Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Weather Shocks, Birth and Early Life Health: Evidence of Different Gender Impacts

Weather Shocks, Birth and Early Life Health: Evidence of Different Gender Impacts Abstract This paper examines the impact of exposure to weather events during gestation on birth weight and anthropometric health of a cohort of children. We explore birth records for the cohort of children born between 2003 and 2013 in Sierra Leone using Demographic Health Surveys linked to temporal variation of rainfall and temperature patterns. We find that in utero droughts (or abnormally low precipitation levels) increase the prevalence of low birth weight with larger effects among boys. However, the effects of those same in utero shocks on the prevalence of stunting up to 59 months later are smaller for boys than for girls. The gender difference in estimated impacts from birth to anthropometric health is attributed to food consumption patterns that favour boys. Our results have policy implications for tracking health outcomes during early childhood using birth and anthropometric health, especially by gender. 1. Introduction Birth weight underpins short- to long-term socioeconomic outcomes (Behrman and Rosenzweig, 2004; Black et al., 2007; Almond and Currie, 2011; Baguet and Dumas, 2019; Conti et al., 2020; Bassino et al., 2022). This is also linked to childhood anthropometric health, which helps to convey health trajectory throughout the lifecycle of an individual (Hoddinott and Kinsey, 2001; Case et al., 2005; Case and Paxson, 2008; Almond et al., 2012). The overarching research question addressed in this paper is to investigate if the effects of shock exposure at the in utero period impacts birth outcomes, and if those effects persist into early childhood. As a secondary question, we also examine the dynamic effects of gestational shocks for boys and girls to check if these are different. While extensive body of literature provides evidence on the adverse effects of in utero weather shocks on anthropometric health, there are emerging studies that demonstrate the adverse effects of same on birth outcomes (Deschênes et al., 2009; Molina and Saldarriaga, 2017; Chen et al., 2020; Abiona and Ajefu, 2022). Up to date, little is known about the role of gestation shocks in the progression of child development from birth to short-term outcomes in a way that could inform policy. This gap implies that policymakers lack adequate information to design intervention programs for early life shock exposure. This study seeks to bridge this gap in knowledge by extending the body of evidence in this area of research. To address the main objective of this paper, we reconnect important childhood health measures, using birth and anthropometric outcomes. This way, we are able to examine the transition impacts of early life shocks from alternative vital statistics. The study uses a cohort of children while focusing on the differential enabling pathways of weather patterns to extend frontiers of evidence regarding the literature on the impacts of extreme weather events on child health in Sub-Saharan Africa (SSA). To achieve this, we match vital health statistics (birth weight and anthropometric health) from the Demographic Health Surveys (DHS) to shocks constructed from the University of Delaware weather data archive in order to investigate the impact of weather shocks on early life health outcomes for rural households in Sierra Leone. The mapping process of height-for-age z-scores (HAZ) of children aged 0–59 months to their birth weight constitutes the novel approach adopted in this research1. We hypothesise diverse weather pathways to evaluate the role of weather shocks on the outcomes. This includes the impact of weather patterns through nutrition, disease environment, water scarcity and other stressor factors. Our analysis shows weak impacts of low precipitation level for seasonal and gestation period variations on birth outcomes but we strongly reject the null hypothesis for anthropometric outcomes. Our findings also show that the impact of weather shocks have alternative pathways. For example, the impacts of shock on birth outcomes are partially documented for the low birth weight indicator with a positive association from seasonal drought. Our findings also show that an incidence of seasonal drought decreases HAZ by approximately 32% while increasing the probability of stunting by 39%. We also estimate results for both extremely low and high precipitation level pathways during gestation, respectively. Our results show that both shocks have an impact on HAZ. Exposure to extremely low precipitation level is associated with an average decrease in HAZ by 41%, while extremely low and high precipitation levels is associated with an increase in the probability of stunting by 43% and 24%, respectively. Heterogeneous impacts show an asymmetric impact of gestational weather shock on birth and anthropometric outcomes by gender. We find that in utero droughts (or abnormally low precipitation levels) increase the prevalence of low birth weight with larger effects among boys. However, the effects of those same in utero shocks on the prevalence of stunting up to 59 months later are smaller for boys than for girls. Similarly, we report impacts of temperature deviation but disproportionately larger impacts of lower than average temperature deviation relative to the higher than average component for HAZ. Results from the trimester-level heatwave model (using threshold temperature shock indicators across three trimesters) present persistent average impact of trimester-level stressor factor from birth following through to HAZ. Our findings fill an important gap on the impacts of exposure to weather events during gestation on the trajectory of health outcomes. This paper adds to the literature on the impacts of drought on the nutritional status of children in low-income countries (Hirvonen et al., 2020) and indirectly to the origin of future disproportionate impacts of early life shocks on socioeconomic outcomes by gender (Feeny et al., 2021). Other research shows that early life events may be embedded within other weak environmental factors such as sanitation. For example, Mulmi et al. (2016) find asymmetric gendered impacts across trimester exposure to vegetation levels. Cornwell and Inder (2015) provide evidence in support of both nutrition and disease environment from rainfall shocks on height-for-age measures. Our conceptual framework builds on these studies to disentangle the different weather pathways further while also exploring impacts by gender. We address two important issues. First, we examine the impact of gestational shocks on the trajectory of health by combining birth and anthropometric health for rural children who have a greater level of exposure to weather shocks and require policy intervention. Second, we examine the impacts by gender to highlight resource allocation behaviours. Household allocation of economic resources affects human capital outcomes in low-income countries but there are conflicting findings in this area of research (Abhishek, 2010; Rodríguez, 2016; Kaul, 2018). Hence, our results provide additional evidence for more efficient household resource allocation decisions. Our results also highlight the unequal growth trajectory by gender to help ensure equitable distribution of household resources. The remainder of this paper is organised as follows. In Section 2, we outline the research context and the conceptual framework of the study. Section 3 discusses the data sources and representation of the weather pathways. Section 4 provides the empirical methods. In Section 5, we present estimated results, and discuss the main findings and conclude in Section 6. 2. Setting and conceptual framework 2.1. Research setting Evidence from African countries shows a need for deeper understanding of the impact of extreme weather anomalies on child health outcomes. This paper extends the literature on the impacts from different components of weather shocks. Our research design bridges methods from different studies on child health outcomes from rainfall shocks (Randell et al., 2020; Thiede and Strube, 2020) and from temperature shocks (Geruso and Spears, 2018; Baker and Anttila-Hughes, 2020; Block et al., 2022). Sierra Leone is a low-income country in SSA and its rural communities are largely dependent on subsistence agriculture. Weather patterns relative to norm for planting and harvest seasons affect agricultural yields, and the nutrition available to rural mothers and their children, both during gestation and after birth. In this study, we use rainfall shocks during the pre-birth agricultural seasons and water density metrics to capture shocks during gestation. The rationale for this approach is that rural households are prone to harvest shocks while the extent of impact is unpredictable for urban households. In general, agricultural yields associated with rainfall patterns are a plausible measure of food security in rural Sierra Leone. This pathway requires a lag, separating planting from the harvesting season, to accurately capture household exposure to food insecurity. The shock covers harvest in rural areas for all births within the same locality over the same period of time. On the other hand, the water density and temperature pathways match climatic conditions to accumulated levels during the specific gestation period of each child. The extreme nature of the climatic conditions from these alternative pathways can be regarded as disease environment for extreme positive rainfall (flood) and water scarcity for extreme negative rainfall (droughts) (Almond et al., 2012; Rocha and Soares, 2015). Similarly, above average temperature levels (known as heatwaves) reflect stressor factors during the gestation period (Schetter, 2011; Torche, 2018). Sierra Leone's wet (planting) season runs from April to October each year (Ngegba et al., 2018) and the major crops include cereals and paddy rice. Other main grains are maize, millet and sorghum with mainly subsistence agricultural practices in rural areas. These crops require consistent rainfall hence the need to cultivate them during the wet season. Crop harvests across Sierra Leone commence in October each year following the planting season from April. Rural households' lean season starts just before the harvests when households may have depleted food preserved from the previous cycle. Agricultural extension programs are provided for farmers in this period to support sustainable agricultural practice (Ngegba et al., 2018). We construct drought shocks for the wet season using precipitation levels in the agricultural cycles for rural Sierra Leone. 2.2. Background theory and relevant literature Foetal origin hypothesis underpins the health impacts of exposure to shocks during gestation and also recognises household resource variability among other factors. Many studies in economics have explored this hypothesis empirically (Deschênes et al., 2009; Almond and Currie, 2011) to provide a framework to understand factors affecting important development milestones from childhood to adulthood. To tackle endogeneity concerns, many studies explore natural events providing exogenous variation to estimate causal impacts. Developing effective policy requires an investigation of comprehensive gestation shocks (Deschênes et al., 2009; Andalón et al., 2016) and early life events (Comfort, 2016; Lee and Li, 2021; Balietti et al., 2022; Freudenreich et al., 2022). These are not mutually exclusive events but most studies include each separately for specific policy guidelines. The main objective of this study is to understand if impacts from each of these weather shocks persist through early childhood. Evaluating the impact of these simultaneous shock components on birth and anthropometric health outcomes helps to understand the potential disproportionate impacts of these foetal shocks on children's development pathways. A large body of literature has investigated the impacts of early life weather events focusing on diverse pathways. Transmission of weather events includes malnutrition (Meng and Qian, 2009; Majid, 2015; Block et al., 2022) and disease environment and stress (Deschênes et al., 2009; Andalón et al., 2016; Lee and Li, 2021). Kudamatsu et al. (2012) provide additional context for how extreme weather patterns relate to disease environment and undernutrition and how this increases the infant mortality rate in Africa. Many studies focus on nutrition as the main channel of transmission between weather patterns and birth outcomes or anthropometric health. This is also the focus of the foetal origin hypothesis presented by Almond and Currie (2011). Transient weather shocks affecting seasonal cycles during gestation may distort foetal growth thereby triggering intergenerational effects through their effects on early life health outcomes. The impact of food security on health outcomes is likely to be greater in rural areas where households predominantly depend on rain-fed agricultural practice. While this is quite important, extreme weather events may also be triggered by other supplementary mechanisms beyond scarce economic resources in low-income countries. For example, a disease environment is often triggered during floods, because of unsuitable infrastructure. This can come from dirty water pools that breed insects such as mosquitos during rainy seasons, or a muddy environment that leads to dirty water flows into streams—the main drinking water source for rural households. This shows that both environment and infrastructure play interconnected roles in intergenerational health transmission during gestation. Another mechanism underlying the rainfall pattern during gestation is water access, which may have a direct impact on health outcomes, mostly for rural households that have barriers to potable water during drought seasons (Daghagh Yazd et al., 2020), and in other cases urban water through other mechanisms (Desbureaux and Rodella, 2019). On the other hand, extreme temperatures—heatwave and cold—during gestation can influence early life health outcomes. We broadly categorise weather pathways into nutrition, water density (water scarcity and disease environment) and heatwave. Figure A1 illustrates the selected pathways of weather shocks2. 