardize DHS administrative region names, aggregating to supersets if any regions changed borders or split during our sample period. We calculate each child's anthropometrics using his/her height and weight following the NCHS/CDC/WHO International Reference Standard (21 ) intended to provide a single measure of child nutritional outcomes comparable across ages and genders.

Methods
Time series data on El Niño are from the NOAA Climate Prediction Center's monthly time series. We designate a country as teleconnected if its temperature is closely coupled to ENSO9, defined as having at least 50% of the population living in locations where local temperature as reported in the UDEL global gridded temperature dataset (30 ) at month t is significantly correlated with the second month lag (t-2) of the ENSO state (NINO3.4 SST index) for at least three months of the year. We designate teleconnection at country level since price effects from ENSO in one subnational region would affect all parts of the country through domestic markets. Our treatment variable of interest is the maximum value of the NINO3.4 index between May and December in a given year. We also record the correlation between ENSO temperatures and precipitation at pixel-level within teleconnected countries, in order to differentiate between places where warmer ENSO leads to dry or wet precipitation anomalies. Our estimates separately identify effects in DHS clusters located in first-level administrative units (e.g., state/province) where more than 50% of land area has three months or more per year with a statistically significant positive correlation between precipitation and the second monthly lag of NINO3.4 (8 ).
Microdata on children's health are from the Demographic and Health Surveys (DHS). We identify all standard DHS surveys from teleconnected countries for which children's anthropometric data are available, generating a sample of 1,253,176 child-level observations in 186 surveys from 51 countries between 1986 and 2018. We calculated children's anthropometrics according to the WHO Anthro Child Growth Standard 1 .
We estimate the effect of ENSO on anthropometric measure Y ict for child i living in country c in year t (Table 1) using an equation of the form for NINO3.4 anomalies defined as above for the tropical year t in which the anthropometrics for child i were measured, as well as country-specific controls for whether the household is urban or rural (F E c×rural ), the mother's education in years, and the mother's age at time of child birth. f (t) captures detrending and seasonality adjustments to the data for each of five major UNICEF world regions using both linear year trends and month of survey fixed effects. We normalize the data by country using fixed effects FEcr separately identified for rural and urban location within each country. This research design corresponds to looking at average differences in child outcomes within the same country, separately for rural and urban areas, under different global ENSO states. This allows for identification of ENSO's effect under minimal assumptions of potential confounding (23 , 31 ). Standard errors are two-way clustered at the level of interview year (N = 33) in order to correct for a common global ENSO shock, as well as at the level of subnational first administrative unit (N=532) in order to adjust for serial correlation in anthropometric indicators over time and space.
Most specifications identify β, the effect of ENSO, separately for children living in teleconnected areas where El Niño conditions tend to produce wet anomalies (β p ) and those with neutral or dry anomalies (β n ).
Depending on the specification, the regression weights observations either to produce an estimate of β that represents the effect of ENSO on the average country in our sample (that is, using the DHS sampling weights for observations, normalized such that all observations across all surveys sum to unity for each country), or an estimate of β that represents the effects on the average child in the countries of our sample (combining normalized DHS sampling weights with population weights for each country). In either case, weights adjust for the fact that countries had different numbers of DHS surveys with different sample sizes over the time period.
For Fig. 2B, we utilize the Frisch-Waugh-Lovell theorem and first residualize our outcome, WAZ, and our independent variable, NINO3.4 SST. That is, we run the following regressions separately for locations with negative / neutral precipitation correlation and positive precipitation correlation, weighted as in equation 1: We plot the relationship between these residuals using an Epanechnikov kernel-weighted local polynomial regression with a bandwidth of 0.7 in the residualized x-variable.