3. Data 3.1. Birth and anthropometric outcomes We use data from the DHS for Sierra Leone. The Sierra Leone DHS follows the usual DHS two-stage cluster-level sampling procedure for national representative datasets. The first stage involves sampling of clusters within districts, while the second stage includes sampling of the household units within selected clusters. The clusters are designated enumeration areas. The Sierra Leone DHS provides linkage geographic coordinates at the enumeration area level. The DHS reports the geolocation of each enumeration area adjusted by 5 km distance across rural areas to ensure confidentiality of the survey locations and anonymity of the study participants. We use birth weight data linked to the birth records and anthropometric data for children from the Sierra Leone DHS for 2008 and 2013. The birth dates span 11 years from 2003 to 2013. We also use household level demographic variables and those specifically relating to women aged 15–49 years. Birth records include vital statistics such as gender, nature of birth, birth order, information on single or multiple births and other childbirth conditions. In addition to birth weight, we use corresponding HAZ for the sample of children in this study. To strictly maintain a cohort of children in our analysis we drop recent births with no documented record of HAZ and children with HAZ but whose birth weights are missing. We follow the literature to use a gestation period of 38–40 weeks (Rocha and Soares, 2015; Abiona and Ajefu, 2022). The outcome variables are the officially documented birth weight during childbirth (measured in kilogrammes) and the World Health Organisation's standardised HAZ measures from the survey data. We also follow the literature to construct an indicator variable for an occasion of low birth weight (LBW), with 1 for weight less than 2500 g at birth and 0 otherwise. HAZ less than −2 standard deviation units is categorised as stunting in the same manner. 3.2. Weather data To measure weather conditions, we use data from the Center for Climatic Research, University of Delaware. We extract rainfall and temperature datasets from weather stations reporting the terrestrial precipitation and air temperature for 1900–2017 (version 5.01). This archive provides estimates of monthly precipitation and temperature on a 0.5° by 0.5° grid for terrestrial areas across the globe. The rainfall and temperature estimates are based on climatologically aided interpolation of available weather station information made available by Matsuura and Willmott (2017). We use the GPS coordinates provided for each DHS enumeration area across Sierra Leone for the 2008 and 2013 surveys to estimate corresponding weather patterns for baseline weather data required to compute shocks3. To address concerns on selective migration or location sorting across enumeration areas, we restrict our analysis to women surveyed in their permanent place of residence only. Figure 1 presents the distribution of survey clusters in the two survey waves. The administrative borders reflect the 153 chiefdoms across Sierra Leone while the dots represent the sampled clusters—enumeration areas—within each chiefdom for the two waves. Figure 1 Open in new tabDownload slide Distribution of Survey Clusters in Sierra Leone DHS Data 3.3. Shock pathways 3.3.1. Agricultural cycles For empirical analysis, we compute the seasonal mean and standard deviation movements for the historical rainfall period within the same enumeration area and define seasonal droughts using standard deviation movements around the long-term average. These shocks cover all enumeration areas (henceforth designated as localities in this paper) across Sierra Leone between 2002 and 2013 using the methodology in equation (1): $$ \begin{align} {\mathrm{Seasonal}\ \mathrm{drought}}_{ly-1}=1\ if\ {\mathrm{Rainfall}}_{ly-1}<\left(\overline{{\mathrm{Rainfall}}_l}-\left[2\ast \left({\mathrm{Rainfall}}_{SD;l}\right)\right]\ \right),\mathrm{and}\ \mathrm{zero}\ \mathrm{otherwise}, \end{align}$$(1) Where |${\mathrm{Rainfall}}_{ly-1}$| indicates the precipitation level for the reference agricultural season relating to the year of birth of each child within a locality l, and |$\overline{{\mathrm{Rainfall}}_l}$| (⁠|${\mathrm{Rainfall}}_{SD;l}$|⁠) is the average (standard deviation) historical precipitation for a locality in the previous planting seasons covering 30 years. Thus, |${\mathrm{Seasonal}\ \mathrm{drought}}_{ly-1}$| represents an agricultural shock indicator. Figure 2 shows how births across months of the year are calibrated with agricultural cycles for our analysis. Figure 2 Open in new tabDownload slide Matching Agricultural Cycle Across Months of Childbirth—2008 3.3.2. Water density: disease environment and water scarcity The disease environment and water scarcity pathways more directly capture an individual child's exposure to weather events during gestation. We classify gestation exposure to the disease environment and water scarcity as rainfall levels related directly to drought and flood events, and not agricultural practice. The drought measure here is completely different from the drought approach in equation (1) due to the different periodic aggregation of the rainfall pattern but with the same threshold of 2 standard deviation movements. Both the drought and flood shocks capture variation in 9-month gestational period rainfall as follows: $$ \begin{equation} {\textrm{Gestation}\ \textrm{drought}}_{lym}=1\ if\ \sum_{m=-8}^0{\mathrm{R}}_{lm}<\left(\overline{{\mathrm{R}}_l}-\left[2\ast \left({\mathrm{R}}_{SD;l}\right)\right]\right), \ \textrm{and}\ \textrm{zero}\ \textrm{otherwise}, \end{equation}$$(2) $$ \begin{equation} \textrm{Gestation}\ {\textrm{flood}}_{lym}=1\ if\ \sum_{m=-8}^0{\mathrm{R}}_{lm}>\left(\overline{{\mathrm{R}}_l}>{\mathrm{R}}_{SD;l}\right),\ \textrm{and}\ \textrm{zero}\ \textrm{otherwise}, \end{equation}$$(3) where |${\mathrm{R}}_{lm}$| indicates the monthly rainfall during the gestation period specific to each child. We compute accumulated rainfall patterns for each child within locality l from the month of conception (m = −8) to delivery (m = 0). |$\overline{{\mathrm{R}}_l}$| is the average historical rainfall level for child-specific gestational period for each locality in the past 30 years covering a similar period. |${\mathrm{R}}_{SD;l}$| represents the associated standard deviation which depicts the volatility of rainfall levels. We follow Rocha and Soares (2015) by using progressive monthly accumulation of low rainfall levels used to capture water scarcity for Brazilian municipalities while extending this to flood. This is similar to the approach used in Carrillo (2020). Note that weather data distribution inconsistencies explain the asymmetric drought and flood composition in our analysis. For seasonal rainfall, the flood shock thresholds (above 1 or 2 standard deviation movements from the historical average) are unavailable across localities. This implies that we are unable to include an indicator for flood in our seasonal shock regressions. On the other hand, the distribution of the gestation shock thresholds differs for the drought and flood compositions where drought exists for both 1 and 2 standard deviation movements below; but we have up to 1 standard deviation above the norm across localities for the flood component. We chose the extreme drought shock while complementing that with the available gestation flood shock for our empirical context. This particularly enables us to cover shocks for the diverse nature of crops planted by smallholder farmers in rural Sierra Leone. 3.3.3. Temperature shocks We now examine the role of weather-related stress on health outcomes by focusing on gestational temperature variation. We follow the standard literature (Molina and Saldarriaga, 2017) by defining temperature variability as the deviation in the air temperature from each locality's historical mean as follows: $$ \begin{equation} {\mathrm{Temperature}\ \mathrm{shock}}_{lym}=\kern0.5em \left[\frac{1}{9}\ \sum_{m=-8}^0\left({\mathrm{temp}}_{lym}-\overline{{\mathrm{temp}}_l}\right)\ \right]/{SD}_l, \end{equation}$$(4) for a child born in locality |$l$| in year y and month m, where m = {−8; −7;...; 0}. The variable |${\mathrm{temp}}_{lym}$| is the monthly air temperature in the corresponding locality for the |$m$|th month preceding a child's month of birth, |$\overline{{\mathrm{temp}}_l}$| is the local historical moving temperature average for similar periodic progression in the past 30 years, and |${SD}_l$|is the standard deviation of the locality's observed temperature over the same period. While |${\mathrm{Temperature}\ \mathrm{shock}}_{lym}$| captures the temperature exposure for each child during gestation, we are interested in the propensity of extreme temperature measures—both cold spells and heatwaves. We capture this in equation (4) by separating the shock variable into absolute negative and positive deviations, respectively, around the historical norm. In the results section, we also complement negative deviation estimations with trimester-level heatwave exposure indicators4. This depicts the propensity of the exposure to heatwaves at the embryonic, foetal and perinatal stages of gestation indicating the first, second and third trimesters, respectively. Summary statistics of the outcome variables and non-standardised weather parameters are reported in Table 1. The number of observations for 2008 and 2013 varies significantly, reflecting a baseline distribution of samples over the two surveys. Average children's age is 24 months in 2008 and 24.6 months in 2013, and 47.6% and 45.6% of the children are males in 2008 and 2013, respectively. Average birth weight is approximately 3.3 kg across surveys but the incidence of low birth weight is 18.7% for 2008 but just 10% in 2013. This pattern is contrary to a relatively smaller difference in stunting rates of 38.6% in 2008 and 41.2% in 2013. Also, the gendered distribution of birth weight and HAZ in Figure 3 shows more even distribution of these variables for girls. About 12.3% (42.3) of boys and 15.1% (35.7) of girls have low birth weight (stunting). The patterns are consistent with the theoretical proposition of stronger girl foetuses relative to boys due to considerably lower deterioration in anthropometric health. Table 1 Summary Statistics . Survey year . . 2008 . 2013 . Variables Mean Std dev Mean Std dev  No. of observations 422 NA 1,335 NA  Age of child (months) 24.009 15.815 24.589 16.407  Gender of child (male) 0.476 0.500 0.456 0.498 Health outcomes, transition variables  Birth weight (kg) 3.314 1.049 3.258 0.664  Low birth weight (indicator) 0.187 0.391 0.104 0.306  HAZ (standard deviation) −1.433 2.086 −1.505 1.913  Stunting (indicator) 0.386 0.487 0.412 0.492 Weather variables  Total yearly rainfall measure (millimetre) 1603 504 2301 435  Rainfall standard deviation (millimetre) 390 87 438 57  Total yearly temperature (⁠|${}^{\circ}\mathrm{C}$|⁠) 186 3 186 4 . Survey year . . 2008 . 2013 . Variables Mean Std dev Mean Std dev  No. of observations 422 NA 1,335 NA  Age of child (months) 24.009 15.815 24.589 16.407  Gender of child (male) 0.476 0.500 0.456 0.498 Health outcomes, transition variables  Birth weight (kg) 3.314 1.049 3.258 0.664  Low birth weight (indicator) 0.187 0.391 0.104 0.306  HAZ (standard deviation) −1.433 2.086 −1.505 1.913  Stunting (indicator) 0.386 0.487 0.412 0.492 Weather variables  Total yearly rainfall measure (millimetre) 1603 504 2301 435  Rainfall standard deviation (millimetre) 390 87 438 57  Total yearly temperature (⁠|${}^{\circ}\mathrm{C}$|⁠) 186 3 186 4 Notes: Summary statistics are reported for 1,757 observations for childbirth records spanning 2003–2008 from the 2008 DHS survey and 2008–2013 from the 2013 DHS survey. Sample includes rural localities only. Low birth weight indicator is designated 1 if birth weight measure is less than 2.5 kg—and 0 otherwise; stunting is regarded as 1 if HAZ is below −2 standard deviations—and 0 otherwise. The sample of observations is captured from women with surviving babies only and restricted to a cohort of children with corresponding measures of birth weight and HAZ. Open in new tab Table 1 Summary Statistics . Survey year . . 2008 . 2013 . Variables Mean Std dev Mean Std dev  No. of observations 422 NA 1,335 NA  Age of child (months) 24.009 15.815 24.589 16.407  Gender of child (male) 0.476 0.500 0.456 0.498 Health outcomes, transition variables  Birth weight (kg) 3.314 1.049 3.258 0.664  Low birth weight (indicator) 0.187 0.391 0.104 0.306  HAZ (standard deviation) −1.433 2.086 −1.505 1.913  Stunting (indicator) 0.386 0.487 0.412 0.492 Weather variables  Total yearly rainfall measure (millimetre) 1603 504 2301 435  Rainfall standard deviation (millimetre) 390 87 438 57  Total yearly temperature (⁠|${}^{\circ}\mathrm{C}$|⁠) 186 3 186 4 . Survey year . . 2008 . 2013 . Variables Mean Std dev Mean Std dev  No. of observations 422 NA 1,335 NA  Age of child (months) 24.009 15.815 24.589 16.407  Gender of child (male) 0.476 0.500 0.456 0.498 Health outcomes, transition variables  Birth weight (kg) 3.314 1.049 3.258 0.664  Low birth weight (indicator) 0.187 0.391 0.104 0.306  HAZ (standard deviation) −1.433 2.086 −1.505 1.913  Stunting (indicator) 0.386 0.487 0.412 0.492 Weather variables  Total yearly rainfall measure (millimetre) 1603 504 2301 435  Rainfall standard deviation (millimetre) 390 87 438 57  Total yearly temperature (⁠|${}^{\circ}\mathrm{C}$|⁠) 186 3 186 4 Notes: Summary statistics are reported for 1,757 observations for childbirth records spanning 2003–2008 from the 2008 DHS survey and 2008–2013 from the 2013 DHS survey. Sample includes rural localities only. Low birth weight indicator is designated 1 if birth weight measure is less than 2.5 kg—and 0 otherwise; stunting is regarded as 1 if HAZ is below −2 standard deviations—and 0 otherwise. The sample of observations is captured from women with surviving babies only and restricted to a cohort of children with corresponding measures of birth weight and HAZ. Open in new tab Figure 3 Open in new tabDownload slide Density of Birth Weight and Age-Standardised Height Measures 4. Empirical methods We model econometric equations using exogenous variation in weather patterns for our identification strategy. Our main empirical approach focuses on modelling weather across pathways to draw inference on the impacts of gestational shocks on birth and early life health outcomes. The baseline equations for the seasonal drought, water density (disease environment and water scarcity) and temperature pathways are specified as follows: Seasonal low precipitation level model $$ \begin{equation} {\mathrm{Health}\ \mathrm{outcome}}_{ilym}={\alpha}_{lm}+{\varnothing}_{ym}+{\beta}_1{\mathrm{Seasonal}\ \mathrm{drought}}_{ly-1}+{X}_i^{\prime }\ {\theta}_x+{Z}_{ct}^{\prime }\ {\theta}_z+{\varepsilon}_{ilym} \end{equation}$$(5) Gestational low and high precipitation levels model $$ \begin{align} &{\mathrm{Health}\ \mathrm{outcome}}_{ilym}={\alpha}_{lm}+{\varnothing}_{ym}+{\mu}_1{\mathrm{Gestation}\ \mathrm{drought}}_{lym}+{\mu}_2\mathrm{Gestation}\ {\mathrm{flood}}_{lym} +{X}_i^{\prime }\ {\theta}_x+{Z}_{ct}^{\prime }\ {\theta}_z+{\varepsilon}_{ilym} \end{align}$$(6) Temperature variation model $$ \begin{equation} {\mathrm{Health}\ \mathrm{outcome}}_{ilym}={\alpha}_{lm}+{\varnothing}_{ym}+\kern0.5em {\delta}_1{\mathrm{Temperature}\ \mathrm{shock}}_{lym}+{X}_i^{\prime }\ {\theta}_x+{Z}_{ct}^{\prime }\ {\theta}_z+{\varepsilon}_{ilym} \end{equation}$$(7) Where |${\mathrm{Health}\ \mathrm{outcome}}_{ilym}$| represents health variables—namely birth weight and HAZ—for an observation i in localitylfor a child born in year y and month m using the cohort of children identified in the data section. We estimate the same set of models for low birth weight and stunting indicators to capture intergenerational health transition impacts of the shocks at the threshold level. |${\alpha}_{lm}$| is the locality-of-birth by month-of-birth fixed effects and |${\varnothing}_{ym}$| is the year-of-birth by month-of-birth fixed effects. |${\beta}_1$| is the parameter of interest in equation (5). This parameter measures the impact of exposure to locality harvests from low rainfall. Agricultural season droughts in the lead up to a child's birth are an important determinant of crop harvests and the nutritional intake of the mother during the child's gestation. The baseline impacts for the disease environment and water scarcity pathways are derived from parameters |${\mu}_1$| and |${\mu}_2$| in equation (6), while equation (7) examines the impact of temperature variation using parameter |${\delta}_1$|⁠. In the regression analysis we include a vector of individual level controls, denoted as X in equations (5)–(7), for the child, mothers and their partners. We control for the gender of the child to capture differential outcomes for boys and girls while including age-in-months fixed effects to control for non-linear HAZ responses to shocks across age in months (Rabassa et al., 2014). Mothers' demographic characteristics include age during survey, age at first birth, and indicators for education category and marital status. Partners' covariates consist of demographic characteristics including the individual's age group, education and occupation categories. Z is an array of household level controls including household size, age of household head and an indicator variable for the gender of the head of household. In these models, the error term (⁠|${\varepsilon}_{ilym}$|⁠) accounts for the unobserved time-variant locality characteristics. We estimate equations (5)–(7) by including locality-of-birth by month-of-birth fixed effects (also referred to as birth cohort fixed effects in this paper) to capture unobserved heterogeneity across localities. Note that the variables in the error term are orthogonal to a child's exposure to shock during gestation, which leads to unbiassed estimates of rainfall and temperature shocks on the outcome variables. Our main assumption is an identically and independently distributed (iid) error term across localities, which is correlated within each locality due to spatial correlation of rainfall and temperature patterns for households in the same locality. Hence, standard errors are clustered at the locality level. Our empirical strategy primarily focuses on the temporal weather variation that comes from the locality-of-birth by month-of-birth fixed effects. This means that our models compare children born in the same locality and same month but in different years. This approach is better than the spatial identification variation and avoids natural location selection concerns that may arise from the data. In the regression analysis, we estimate each pathway separately. For equation (5), the identification strategy exploits seasonal rainfall variation and locality concentration of births. This specification uses birth cohort fixed effects to identify impacts of weather shocks on health outcomes. Exogenous variation in the agricultural cycle rainfall across localities is important for the causal interpretation of estimates in this model. Our data shows evidence of established widespread locality-level drought distribution for both cross-regional differences and temporal rainfall variation over a 12-year period that is deemed sufficient to identify the exogenous impact of harvests5. For direct gestational shocks encompassing disease environment, water scarcity and temperature, identification is based on rainfall and temperature variation across individual-level gestational period. This methodology provides a more extensive variation in the weather patterns plausible for specific impacts of weather shocks. 5. Results 5.1. Main results 5.1.1. Gestational low precipitation levels Table 2 presents coefficient estimates of impacts of seasonal drought shock on birth weight and anthropometric outcomes from regressions of equation (5). These are treatment effect estimates to test the impact of gestational nutrition on health outcomes. These results capture locality rainfall variations underlying agricultural harvests, and how they affect birth outcomes and anthropometric health in Panels A and B. We mainly provide Table 2 results from alternative models for equation (5) where we exclude the year-of-birth by month-of-birth fixed effects in columns (1) and (3) while reporting fully-specified models in columns (2) and (4). Each regression model includes all controls, month-of-birth fixed effect and locality-of-birth by month-of-birth fixed effects. In Panel A, baseline estimates6 for low birth weight show a positive correlation of 8.3 percentage points with seasonal drought shock. We also estimate that seasonal drought shock incidence is associated with a decline in the HAZ by 0.47 standard deviations (Panel B). This result is reinforced by the estimated increase of stunting by 15.9 percentage points. Estimated coefficients are statistically significant at traditional levels across Panels A and B. Table 2 Impact of Seasonal Drought Shock on Health Outcomes in Rural Sierra Leone . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Seasonal drought −0.086 −0.084 0.070 0.083* (0.098) (0.099) (0.047) (0.046)  Constant 3.112*** 3.111*** 0.141* 0.118 (0.180) (0.178) (0.075) (0.079)  R-squared 0.371 0.378 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Seasonal drought −0.438* −0.473** 0.144** 0.159*** (0.236) (0.240) (0.060) (0.060)  Constant −0.553 −0.535 0.259** 0.204* (0.437) (0.443) (0.117) (0.119)  R-squared 0.357 0.362 0.292 0.302  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Seasonal drought −0.086 −0.084 0.070 0.083* (0.098) (0.099) (0.047) (0.046)  Constant 3.112*** 3.111*** 0.141* 0.118 (0.180) (0.178) (0.075) (0.079)  R-squared 0.371 0.378 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Seasonal drought −0.438* −0.473** 0.144** 0.159*** (0.236) (0.240) (0.060) (0.060)  Constant −0.553 −0.535 0.259** 0.204* (0.437) (0.443) (0.117) (0.119)  R-squared 0.357 0.362 0.292 0.302  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 Notes: Table 2 presents coefficient estimates of regressions of seasonal drought shock on multiple outcomes for rural surviving births in Sierra Leone between 2003 and 2013. Panels A and B present estimated results for child health outcomes: birth weight (kg) and anthropometric health—using age-standardised height z-scores—HAZ (sd). Low birth weight and stunting indicators are used as supplementary outcomes. These measure proportion of child observations below 2.5 kg for birth weight and below −2 standard deviations for HAZ. Explanatory drought variable follows equation (1) for seasonal drought representing 2 standard deviations below the norm. Each column for each panel presents a separate regression. Columns (1) and (3) present estimated coefficients from regressions including all controls and locality-of-birth by month-of-birth fixed effects while columns (2) and (4) additionally include year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. Robust standard errors, clustered at the locality level, are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab Table 2 Impact of Seasonal Drought Shock on Health Outcomes in Rural Sierra Leone . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Seasonal drought −0.086 −0.084 0.070 0.083* (0.098) (0.099) (0.047) (0.046)  Constant 3.112*** 3.111*** 0.141* 0.118 (0.180) (0.178) (0.075) (0.079)  R-squared 0.371 0.378 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Seasonal drought −0.438* −0.473** 0.144** 0.159*** (0.236) (0.240) (0.060) (0.060)  Constant −0.553 −0.535 0.259** 0.204* (0.437) (0.443) (0.117) (0.119)  R-squared 0.357 0.362 0.292 0.302  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Seasonal drought −0.086 −0.084 0.070 0.083* (0.098) (0.099) (0.047) (0.046)  Constant 3.112*** 3.111*** 0.141* 0.118 (0.180) (0.178) (0.075) (0.079)  R-squared 0.371 0.378 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Seasonal drought −0.438* −0.473** 0.144** 0.159*** (0.236) (0.240) (0.060) (0.060)  Constant −0.553 −0.535 0.259** 0.204* (0.437) (0.443) (0.117) (0.119)  R-squared 0.357 0.362 0.292 0.302  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 Notes: Table 2 presents coefficient estimates of regressions of seasonal drought shock on multiple outcomes for rural surviving births in Sierra Leone between 2003 and 2013. Panels A and B present estimated results for child health outcomes: birth weight (kg) and anthropometric health—using age-standardised height z-scores—HAZ (sd). Low birth weight and stunting indicators are used as supplementary outcomes. These measure proportion of child observations below 2.5 kg for birth weight and below −2 standard deviations for HAZ. Explanatory drought variable follows equation (1) for seasonal drought representing 2 standard deviations below the norm. Each column for each panel presents a separate regression. Columns (1) and (3) present estimated coefficients from regressions including all controls and locality-of-birth by month-of-birth fixed effects while columns (2) and (4) additionally include year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. Robust standard errors, clustered at the locality level, are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab The results show marginal statistical precision for estimated results for birth outcomes (precisely low birth weight)7. Lack of statistical power for treatment effects for birth weight may be due to the small sample composition captured due to the data restriction imposed while linking it to HAZ. In particular, the weak birth weight findings may be partially driven by the imprecise birth weight measures with the documented uneven birth weight patterns using our sample (Figure 3). However, these results do not necessarily contradict evidence documenting the impact of nutrition during gestation on foetal development stages in related studies (Burlando, 2014; Andalón et al., 2016). Robust precision and persistence of estimates across anthropometric health outcomes can be attributable to limited implication of the above restriction on the composition of the reference HAZ sample. Using coefficient estimates from the baseline results to evaluate the impacts of these estimates, an incidence of seasonal drought shock is associated with a 67% increase in the likelihood of low birth weight; a 32% decrease in HAZ; and a 39% increase in the probability of stunting when compared to the mean. Our research design is opposed to a relationship between dry season shocks and birth and/or anthropometric health. To test this alternative pathway, we run tests for both birth and anthropometric outcomes using the underlying drought variation in rainfall for consecutive out-of-season months8. In Table A1, we present the coefficients for the fully-specified fixed effect models—which includes locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects—with a full set of controls. All coefficients are quite small and statistically insignificant, supporting our initial proposition that estimated impacts for the anthropometric health outcomes are associated with only agricultural shocks. This result provides additional credibility for nutrition as the underlying channel for the estimated results in Table 2. 5.1.2. Gestational low and high precipitation levels Our theoretical framework recognises the water density pathway by linking this to the gestation period for various dimensions of early life impacts. While other studies have examined these direct exposures relating to environmental factors such as water pollution and vegetation on a range of child health outcomes (Brainerd and Menon, 2014; Mulmi et al., 2016), this paper is the first to investigate weather-related gestation outcomes for the transition of birth to anthropometric health. Potential disaggregated pathways for water in this framework include potable water accessibility, disease environment, and infrastructure. Access to potable water is an important direct pathway of rainfall to birth outcomes, which may persist to short-term. This is particularly important in rural Sierra Leone where households rely on streams, rivers and untreated wells for domestic water. On one hand, we capture access to water directly tied to the gestational months by measuring the foetal exposure to drought during conception. On the other hand, we use a flood indicator to measure the impact of disease environment in the same setting. We combine both water access and disease environment as explanatory variables of interest in the same model using equation (6). We report results of water density pathways in Table 3. While there are no notable results for birth outcomes, both gestation drought and flood shocks demonstrate results consistent with the expected pattern for anthropometric health. The drought shock shows a strong and persistent impact on HAZ, reinforcing the role of water scarcity during gestation. Our results also uphold evidence for the disease environment for stunting. While it is difficult to interpret these results as disproportionate impacts due to the asymmetric composition of the drought and flood shocks from equations (2) and (3), there is suggestive evidence that the resultant drought shock may constitute other stress factors related to water scarcity for rural women in Sierra Leone. In Panel B column (2), the estimated impact of drought shock on HAZ is a reduction of 0.61 standard deviations. Similarly, estimated impacts for stunting in column (4) show an increase of 17.5 and 9.7 percentage points for drought and flood respectively. These results translate to a decrease in HAZ of 41% association with gestational drought—with a corresponding increase in the likelihood of stunting by 43% for gestational drought shocks and 24% for flood shocks. Table 3 Impact of gestation drought and flood shocks on health outcomes in rural Sierra Leone . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Gestation drought −0.070 −0.083 0.037 0.062 (0.104) (0.109) (0.050) (0.049)  Gestation flood 0.064 0.067 0.037* 0.033 (0.047) (0.050) (0.022) (0.022)  Constant 3.100*** 3.113*** 0.132* 0.114 (0.181) (0.178) (0.075) (0.079)  R-squared 0.371 0.379 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Gestation drought −0.598*** −0.606** 0.167*** 0.175*** (0.225) (0.235) (0.057) (0.060)  Gestation flood −0.145 −0.149 0.109*** 0.097*** (0.126) (0.137) (0.033) (0.035)  Constant −0.548 −0.558 0.259** 0.224* (0.436) (0.445) (0.115) (0.120)  R-squared 0.360 0.364 0.298 0.306  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Gestation drought −0.070 −0.083 0.037 0.062 (0.104) (0.109) (0.050) (0.049)  Gestation flood 0.064 0.067 0.037* 0.033 (0.047) (0.050) (0.022) (0.022)  Constant 3.100*** 3.113*** 0.132* 0.114 (0.181) (0.178) (0.075) (0.079)  R-squared 0.371 0.379 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Gestation drought −0.598*** −0.606** 0.167*** 0.175*** (0.225) (0.235) (0.057) (0.060)  Gestation flood −0.145 −0.149 0.109*** 0.097*** (0.126) (0.137) (0.033) (0.035)  Constant −0.548 −0.558 0.259** 0.224* (0.436) (0.445) (0.115) (0.120)  R-squared 0.360 0.364 0.298 0.306  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 Notes: Table 3 presents coefficient estimates of regressions of gestation drought and flood shocks on multiple outcomes for rural surviving births in Sierra Leone between 2003 and 2013. Panels A and B present estimated results for child health outcomes: birth weight (kg) and anthropometric health—using age-standardised height z-scores—HAZ (sd). Low birth weight and stunting indicators are used as supplementary outcomes. These measure proportion of child observations below 2.5 kg for birth weight and below −2 standard deviations for HAZ. Explanatory variables follow equations (2) and (3) for gestation shocks representing below 2 standard deviations of the norm for gestational drought and above 1 standard deviation of the norm for gestational flood, respectively. Each column for each panel presents a separate regression. Columns (1) and (3) present estimated coefficients from regressions including all controls and locality-of-birth by month-of-birth fixed effects while columns (2) and (4) additionally include year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. Robust standard errors, clustered at the locality level, are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab Table 3 Impact of gestation drought and flood shocks on health outcomes in rural Sierra Leone . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Gestation drought −0.070 −0.083 0.037 0.062 (0.104) (0.109) (0.050) (0.049)  Gestation flood 0.064 0.067 0.037* 0.033 (0.047) (0.050) (0.022) (0.022)  Constant 3.100*** 3.113*** 0.132* 0.114 (0.181) (0.178) (0.075) (0.079)  R-squared 0.371 0.379 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Gestation drought −0.598*** −0.606** 0.167*** 0.175*** (0.225) (0.235) (0.057) (0.060)  Gestation flood −0.145 −0.149 0.109*** 0.097*** (0.126) (0.137) (0.033) (0.035)  Constant −0.548 −0.558 0.259** 0.224* (0.436) (0.445) (0.115) (0.120)  R-squared 0.360 0.364 0.298 0.306  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Gestation drought −0.070 −0.083 0.037 0.062 (0.104) (0.109) (0.050) (0.049)  Gestation flood 0.064 0.067 0.037* 0.033 (0.047) (0.050) (0.022) (0.022)  Constant 3.100*** 3.113*** 0.132* 0.114 (0.181) (0.178) (0.075) (0.079)  R-squared 0.371 0.379 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Gestation drought −0.598*** −0.606** 0.167*** 0.175*** (0.225) (0.235) (0.057) (0.060)  Gestation flood −0.145 −0.149 0.109*** 0.097*** (0.126) (0.137) (0.033) (0.035)  Constant −0.548 −0.558 0.259** 0.224* (0.436) (0.445) (0.115) (0.120)  R-squared 0.360 0.364 0.298 0.306  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 Notes: Table 3 presents coefficient estimates of regressions of gestation drought and flood shocks on multiple outcomes for rural surviving births in Sierra Leone between 2003 and 2013. Panels A and B present estimated results for child health outcomes: birth weight (kg) and anthropometric health—using age-standardised height z-scores—HAZ (sd). Low birth weight and stunting indicators are used as supplementary outcomes. These measure proportion of child observations below 2.5 kg for birth weight and below −2 standard deviations for HAZ. Explanatory variables follow equations (2) and (3) for gestation shocks representing below 2 standard deviations of the norm for gestational drought and above 1 standard deviation of the norm for gestational flood, respectively. Each column for each panel presents a separate regression. Columns (1) and (3) present estimated coefficients from regressions including all controls and locality-of-birth by month-of-birth fixed effects while columns (2) and (4) additionally include year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. Robust standard errors, clustered at the locality level, are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab 5.1.3. Gestational temperature We also estimate the impact of gestation temperature shock using the deviation measure in equation (4). Results of the regression of health outcomes on temperature—equation 6—are reported in Table 4. Similar to the pattern shown for rainfall shocks, the estimated impacts for birth outcomes lack statistical power. However, the estimated patterns are considerably strong and statistically significant for anthropometric outcomes. Coefficient estimates are also consistent with a priori theoretical expectation especially with the distribution of impacts across negative and positive temperature shocks (columns (3) and (4)). Negative (positive) temperature deviation is associated with an average increase (decrease) of 0.80 (0.25) standard deviations in HAZ and an average decrease (increase) in stunting of 32.4 (1.8) percentage points. All coefficient estimates are statistically significant except for estimated positive temperature for stunting. The results generally indicate that there is a positive association between linear temperature deviation and HAZ while the negative deviation increase is disproportionate relative to the observed decline from the positive temperature deviation. This asymmetric pattern from disaggregated temperature could be explained by the lack of extreme positive temperature shocks (heatwaves) observed in the data wherein the negative temperature shocks are also mild. In general, the negative temperature measure reflects favourable gestational temperature weather conditions. We complement our findings