The negative association between the ENSO index and undernutrition remains robust to a variety of different model specifications controlling for a variety of different plausible observable and unobservable factors.  (3) normalizes data by rural/urban location, and adds child-level controls, and results in a decrease in the coefficient for positive precipitation correlation locations but no change for other locations, consistent with anthropometric measures being standardized and comparable across age and sex of child. Model (4) flexibly detrends the data separately by UNICEF region and adds country-specific rural/urban fixed effects, resulting in no statistical change in the coefficients while increasing significance levels to 1% confidence. Model (5) adds country-specific interview month fixed effects to remove seasonality separately by country. This stricter specification makes estimates noisier, but they are not statistically different from those in (4). Model (6) controls for mother and child characteristics separately by country, resulting in no significant change in results. Finally, model (7) replaces the country fixed effects with fixed effects in first level administrative unit, thus comparing children living in the same state/province interviewed during different ENSO states. This stricter specification also does not lead to changes in results. Panel B performs all the same specifications using the binary outcome of undernourishment (below -2σ in weight for age). The pattern across specifications is similar: effects are consistent in direction and magnitude across all models, with statistical significance emerging after the data are detrended in model (2).  (1) and (5) use weights as in main results in Table 1 in order produce estimates interpreted as ENSO's effect on a child in the average country. These weights use the DHS sampling weights, adjust for survey size 14 differences across countries, and then adjust for different numbers of DHS surveys in different countries: Models (2) and (6) weight observations such that results are interpreted as the effect on the average child in the sample countries. The weights are the same as above, but now weight across countries by country population: Results with these weights suggest larger effects of ENSO, although estimates are less precisely identified in the case of weight for age (2). Given that India represents 23% of the data and is by far the country with largest population, models (3) and (7) use average child weights but exclude India and find that results on the average child are larger than in (1) and (5) but not statistically different. In order to gauge whether the results might be driven by observation weights, models (4) and (8) (5) and (10) is not statistically significant, while La Niña-like events exhibit statistically significant effects on increasing weight for age and reducing undernutrition, with opposite effects in places where precipitation correlation with NINO3.4 is positive.  Tables S7-S8 test whether results are robust to alternative definitions of teleconnection. Figure S1 shows the global map of ENSO correlations to precipitation and temperature used for teleconnection assignment.   Figure S2 and Tables S10-S11 explore whether the ENSO state might affect the timing of DHS surveys within the year, and therefore spuriously lead to changes in child anthropometrics due to seasonality. Figure   S2 plots sample sizes by month and ENSO state with 95% CIs to show variation in these patterns across DHS surveys. There is no statistically different monthly pattern across ENSO states.   (4) to adjust for arbitrary serial and spatial autocorrelation at these geographic units. Inference is also consistent under two-way clustering at UNICEF region and decade (5), country and year (6), or admin1 and year (7, the main specification). Note that (7) is designed to address possible spatial and temporal autocorrelation through use of two-way clustering of standard errors (32 ) at the levels of interview year and first subnational administrative subunit (e.g., state or province, at which level the DHS is representative). This approach is conservative, and controls for arbitrary autocorrelation within both the level of treatment (i.e. tropical year) across space as well as within the DHS sample frame across multiple surveys at different times.
Finally, Figure S3 shows that a randomization inference test rejects the possibility that the structure of the data is spuriously resulting in the estimated effects on child anthropometrics (33 ). The procedure randomly permutes yearly NINO3.4 values across years, and estimates the main specification on child weight for age (as in Table 1 column 1, excluding the observations with positive precipitation correlation to NINO3.4): The figure plots the distribution of β n estimated from the 2,000 random permutations of NINO3.4. This, in effect, creates 2,000 placebo datasets where every child surveyed in tropical year T, i.e., between May of calendar year T and April of calendar year T+1, is assigned a random NINO3.4 anomaly value from the time series of NINO3.4 without replacement. The distribution of estimates from running our main specification on these 2,000 datasets is the blue shaded region. It is centered at 0, providing evidence that this research design is not biased by, for example, temporal autocorrelation. Further, the coefficient we estimate using the actual NINO3.4 variation (-0.025) is substantially different from zero (p = 0.08). This Fisher randomization inference test illustrates that the residual annual variation in NINO3.4 and the child anthropometric data is meaningful compared to arbitrary degrees of freedom at the annual level.   