with an alternative heatwave indicator for average trimester-level stress factors. Table A2 shows that second trimester heatwave is strongly associated with both birth and anthropometric outcomes as opposed to the divergent impacts in Table 4. Table 4 Impact of gestational temperature shocks on health outcomes in rural Sierra Leone Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator) Birth weight (kg) Low birth weight (indicator)  Linear temperature deviation −0.006 0.016 (0.020) (0.010)  Negative temperature deviation −0.044 −0.024 (0.098) (0.048)  Positive temperature deviation −0.015 0.014 (0.028) (0.014)  Constant 3.113*** 0.134* 3.122*** 0.135* (0.181) (0.076) (0.180) (0.078)  R-squared 0.370 0.283 0.370 0.283 Mean dependent variable (birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (SD) Stunting (indicator) HAZ (SD) Stunting (indicator)  Linear temperature deviation −0.334*** 0.066*** (0.061) (0.017)  Negative temperature deviation 0.801*** −0.324*** (0.288) (0.079)  Positive temperature deviation −0.248*** 0.018 (0.083) (0.022)  Constant −1.191*** 0.403*** −1.281*** 0.453*** (0.434) (0.116) (0.433) (0.114)  R-squared 0.344 0.278 0.345 0.284  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Observations 1,677 1,677 1,677 1,677 Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator) Birth weight (kg) Low birth weight (indicator)  Linear temperature deviation −0.006 0.016 (0.020) (0.010)  Negative temperature deviation −0.044 −0.024 (0.098) (0.048)  Positive temperature deviation −0.015 0.014 (0.028) (0.014)  Constant 3.113*** 0.134* 3.122*** 0.135* (0.181) (0.076) (0.180) (0.078)  R-squared 0.370 0.283 0.370 0.283 Mean dependent variable (birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (SD) Stunting (indicator) HAZ (SD) Stunting (indicator)  Linear temperature deviation −0.334*** 0.066*** (0.061) (0.017)  Negative temperature deviation 0.801*** −0.324*** (0.288) (0.079)  Positive temperature deviation −0.248*** 0.018 (0.083) (0.022)  Constant −1.191*** 0.403*** −1.281*** 0.453*** (0.434) (0.116) (0.433) (0.114)  R-squared 0.344 0.278 0.345 0.284  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Observations 1,677 1,677 1,677 1,677 Notes: Table 4 presents coefficient estimates of regressions of gestation temperature shocks on multiple outcomes for rural surviving births in Sierra Leone between 2003 and 2013. Panels A and B present estimated results for child health outcomes: birth weight (kg) and anthropometric health—using age standardised height z-scores—HAZ (sd). Low birth weight and stunting indicators are used as supplementary outcomes. These measure proportion of child observations below 2.5 kg for birth weight and below −2 standard deviations for HAZ. Explanatory heatwave variable follows equation (4) for gestational temperature reference relative to historical timeline of the gestation period. Each column for each panel presents a separate regression. Columns (1) and (2) present estimated coefficients for linear deviation from historical air temperature shock while columns (3) and (4) disaggregate deviation to negative for absolute values for deviation lower than norm and positive for deviations above norm. All regressions include all controls, locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. Robust standard errors clustered at the locality level are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab Table 4 Impact of gestational temperature shocks on health outcomes in rural Sierra Leone Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator) Birth weight (kg) Low birth weight (indicator)  Linear temperature deviation −0.006 0.016 (0.020) (0.010)  Negative temperature deviation −0.044 −0.024 (0.098) (0.048)  Positive temperature deviation −0.015 0.014 (0.028) (0.014)  Constant 3.113*** 0.134* 3.122*** 0.135* (0.181) (0.076) (0.180) (0.078)  R-squared 0.370 0.283 0.370 0.283 Mean dependent variable (birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (SD) Stunting (indicator) HAZ (SD) Stunting (indicator)  Linear temperature deviation −0.334*** 0.066*** (0.061) (0.017)  Negative temperature deviation 0.801*** −0.324*** (0.288) (0.079)  Positive temperature deviation −0.248*** 0.018 (0.083) (0.022)  Constant −1.191*** 0.403*** −1.281*** 0.453*** (0.434) (0.116) (0.433) (0.114)  R-squared 0.344 0.278 0.345 0.284  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Observations 1,677 1,677 1,677 1,677 Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator) Birth weight (kg) Low birth weight (indicator)  Linear temperature deviation −0.006 0.016 (0.020) (0.010)  Negative temperature deviation −0.044 −0.024 (0.098) (0.048)  Positive temperature deviation −0.015 0.014 (0.028) (0.014)  Constant 3.113*** 0.134* 3.122*** 0.135* (0.181) (0.076) (0.180) (0.078)  R-squared 0.370 0.283 0.370 0.283 Mean dependent variable (birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (SD) Stunting (indicator) HAZ (SD) Stunting (indicator)  Linear temperature deviation −0.334*** 0.066*** (0.061) (0.017)  Negative temperature deviation 0.801*** −0.324*** (0.288) (0.079)  Positive temperature deviation −0.248*** 0.018 (0.083) (0.022)  Constant −1.191*** 0.403*** −1.281*** 0.453*** (0.434) (0.116) (0.433) (0.114)  R-squared 0.344 0.278 0.345 0.284  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Observations 1,677 1,677 1,677 1,677 Notes: Table 4 presents coefficient estimates of regressions of gestation temperature shocks on multiple outcomes for rural surviving births in Sierra Leone between 2003 and 2013. Panels A and B present estimated results for child health outcomes: birth weight (kg) and anthropometric health—using age standardised height z-scores—HAZ (sd). Low birth weight and stunting indicators are used as supplementary outcomes. These measure proportion of child observations below 2.5 kg for birth weight and below −2 standard deviations for HAZ. Explanatory heatwave variable follows equation (4) for gestational temperature reference relative to historical timeline of the gestation period. Each column for each panel presents a separate regression. Columns (1) and (2) present estimated coefficients for linear deviation from historical air temperature shock while columns (3) and (4) disaggregate deviation to negative for absolute values for deviation lower than norm and positive for deviations above norm. All regressions include all controls, locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. Robust standard errors clustered at the locality level are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab 5.2. Heterogeneous impacts by gender We estimate heterogeneous impacts of seasonal and gestational rainfall shocks across boys and girls to examine the gender dimension of our treatment effects. Table 5 presents fully specified models for birth outcomes (birth weight and low birth weight) from seasonal shock (Panel A) and gestation shocks (Panel B) for a comparable sample of observations across boys and girls. Table 6 presents similar heterogeneous results for HAZ. In Table 5, we document striking results where previously muted coefficient estimates are statistically significant when we estimate the impacts of shocks on birth weight for boys. Here, the incidence of seasonal drought shock is associated with a decrease in birth weight of 412 g, which corresponds to an increase in the incidence of low birth weight by 12.2 percentage points. Gestation drought shock shows an association of a decline in birth weight of 391 g while the estimated impact of flood shock is an increase of 132 grams—not statistically significant. Coefficient estimates for birth weight of girls are much smaller in magnitude and not precisely estimated (columns (3) and (4)). These pronounced asymmetric impacts for boys are consistent with a priori theoretical evidence of relatively weaker boy foetuses with similar exposure to adverse shocks when compared to girls. Surprisingly, Table 6 reports a concentration of treatment effects for girls when focussing on girls. Here, coefficient estimates for boys are generally smaller compared to estimated effects for girls in a similar context and mostly statistically insignificant9. Results for girls in columns (3) and (4) are also much bigger than the baseline coefficients for the total sample of observations reported in Tables 2 and 3 respectively. The results also reflect asymmetric impacts of gestation drought and flood shocks previously documented in the baseline results more prominently10. Table 5 Gendered impact of weather shocks on birth outcomes . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Seasonal drought −0.412*** 0.122** −0.076 0.129 (0.144) (0.061) (0.143) (0.096)  Constant 3.223*** 0.060 3.358*** 0.052 (0.257) (0.103) (0.218) (0.137)  R-squared 0.562 0.458 0.432 0.390  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Panel B: direct gestation Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Gestation drought −0.391** 0.066 0.082 0.067 (0.167) (0.070) (0.176) (0.102)  Gestation flood 0.132 0.003 0.037 0.026 (0.100) (0.035) (0.054) (0.034)  Constant 3.212*** 0.060 3.371*** 0.051 (0.255) (0.103) (0.218) (0.138)  R-squared 0.563 0.455 0.433 0.388  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Observations 703 703 825 825 . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Seasonal drought −0.412*** 0.122** −0.076 0.129 (0.144) (0.061) (0.143) (0.096)  Constant 3.223*** 0.060 3.358*** 0.052 (0.257) (0.103) (0.218) (0.137)  R-squared 0.562 0.458 0.432 0.390  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Panel B: direct gestation Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Gestation drought −0.391** 0.066 0.082 0.067 (0.167) (0.070) (0.176) (0.102)  Gestation flood 0.132 0.003 0.037 0.026 (0.100) (0.035) (0.054) (0.034)  Constant 3.212*** 0.060 3.371*** 0.051 (0.255) (0.103) (0.218) (0.138)  R-squared 0.563 0.455 0.433 0.388  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Observations 703 703 825 825 Notes: Table 5 reports heterogeneous impacts of estimated impacts of rainfall shocks on birth outcomes for boys and girls. Panel A presents results for seasonal drought, while Panel B presents estimated results for gestation drought and flood. These results refer to the baseline results in Columns (2) and (4) for Panel A of Tables 2 and 3. All regressions include all controls, locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. See Tables 2 and 3 for additional notes. Robust standard errors clustered at the locality level are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab Table 5 Gendered impact of weather shocks on birth outcomes . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Seasonal drought −0.412*** 0.122** −0.076 0.129 (0.144) (0.061) (0.143) (0.096)  Constant 3.223*** 0.060 3.358*** 0.052 (0.257) (0.103) (0.218) (0.137)  R-squared 0.562 0.458 0.432 0.390  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Panel B: direct gestation Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Gestation drought −0.391** 0.066 0.082 0.067 (0.167) (0.070) (0.176) (0.102)  Gestation flood 0.132 0.003 0.037 0.026 (0.100) (0.035) (0.054) (0.034)  Constant 3.212*** 0.060 3.371*** 0.051 (0.255) (0.103) (0.218) (0.138)  R-squared 0.563 0.455 0.433 0.388  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Observations 703 703 825 825 . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Seasonal drought −0.412*** 0.122** −0.076 0.129 (0.144) (0.061) (0.143) (0.096)  Constant 3.223*** 0.060 3.358*** 0.052 (0.257) (0.103) (0.218) (0.137)  R-squared 0.562 0.458 0.432 0.390  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Panel B: direct gestation Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Gestation drought −0.391** 0.066 0.082 0.067 (0.167) (0.070) (0.176) (0.102)  Gestation flood 0.132 0.003 0.037 0.026 (0.100) (0.035) (0.054) (0.034)  Constant 3.212*** 0.060 3.371*** 0.051 (0.255) (0.103) (0.218) (0.138)  R-squared 0.563 0.455 0.433 0.388  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Observations 703 703 825 825 Notes: Table 5 reports heterogeneous impacts of estimated impacts of rainfall shocks on birth outcomes for boys and girls. Panel A presents results for seasonal drought, while Panel B presents estimated results for gestation drought and flood. These results refer to the baseline results in Columns (2) and (4) for Panel A of Tables 2 and 3. All regressions include all controls, locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. See Tables 2 and 3 for additional notes. Robust standard errors clustered at the locality level are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab Table 6 Gendered impact of weather shocks on anthropometric health outcomes . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Seasonal drought 0.317 0.068 −1.244*** 0.245*** (0.414) (0.115) (0.361) (0.094)  Constant −0.550 0.305 −1.304** 0.396** (0.814) (0.193) (0.597) (0.190)  R-squared 0.402 0.384 0.484 0.440  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Panel B: direct gestation HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Gestation drought −0.120 0.146 −1.104*** 0.220** (0.426) (0.114) (0.360) (0.088)  Gestation flood −0.150 0.056 −0.041 0.091* (0.258) (0.071) (0.192) (0.051)  Constant −0.565 0.310 −1.279** 0.442** (0.813) (0.193) (0.600) (0.189)  R-squared 0.402 0.387 0.482 0.442  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Observations 703 703 825 825 . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Seasonal drought 0.317 0.068 −1.244*** 0.245*** (0.414) (0.115) (0.361) (0.094)  Constant −0.550 0.305 −1.304** 0.396** (0.814) (0.193) (0.597) (0.190)  R-squared 0.402 0.384 0.484 0.440  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Panel B: direct gestation HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Gestation drought −0.120 0.146 −1.104*** 0.220** (0.426) (0.114) (0.360) (0.088)  Gestation flood −0.150 0.056 −0.041 0.091* (0.258) (0.071) (0.192) (0.051)  Constant −0.565 0.310 −1.279** 0.442** (0.813) (0.193) (0.600) (0.189)  R-squared 0.402 0.387 0.482 0.442  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Observations 703 703 825 825 Notes: Table 6 reports heterogeneous impacts of estimated impacts of rainfall shocks on birth-to-HAZ and HAZ for boys and girls. Panel A presents results for seasonal drought, while Panel B presents estimated results for gestation drought and flood. These results refer to the baseline results in Columns (2) and (4) for Panels B and C of Tables 2 and 3. All regressions include all controls, locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. See Tables 2 and 3 for additional notes. Robust standard errors clustered at the locality level are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab Table 6 Gendered impact of weather shocks on anthropometric health outcomes . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Seasonal drought 0.317 0.068 −1.244*** 0.245*** (0.414) (0.115) (0.361) (0.094)  Constant −0.550 0.305 −1.304** 0.396** (0.814) (0.193) (0.597) (0.190)  R-squared 0.402 0.384 0.484 0.440  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Panel B: direct gestation HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Gestation drought −0.120 0.146 −1.104*** 0.220** (0.426) (0.114) (0.360) (0.088)  Gestation flood −0.150 0.056 −0.041 0.091* (0.258) (0.071) (0.192) (0.051)  Constant −0.565 0.310 −1.279** 0.442** (0.813) (0.193) (0.600) (0.189)  R-squared 0.402 0.387 0.482 0.442  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Observations 703 703 825 825 . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Seasonal drought 0.317 0.068 −1.244*** 0.245*** (0.414) (0.115) (0.361) (0.094)  Constant −0.550 0.305 −1.304** 0.396** (0.814) (0.193) (0.597) (0.190)  R-squared 0.402 0.384 0.484 0.440  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Panel B: direct gestation HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Gestation drought −0.120 0.146 −1.104*** 0.220** (0.426) (0.114) (0.360) (0.088)  Gestation flood −0.150 0.056 −0.041 0.091* (0.258) (0.071) (0.192) (0.051)  Constant −0.565 0.310 −1.279** 0.442** (0.813) (0.193) (0.600) (0.189)  R-squared 0.402 0.387 0.482 0.442  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Observations 703 703 825 825 Notes: Table 6 reports heterogeneous impacts of estimated impacts of rainfall shocks on birth-to-HAZ and HAZ for boys and girls. Panel A presents results for seasonal drought, while Panel B presents estimated results for gestation drought and flood. These results refer to the baseline results in Columns (2) and (4) for Panels B and C of Tables 2 and 3. All regressions include all controls, locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. See Tables 2 and 3 for additional notes. Robust standard errors clustered at the locality level are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab The pattern shows that the main results are disproportionately driven by the impacts of droughts on girls. This difference by gender for birth to anthropometric health is interesting evidence on the impact of early life shocks on health outcomes. The reversal of shock impacts between birth and short-term periods is quite intriguing. One explanation is that this may be related to potential behavioural responses in allocation of household resources, which coincidentally reduces the impact of shocks on boys in the short-term. We investigate this further by estimating the impact of most recent agricultural shock on food intake across boys and girls. Results in Table A3 demonstrate that rainfall deviation generally increases nutrient intake using food consumption score. More importantly, positive shock increases the nutrient mix for boys in a manner that is not evident for girls while negative shock shows a more prominent adverse effect for girls. Note that estimated impacts for girls are not statistically significant. While similar results have been presented in the literature on aggregate diet composition (Aguilar and Vicarelli, 2022), our results taken together indicate potential discriminatory behaviour against girls. These behavioural responses may be exacerbated by the information passed to parents to reverse boys' stunting classifications due to exposure to gestational shock. This outcome thrives amidst a general disincentive to invest in girl child human capital outcomes in Africa (Iddrisu et al., 2018). 5.3. Endogenous and selection issues One important limitation of the empirical analysis of gestational exposure to weather shocks and birth outcomes is the confounding nature of weather events and the household decision making process. We check for sample selection into childbearing as fertility decisions are endogenous and may be affected by weather conditions. To investigate this, we estimate the impact of recent agricultural shocks on mothers' use of contraceptives. Results show that there is no impact of recent seasonal harvest shocks on the adoption of contraception devices (Table A5). Similarly, we address potential selection bias in estimated coefficients with successful births using data on miscarried pregnancies from the DHS surveys used in this study. The data provides information on women with miscarried pregnancies and provides details such as the month and year of the event. Similar to the main results, we focus on the rural households surveyed in their permanent place of residence during the survey to tackle issues of adaptation and location sorting. We estimate foetal mortality models by conducting regressions of equations (5) and (6) using alternative measures of miscarriage. This includes counts of miscarriage cases and fraction of miscarried pregnancies during a conception year and across localities. The results are reported in Table A4. Estimated coefficients for all categories of outcome and gestational shocks (from precipitation and temperature) are small and generally imprecise. This result shows that the coefficient estimates reported for the main results are not likely to be biassed upwards by potential selection bias arising from miscarried pregnancies linked to adverse weather events11. 6. Discussion and conclusion In this paper, we match the birth outcomes of children to their anthropometric health (HAZ) in the short-term to examine the impact of early life events on their growth trajectory. While we document persistence in the impact of shocks on HAZ outcomes, we observe only marginal effects on birth outcomes. The pathways for the weather shocks include agricultural seasons' rainfall patterns that impact maternal nutrition and food security, droughts which create water scarcity and the asymmetric impact of negative temperature deviation relative to positive deviation. We complement the later by using trimester level heatwave indicator relating to stressor factor during conception. Heterogeneous results show that the impacts of gestational weather shocks which are stronger for boys immediately after birth have been reversed towards girls who show weaker health at a later stage of early life. Supporting analyses show evidence of discriminatory food allocation against girls which explains the gendered-differential reversal during the transition period. Our findings contribute to the research on the impact of weather shocks on health outcomes for low-income countries by examining the persistence and pathways of these effects. Lack of robust evidence on the transition of health outcomes across different stages of life and from different components of weather shocks hinders effective policy. This study is important for the SSA region due to the likelihood of increasing exposure to unpredictable weather events from climate change. Our findings can help provide structured evidence addressing both deficiencies in the research context and policy guidelines for similar settings. This paper provides an understanding of the importance of development milestones using links between birth and anthropometric health outcomes that have not been explored in the literature. The study context is placed within the critical programming period which captures both seasonal and gestational shocks to provide specific guidance for policy interventions. Our results compare to earlier studies on the impacts of early life events in Africa (Thai and Falaris, 2014; Abiona, 2017). Our main results show an approximately 32% (41%) decline in HAZ, corresponding to a 39% (43%) in likelihood of stunting, in response to incidence of seasonal (gestation) droughts across children aged 0 to 59 months. These results are comparable to estimated impacts of drought shocks on HAZ for Malawi where early life droughts lead to a decrease in age-standardised weight z-scores and HAZ of up to 43% and 27%, respectively (Abiona, 2017). Our findings also align with those from other studies in that the impacts of weather events are driven by droughts rather than impacts of floods on early life health. Other settings with similar results for the impacts of weather shocks on child height include Rabassa et al. (2014) for Nigeria, Thai and Falaris (2014) for Vietnam and Groppo and Kraehnert (2016) for Mongolia. This study presents a clear policy framework to consider in designing intervention programs to tackle early life weather shocks. There are indications that extremely low rainfall patterns are more disruptive for early life growth trajectories than excess rainfall patterns. The twist on impacts by gender between birth and anthropometric health outcomes is another dimension to guide policy. While our finding on the reversal of the impacts of shocks from boys to girls in the short-term is not new in the literature, food consumption patterns in favour of boys in periods of abundance consolidate the evidence in this area. Our findings provide additional evidence from the health transition dimension for more equitable household resource allocation decisions. In the absence of more robust evidence it is unclear if the disproportionate allocation of resources is due to deliberate efforts to sideline girls (Sahn and Stifel, 2002; Maitra and Rammohan, 2011). This study shows that timing of shocks relative to the period of conception, gestation, birth or breastfeeding is also an important factor for consideration of policy options. More broadly, the scope of policy aiming to address impacts of environmental shocks may be more targeted with an understanding of the residual impacts of the nature of the shocks especially as it relates to gender balance. Ways to mitigate these impacts include making primary healthcare facilities more accessible to rural households, providing extra support to pregnant or breastfeeding women, educating parents that boys and girls have the same nutritional needs regardless of the birth outcomes. Acknowledgement I wish to express my deep appreciation to UTS Business Research Grant Committee for providing financial support to carry out this research. I am also grateful to workshop attendants at the Centre for Health Economics Research and Evaluation (CHERE), University of Technology Sydney, and seminar participants at the 2021 International Health Economics Association (iHEA) Congress for their comments and suggestions. I am indebted to two anonymous referees who reviewed the paper and provided extensive feedback that helped shape and improve the overall quality of the paper. The survey data used in this article can be obtained from the website of DHS using the following links: https://dhsprogram.com/methodology/survey/survey-display-324.cfm and https://dhsprogram.com/methodology/survey/survey-display-450.cfm. The findings and opinions expressed in this paper are exclusively those of the author. The author is also solely responsible for the content and any errors. Supplementary material Supplementary material is available at Journal of African Economies online. Data availability The data underlying this article are available in the article and in its online supplementary material. Footnotes 1 The DHS data collects extensive contemporaneous HAZ measures (for children aged 0–59 months) in reference to the year and month of survey. It also collects historic record of birth weight for each child at birth as reported by their parents. Note that the gap between the period of birth weight record and HAZ measure is determined by the timeline of the age-in-months of each child at the date of the survey. 2 This list focuses on restricted rural pathways based on data availability and does not include exhaustive weather pathways. 3 To achieve this, we match GPS coordinates of each enumeration area to the GPS of the four closest weather stations from UDEL's weather data archive to obtain estimates of rainfall and temperature. This process used inverse weighted distance average of the weather parameters from the four closest weather stations to the GPS of the enumeration area similar to the approach used in other literature that we follow to construct the weather parameters in this study (Rocha and Soares, 2015). This approach may lead to small variation of absolute weather patterns especially for agricultural cycles of neighbouring enumeration areas with limited distribution of weather stations. We partially address this problem by using deviation of weather measures from the historical norm designated as shocks to provide an additional level of variation in this paper (see equations 1–4 below). 4 We depict the trimester-level shocks using the threshold method for depicting warmer temperatures in the literature (Jagnani et al., 2021). This is designated as 1 if |${\mathrm{temp}}_{lym}>25{}^{\circ}\mathrm{C}\ \mathrm{in}\ \mathrm{at}\ \mathrm{least}\ \mathrm{one}\ \mathrm{of}\ \mathrm{the}\ \mathrm{three}\ \mathrm{months}$| of the: first trimester for months |$m=-8\ to-6$|⁠; second trimester for months |$m=-5\ to-3$| and third trimester for months |$m=-2\ to\ 0.$| Note that each trimester is treated separately and each trimester indicator is designated 0 otherwise. 5 We include the agricultural cycle of 2002 to capture shocks for children born in early 2003. Overall, we use large spatial–temporal variation of the seasonal rainfall level when compared to the locality average to identify explanatory variables of interest. A similar pattern exists for the gestation period shocks in equations (2) and (3). 6 We refer to the fully-specified models—columns (2) and (4)—as the baseline results. 7 Nevertheless, the pattern of the coefficient estimates for birth weight and low birth weight are largely consistent with a priori theoretical expectation as seasonal drought is negatively (positively) correlated with birth weight (low birth weight indicator). 8 The out of season period for Sierra Leone (November, December, January, February and March) comprises the dry season months when extensive crop harvesting of the previous planting season takes place. There is minimal crop cultivation during this period compared to the rainy season. 9 A formal t-test for equality of coefficient estimates across boys and girls indicates differential estimates for all cases except one. 10 We also use gender interaction term in an alternative specification for sensitivity analysis of our gendered-differential results (results available in the supplementary file published with this paper). Overall, the results reinforce the main findings that the gender difference in impacts reverses from infancy to young children. 11 The infant mortality literature relates to the selection effect discussed in Section 5.3. Findings from the existing literature suggest that our estimates are likely to be conservative estimates (i.e. underestimate the true effects of weather shocks on child nutrition). References Abhishek C. ( 2010 ) ‘ Supply Shocks and Gender Bias in Child Health Investments: Evidence From the ICDS Programme in India ’, The B.E. Journal of Economic Analysis & Policy , 10 ( 1 ): 1 – 26 . Google Scholar OpenURL Placeholder Text WorldCat Abiona O. ( 2017 ) ‘ Adverse Effects of Early Life Extreme Precipitation Shocks on Short-Term Health and Adulthood Welfare Outcomes ’, Review of Development Economics , 21 ( 4 ): 1229 – 54 . Google Scholar Crossref Search ADS WorldCat Abiona O. ( 2022 ) ‘ Malnutrition Pathway for the Impact of in utero Drought Shock on Child Growth Indicators in Rural Households ’, Environment and Development Economics , 27 ( 1 ): 20 – 39 . Google Scholar Crossref Search ADS WorldCat Abiona O. , Ajefu J. ( 2022 ) ‘ The Impact of Timing of in-utero Drought Shocks on Birth Outcomes in Rural Households: Evidence from Sierra Leone ’, Journal of Population Economics . Forthcoming . https://doi.org/10.1007/s00148-022-00926-w. Google Scholar OpenURL Placeholder Text WorldCat Aguilar A. , Vicarelli M. ( 2022 ) ‘ El Niño and Children: Medium-Term Effects of Early-Life Weather Shocks on Cognitive and Health Outcomes ’, World Development , 150 ( C ): 105690. https://doi.org/10.1016/j.worlddev.2021.105690. Google Scholar OpenURL Placeholder Text WorldCat Almond D. , Currie J. ( 2011 ) ‘ Killing Me Softly: The Fetal Origins Hypothesis ’, Journal of Economic Perspectives , 25 ( 3 ): 153 – 72 . Google Scholar Crossref Search ADS PubMed WorldCat Almond D. , Currie J., Herrmann M. ( 2012 ) ‘ From Infant to Mother: Early Disease Environment and Future Maternal Health ’, Labour Economics , 19 ( 4 ): 475 – 83 . Google Scholar Crossref Search ADS WorldCat Andalón M. , Azevedo J. P., Rodríguez-Castelán C., Sanfelice V., Valderrama-González D. ( 2016 ) ‘ Weather Shocks and Health at Birth in Colombia ’, World Development , 82 : 69 – 82 . Google Scholar Crossref Search ADS WorldCat Baguet M. , Dumas C. ( 2019 ) ‘ How Does Birth Weight Affect Health and Human Capital? A Short- and Long-Term Evaluation ’, Health Economics , 28 ( 5 ): 597 – 617 . Google Scholar Crossref Search ADS PubMed WorldCat Baker R. E. , Anttila-Hughes J. ( 2020 ) ‘ Characterizing the Contribution of High Temperatures to Child Undernourishment in Sub-Saharan Africa ’, Scientific Reports , 10 : 18796 . Google Scholar Crossref Search ADS PubMed WorldCat Balietti A. , Datta S., Veljanoska S. ( 2022 ) ‘ Air Pollution and Child Development in India ’, Journal of Environmental Economics and Management , 113 : 102624. https://doi.org/10.1016/j.jeem.2022.102624. Google Scholar OpenURL Placeholder Text WorldCat Bassino J.-B. , Lagoarde-Segot T., Woitek U. ( 2022 ) ‘ Prenatal Climate Shocks and Adult Height in Developing Countries. Evidence from Japan (1872–1917) ’, Economics & Human Biology , 45 ( 101115 ): 101115. https://doi.org/10.1016/j.ehb.2022.101115. Google Scholar OpenURL Placeholder Text WorldCat Behrman J. R. , Rosenzweig M. R. ( 2004 ) ‘ Returns to Birthweight ’, The Review of Economics and Statistics , 86 ( 2 ): 586 – 601 . Google Scholar Crossref Search ADS WorldCat Black S. E. , Devereux P. J., Salvanes K. G. ( 2007 ) ‘ From the Cradle to the Labor Market? The Effect of Birthweight on Adult Outcomes ’, Quarterly Journal of Economics , 122 ( 1 ): 409 – 39 . Google Scholar Crossref Search ADS WorldCat Block S. , Haile B., You L., Headey D. ( 2022 ) ‘ Heat Shocks, Maize Yields, and Child Height in Tanzania ’, Food Security , 14 : 93 – 109 . Google Scholar Crossref Search ADS WorldCat Brainerd E. , Menon N. ( 2014 ) ‘ Seasonal Effects of Water Quality: The Hidden Costs of the Green Revolution to Infant and Child Health in India ’, Journal of Development Economics , 107 ( C ): 49 – 64 . Google Scholar OpenURL Placeholder Text WorldCat Burlando A. ( 2014 ) ‘ Transitory Shocks and Birth Weights: Evidence from a Blackout in Zanzibar ’, Journal of Development Economics , 108 ( C ): 154 – 68 . Google Scholar OpenURL Placeholder Text WorldCat Carrillo B. ( 2020 ) ‘ Early Rainfall Shocks and Later-Life Outcomes: Evidence from Colombia ’, The World Bank Economic Review , 34 ( 1 ): 179 – 209 . Google Scholar Crossref Search ADS WorldCat Case A. , Angela F., Christina P. ( 2005 ) ‘ The Lasting Impact of Childhood Health and Circumstance ’, Journal of Health Economics , 24 ( 2 ): 365 – 89 . Google Scholar Crossref Search ADS PubMed WorldCat Case A. , Paxson C. ( 2008 ) ‘ Stature and Status: Height, Ability, and Labor Market Outcomes ’, Journal of Political Economy , 116 ( 3 ): 499 – 532 . Google Scholar Crossref Search ADS PubMed WorldCat Chen X. , Tan C. M., Zhang X., Zhang X. ( 2020 ) ‘ The Effects of Prenatal Exposure to Temperature Extremes on Birth Outcomes: The Case of China ’, Journal of Population Economics , 33 ( 4 ): 1263 – 302 . Google Scholar Crossref Search ADS PubMed WorldCat Comfort A. B. ( 2016 ) ‘ Long-Term Effect of in utero Conditions on Maternal Survival Later in Life: Evidence from Sub-Saharan Africa ’, Journal of Population Economics , 29 ( 2 ): 493 – 527 . Google Scholar Crossref Search ADS WorldCat Conti G. , Hanson M., Inskip H., Crozier S., Cooper C., Godfrey K. ( 2020 ) Beyond Birthweight: The Origins of Human Capital, IZA Working Paper 13296. Institute of Labor Economics. Cornwell K. , Inder B. ( 2015 ) ‘ Child Health and Rainfall in Early Life ’, Journal of Development Studies , 51 ( 7 ): 865 – 80 . Google Scholar Crossref Search ADS WorldCat Daghagh Yazd S. , Wheeler S. A., Zuo A. ( 2020 ) ‘ Understanding the Impacts of Water Scarcity and Socio-Economic Demographics on Farmer Mental Health in the Murray-Darling Basin ’, Ecological Economics , 169 ( C ): 106564. https://doi.org/10.1016/j.ecolecon.2019.106564. Google Scholar OpenURL Placeholder Text WorldCat Desbureaux S. , Rodella A.-S. ( 2019 ) ‘ Drought in the City: The Economic Impact of Water Scarcity in Latin American Metropolitan Areas ’, World Development , 114 ( C ): 13 – 27 . Google Scholar OpenURL Placeholder Text WorldCat Deschênes O. , Greenstone M., Guryan J. ( 2009 ) ‘ Climate Change and Birthweight ’, The American Economic Review , 99 ( 2 ): 211 – 7 . Google Scholar Crossref Search ADS PubMed WorldCat Feeny S. , Mishra A., Trinh T.-A., Ye L., Zhu A. ( 2021 ) ‘ Early-Life Exposure to Rainfall Shocks and Gender Gaps in Employment: Findings from Vietnam ’, Journal of Economic Behavior and Organization , 183 ( C ): 533 – 54 . Google Scholar OpenURL Placeholder Text WorldCat Freudenreich H. , Aladysheva A., Brück T. ( 2022 ) ‘ Weather Shocks Across Seasons and Child Health: Evidence from a Panel Study in the Kyrgyz Republic ’, World Development , 155 ( 105801 ): 105801. https://doi.org/10.1016/j.worlddev.2021.105801. Google Scholar OpenURL Placeholder Text WorldCat Geruso M. , Spears D. ( 2018 ) ‘ Heat, Humidity, and Infant Mortality in the Developing World ’, NBER Working Paper Series. Working Paper 24870 . http://www.nber.org/papers/w24870. Groppo V. , Kraehnert K. ( 2016 ) ‘ Extreme Weather Events and Child Height: Evidence from Mongolia ’, World Development , 86 : 59 – 78 . Google Scholar Crossref Search ADS WorldCat Hirvonen K. , Sohnesen T. P., Bundervoet T. ( 2020 ) ‘ Impact of Ethiopia's 2015 Drought on Child Undernutrition ’, World Development , 131 ( 104964 ): 104964. https://doi.org/10.1016/j.worlddev.2020.104964. Google Scholar OpenURL Placeholder Text WorldCat Hoddinott J. , Kinsey B. ( 2001 ) ‘ Child Growth in the Time of Drought ’, Oxford Bulletin of Economics and Statistics , 63 ( 4 ): 409 – 36 . Google Scholar Crossref Search ADS WorldCat Iddrisu A. M. , Danquah M., Quartey P., Ohemeng W. ( 2018 ) ‘ Gender Bias in Households’ Educational Expenditures: Does the Stage of Schooling Matter? ’ World Development Perspectives , 10–12 : 15 – 23 . Google Scholar OpenURL Placeholder Text WorldCat Jagnani M. , Barrett C. B., Liu Y., You L. ( 2021 ) ‘ Within-Season Producer Response to Warmer Temperatures: Defensive Investments by Kenyan Farmers ’, The Economic Journal , 131 ( 633 ): 392 – 419 . Google Scholar Crossref Search ADS WorldCat Kaul T. ( 2018 ) ‘ Intra-Household Allocation of Educational Expenses: Gender Discrimination and Investing in the Future ’, World Development , 104 ( C ): 336 – 43 . Google Scholar OpenURL Placeholder Text WorldCat Kudamatsu M. , Persson T., Strömberg D. ( 2012 ) ‘ Weather and Infant Mortality in Africa ’, CEPR Discussion Paper , No. DP9222 . https://ssrn.com/abstract=2210191. Google Scholar OpenURL Placeholder Text WorldCat Lee W.