Figure S3: Randomization inference test indicates association of ENSO state and nutrition unlikely due to chance. The figure plots the distribution of coefficients β n from the following regression matching the estimation in column 1 of Table 1 after excluding locations with positive precipitation correlation: Y ict = α + β n N IN O t + γX i + f (t globalregion ) + F E (c×rural) + ε ict . Y is child weight for age. Estimates of β n are produced after values of NINO3.4 are reshuffled across tropical years. The process is repeated with 2,000 random permutations of NINO3.4. The randomized permutation exercise rejects the possibility that the estimated β n of -0.025 (shown by the vertical line) is a spurious result (p = 0.08). Estimates are from OLS regressions with controls consisting of: fixed effects (indicators) for each country; country-specific controls for mother's age at child's birth and total years of mother's education; country-specific fixed effects for rural vs. urban; and UNICEF world region-specific linear trends in survey year as well as fixed effects for month of interview. Observations are reweighted using DHS sample weights and country size weights in order for estimates to represent an effect on the average country.  Notes: 1-4 examine weight for age z-scores, while 5-8 examine a binary variable for whether the child is undernourished by WHO standards (below -2σ in weight for age). Estimates are from OLS regressions with controls consisting of: fixed effects (indicators) for each country; country-specific controls for mother's age at child's birth and total years of mother's education; country-specific fixed effects for rural vs. urban; and UNICEF world region-specific linear trends in survey year as well as fixed effects for month of interview. Standard errors are two-way clustered at the level of tropical year and admin1 (province) regions. 1 and 5 use weights to represent the effect on an average country in the sample, 2 and 6 weigh observations to represent the average child in the sample; 3 and 7 use the same weights but exclude India given its dominance in the sample. 4 and 8 use no weights. Asterisks indicate statistical significance at the 1% (***), 5% (**) and 10% (*) levels. Notes: 1-5 examine weight for age z-scores, while 6-10 examine a binary variable for whether the child is undernourished by WHO standards (below -2σ in weight for age).  5) and (10) use indicator variables for El Niño-like and La Niña-like states designating all years where the maximum of a three month rolling mean of monthly values was greater than 0.5°C (Niño-like) or less than -0.5°C (Niña-like) from its reference climatology following NOAA CPC guidelines. Estimates are from OLS regressions with controls consisting of: fixed effects (indicators) for each country; country-specific controls for mother's age at child's birth and total years of mother's education; country-specific fixed effects for rural vs. urban; and UNICEF world region-specific linear trends in survey year as well as fixed effects for month of interview. Standard errors are two-way clustered at the level of tropical year and admin1 (province) regions, and observations are reweighted using DHS sample weights and country size weights in order for estimates to represent an effect on the average country. Asterisks indicate statistical significance at the 1% (***), 5% (**) and 10% (*) levels. Notes: Estimates are from OLS regressions with controls consisting of: fixed effects (indicators) for each country; country-specific controls for mother's age at child's birth and total years of mother's education; country-specific fixed effects for rural vs. urban; and UNICEF world region-specific linear trends in survey year as well as fixed effects for month of interview. Standard errors are two-way clustered at the level of tropical year and admin1 (province) regions, and observations are reweighted using DHS sample weights and country size weights in order for estimates to represent an effect on the average child in the sample countries. Asterisks indicate statistical significance at the 1% (***), 5% (**) and 10% (*) levels. Notes: Different anthropometric effects of ENSO are concentrated on short run measures (1-3) weight for age, weight for height, and body mass index z-scores, which all measure shorter-run effects of scarce nutrition, show evidence of contemporaneous ENSO effects measured in°C. Estimates are from OLS regressions with controls consisting of: fixed effects (indicators) for each country; countryspecific controls for mother's age at child's birth and total years of mother's education; countryspecific fixed effects for rural vs. urban; fixed effects for month of interview, and fixed effects for decade of survey year. Standard errors are two-way clustered at the level of tropical year and admin1 (province) regions, and observations are reweighted using DHS sample weights and country size weights in order for estimates to represent an effect on the average country. (4-5) WHO threshold outcomes show ENSO increases the likelihood of being undernourished (below -2σ in weight for age) or wasted (below -2σ in weight for height). Asterisks indicate statistical significance at the 1% (***), 5% (**) and 10% (*) levels. Notes: Different anthropometric effects of ENSO are concentrated on short run measures (1-3) weight for age, weight for height, and body mass index z-scores, which all measure shorter-run effects of scarce nutrition, show evidence of contemporaneous ENSO effects measured in°C. Estimates are from OLS regressions with controls consisting of: fixed effects (indicators) for each country; country-specific controls for mother's age at child's birth and total years of mother's education; country-specific fixed effects for rural vs. urban; and UNICEF world region-specific linear trends in survey year as well as fixed effects for month of interview. Standard errors are two-way clustered at the level of tropical year and admin1 (province) regions, and observations are reweighted using DHS sample weights and country size weights in order for estimates to represent an effect on the average country. (4-5) WHO threshold outcomes show ENSO increases the likelihood of being undernourished (below -2σ in weight for age), but shows a statistically insignificant effect on wasting (below -2σ in weight for height). Asterisks indicate statistical significance at the 1% (***), 5% (**) and 10% (*) levels. for each country; country-specific controls for mother's age at child's birth and total years of mother's education; country-specific fixed effects for rural vs. urban; and UNICEF world region-specific linear trends in survey year as well as fixed effects for month of interview. Standard errors are two-way clustered at the level of tropical year and admin1 (province) regions, and observations are reweighted using DHS sample weights and country size weights in order for estimates to represent an effect on the average country. (4-5) WHO threshold outcomes show ENSO increases the likelihood of being undernourished (below -2σ in weight for age) and wasted (below -2σ in weight for height). Asterisks indicate statistical significance at the 1% (***), 5% (**) and 10% (*) levels. for each country; country-specific controls for mother's age at child's birth and total years of mother's education; country-specific fixed effects for rural vs. urban; and UNICEF world region-specific linear trends in survey year as well as fixed effects for month of interview. Standard errors are two-way clustered at the level of tropical year and admin1 (province) regions, and observations are reweighted using DHS sample weights and country size weights in order for estimates to represent an effect on the average country. (4-5) WHO threshold outcomes show ENSO increases the likelihood of being undernourished (below -2σ in weight for age) and wasted (below -2σ in weight for height). Asterisks indicate statistical significance at the 1% (***), 5% (**) and 10% (*) levels. Notes: Different anthropometric effects of ENSO are concentrated on short run measures (1-3) weight for age, weight for height, and body mass index z-scores, which all measure shorter-run effects of scarce nutrition, show evidence of contemporaneous ENSO effects measured in • C. Estimates are from OLS regressions with controls consisting of: fixed effects (indicators) for each country; countryspecific controls for mother's age at child's birth and total years of mother's education; countryspecific fixed effects for rural vs. urban; country-specific fixed effects for month of interview, and linear and quadratic trends in survey year. Standard errors are two-way clustered at the level of tropical year and admin1 (province) regions, and observations are reweighted using DHS sample weights and country size weights in order for estimates to represent an effect on the average country. Asterisks indicate statistical significance at the 1% (***), 5% (**) and 10% (*) levels. Notes: Estimate is from OLS regressions with controls consisting of: fixed effects (indicators) for each country; country-specific controls for mother's age at child's birth and total years of mother's education; country-specific fixed effects for rural vs. urban; and UNICEF world region-specific linear trends in survey year as well as fixed effects for month of interview. Standard errors are two-way clustered at the level of tropical year and admin1 (province) regions, and observations are reweighted using DHS sample weights and country size weights in order for estimates to represent an effect on the average country. Asterisks indicate statistical significance at the 1% (***), 5% (**) and 10% (*) levels. Notes: Anthropometric effects of ENSO on weight for age show evidence of contemporaneous ENSO effects measured in • C. Estimates are from OLS regressions with controls consisting of: fixed effects (indicators) for each country; country-specific controls for mother's age at child's birth and total years of mother's education; country-specific fixed effects for rural vs. urban; fixed effects for month of interview; and UNICEF world region-specific linear trends in survey year. Inference remains consistent whether standard errors are not clustered (1); clustered at UNICEF region (2), country (3) or admin1/state (4) to adjust for arbitrarily serial and spatial autocorrelation at these geographic units, or two-way clustered at UNICEF region & decade (5), country & year (6), or admin1 & year (7, the main specification). Observations are reweighted using DHS sample weights and country size weights in order for estimates to represent an effect on the average country. Asterisks indicate statistical significance at the 1% (***), 5% (**) and 10% (*) levels.