-S. , Li B. G. ( 2021 ) ‘ Extreme Weather and Mortality: Evidence from Two Millennia of Chinese Elites ’, Journal of Health Economics . Forthcoming . https://doi.org/10.1016/j.jhealeco.2020.102401. Google Scholar OpenURL Placeholder Text WorldCat Maitra P. , Rammohan A. ( 2011 ) ‘ The Link between Infant Mortality and Child Nutrition in India: Is There any Evidence of a Gender Bias? ’ Journal of the Asia Pacific Economy , 16 ( 1 ): 81 – 110 . Google Scholar Crossref Search ADS WorldCat Majid M. F. ( 2015 ) ‘ The Persistent Effects of In Utero Nutrition Shocks over the Life Cycle: Evidence from Ramadan Fasting ’, Journal of Development Economics , 117 : 48 – 57 . Google Scholar Crossref Search ADS WorldCat Matsuura K. , Willmott C. ( 2017 ) ‘Terrestrial air temperature and precipitation: 1900–2017’, in Gridded Monthly Time Series ' Version 5.01 . Center for Climatic Research, University of Delaware , Delaware, United States of America . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Meng X. , Qian N. ( 2009 ) The long-term consequences of famine on survivors: evidence from a unique natural experiment using China's great famine. NBER working paper no. 14917 . Cambridge : National Bureau of Economic Research, Inc . Molina O. , Saldarriaga V. ( 2017 ) ‘ The Perils of Climate Change: In utero Exposure to Temperature Variability and Birth Outcomes in the Andean Region ’, Economics & Human Biology , 24 : 111 – 24 . Google Scholar Crossref Search ADS WorldCat Mulmi P. , Block S. A., Shively G. E., Masters W. A. ( 2016 ) ‘ Climatic Conditions and Child Height: Sex-Specific Vulnerability and the Protective Effects of Sanitation and Food Markets in Nepal ’, Economics and Human Biology , 23 ( C ): 63 – 75 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Ngegba M. P. , Moriba P., Kandeh J. B. A., Moiwo J. P., Massaquoi S. B. ( 2018 ) ‘ Assessing Efficiency of Action Aid Sierra Leone (AASL) Extension Services in Sierra Leone ’, International Journal of Agricultural Extension , 6 ( 2 ): 129 – 38 . Google Scholar Crossref Search ADS WorldCat Rabassa M. , Skoufias E., Jacoby H. ( 2014 ) ‘ Weather and Child Health in Rural Nigeria ’, Journal of African Economies , 23 : 464 – 92 . Google Scholar Crossref Search ADS WorldCat Randell H. , Gray C., Grace K. ( 2020 ) ‘ Stunted from the Start: Early Life Weather Conditions and Child Undernutrition in Ethiopia ’, Social Science & Medicine , 261 ( C ): 113234. https://doi.org/10.1016/j.socscimed.2020.113234. Google Scholar OpenURL Placeholder Text WorldCat Rocha R. , Soares R. R. ( 2015 ) ‘ Water Scarcity and Birth Outcomes in the Brazilian Semiarid ’, Journal of Development Economics , 112 : 72 – 91 . Google Scholar Crossref Search ADS WorldCat Rodríguez L. ( 2016 ) ‘ Intrahousehold Inequalities in Child Rights and Well-Being. A Barrier to Progress? ’ World Development , 83 ( C ): 111 – 34 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Sahn D. E. , Stifel D. C. ( 2002 ) ‘ Parental Preferences for Nutrition of Boys and Girls: Evidence from Africa ’, Journal of Development Studies , 39 ( 1 ): 21 – 45 . Google Scholar Crossref Search ADS WorldCat Schetter C. D. ( 2011 ) ‘ Psychological Science on Pregnancy: Stress Processes, Biopsychosocial Models, and Emerging Research Issues ’, Annual Review of Psychology , 62 : 531 – 58 . Google Scholar Crossref Search ADS PubMed WorldCat Thai T. Q. , Falaris E. M. ( 2014 ) ‘ Child Schooling, Child Health, and Rainfall Shocks: Evidence from Rural Vietnam ’, The Journal of Development Studies , 50 ( 7 ): 1025 – 37 . Google Scholar Crossref Search ADS WorldCat Thiede B. , Strube J. ( 2020 ) ‘ Climate Variability and Child Nutrition: Findings from Sub-Saharan Africa ’, Global Environmental Change , 65 ( C ): 102192. https://doi.org/10.1016/j.gloenvcha.2020.102192. Google Scholar OpenURL Placeholder Text WorldCat Torche F. ( 2018 ) ‘ Prenatal Exposure to an Acute Stressor and children's Cognitive Outcomes ’, Demography , 55 ( 5 ): 1611 – s39 . Google Scholar Crossref Search ADS PubMed WorldCat © The Author(s) 2023. Published by Oxford University Press on behalf of the Centre for the Study of African Economies, all rights reserved. For Permissions, please email: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of African Economies Oxford University Press

Weather Shocks, Birth and Early Life Health: Evidence of Different Gender Impacts

Abiona, Olukorede
Journal of African Economies , Volume Advance Article – Jan 16, 2023
Download PDF
21 pages

Loading next page...
 
/lp/oxford-university-press/weather-shocks-birth-and-early-life-health-evidence-of-different-afl0afmAa9
Publisher
Oxford University Press
Copyright
Copyright © 2023 Centre for the Study of African Economies
ISSN
0963-8024
eISSN
1464-3723
DOI
10.1093/jae/ejac031
Publisher site
See Article on Publisher Site

Abstract

Abstract This paper examines the impact of exposure to weather events during gestation on birth weight and anthropometric health of a cohort of children. We explore birth records for the cohort of children born between 2003 and 2013 in Sierra Leone using Demographic Health Surveys linked to temporal variation of rainfall and temperature patterns. We find that in utero droughts (or abnormally low precipitation levels) increase the prevalence of low birth weight with larger effects among boys. However, the effects of those same in utero shocks on the prevalence of stunting up to 59 months later are smaller for boys than for girls. The gender difference in estimated impacts from birth to anthropometric health is attributed to food consumption patterns that favour boys. Our results have policy implications for tracking health outcomes during early childhood using birth and anthropometric health, especially by gender. 1. Introduction Birth weight underpins short- to long-term socioeconomic outcomes (Behrman and Rosenzweig, 2004; Black et al., 2007; Almond and Currie, 2011; Baguet and Dumas, 2019; Conti et al., 2020; Bassino et al., 2022). This is also linked to childhood anthropometric health, which helps to convey health trajectory throughout the lifecycle of an individual (Hoddinott and Kinsey, 2001; Case et al., 2005; Case and Paxson, 2008; Almond et al., 2012). The overarching research question addressed in this paper is to investigate if the effects of shock exposure at the in utero period impacts birth outcomes, and if those effects persist into early childhood. As a secondary question, we also examine the dynamic effects of gestational shocks for boys and girls to check if these are different. While extensive body of literature provides evidence on the adverse effects of in utero weather shocks on anthropometric health, there are emerging studies that demonstrate the adverse effects of same on birth outcomes (Deschênes et al., 2009; Molina and Saldarriaga, 2017; Chen et al., 2020; Abiona and Ajefu, 2022). Up to date, little is known about the role of gestation shocks in the progression of child development from birth to short-term outcomes in a way that could inform policy. This gap implies that policymakers lack adequate information to design intervention programs for early life shock exposure. This study seeks to bridge this gap in knowledge by extending the body of evidence in this area of research. To address the main objective of this paper, we reconnect important childhood health measures, using birth and anthropometric outcomes. This way, we are able to examine the transition impacts of early life shocks from alternative vital statistics. The study uses a cohort of children while focusing on the differential enabling pathways of weather patterns to extend frontiers of evidence regarding the literature on the impacts of extreme weather events on child health in Sub-Saharan Africa (SSA). To achieve this, we match vital health statistics (birth weight and anthropometric health) from the Demographic Health Surveys (DHS) to shocks constructed from the University of Delaware weather data archive in order to investigate the impact of weather shocks on early life health outcomes for rural households in Sierra Leone. The mapping process of height-for-age z-scores (HAZ) of children aged 0–59 months to their birth weight constitutes the novel approach adopted in this research1. We hypothesise diverse weather pathways to evaluate the role of weather shocks on the outcomes. This includes the impact of weather patterns through nutrition, disease environment, water scarcity and other stressor factors. Our analysis shows weak impacts of low precipitation level for seasonal and gestation period variations on birth outcomes but we strongly reject the null hypothesis for anthropometric outcomes. Our findings also show that the impact of weather shocks have alternative pathways. For example, the impacts of shock on birth outcomes are partially documented for the low birth weight indicator with a positive association from seasonal drought. Our findings also show that an incidence of seasonal drought decreases HAZ by approximately 32% while increasing the probability of stunting by 39%. We also estimate results for both extremely low and high precipitation level pathways during gestation, respectively. Our results show that both shocks have an impact on HAZ. Exposure to extremely low precipitation level is associated with an average decrease in HAZ by 41%, while extremely low and high precipitation levels is associated with an increase in the probability of stunting by 43% and 24%, respectively. Heterogeneous impacts show an asymmetric impact of gestational weather shock on birth and anthropometric outcomes by gender. We find that in utero droughts (or abnormally low precipitation levels) increase the prevalence of low birth weight with larger effects among boys. However, the effects of those same in utero shocks on the prevalence of stunting up to 59 months later are smaller for boys than for girls. Similarly, we report impacts of temperature deviation but disproportionately larger impacts of lower than average temperature deviation relative to the higher than average component for HAZ. Results from the trimester-level heatwave model (using threshold temperature shock indicators across three trimesters) present persistent average impact of trimester-level stressor factor from birth following through to HAZ. Our findings fill an important gap on the impacts of exposure to weather events during gestation on the trajectory of health outcomes. This paper adds to the literature on the impacts of drought on the nutritional status of children in low-income countries (Hirvonen et al., 2020) and indirectly to the origin of future disproportionate impacts of early life shocks on socioeconomic outcomes by gender (Feeny et al., 2021). Other research shows that early life events may be embedded within other weak environmental factors such as sanitation. For example, Mulmi et al. (2016) find asymmetric gendered impacts across trimester exposure to vegetation levels. Cornwell and Inder (2015) provide evidence in support of both nutrition and disease environment from rainfall shocks on height-for-age measures. Our conceptual framework builds on these studies to disentangle the different weather pathways further while also exploring impacts by gender. We address two important issues. First, we examine the impact of gestational shocks on the trajectory of health by combining birth and anthropometric health for rural children who have a greater level of exposure to weather shocks and require policy intervention. Second, we examine the impacts by gender to highlight resource allocation behaviours. Household allocation of economic resources affects human capital outcomes in low-income countries but there are conflicting findings in this area of research (Abhishek, 2010; Rodríguez, 2016; Kaul, 2018). Hence, our results provide additional evidence for more efficient household resource allocation decisions. Our results also highlight the unequal growth trajectory by gender to help ensure equitable distribution of household resources. The remainder of this paper is organised as follows. In Section 2, we outline the research context and the conceptual framework of the study. Section 3 discusses the data sources and representation of the weather pathways. Section 4 provides the empirical methods. In Section 5, we present estimated results, and discuss the main findings and conclude in Section 6. 2. Setting and conceptual framework 2.1. Research setting Evidence from African countries shows a need for deeper understanding of the impact of extreme weather anomalies on child health outcomes. This paper extends the literature on the impacts from different components of weather shocks. Our research design bridges methods from different studies on child health outcomes from rainfall shocks (Randell et al., 2020; Thiede and Strube, 2020) and from temperature shocks (Geruso and Spears, 2018; Baker and Anttila-Hughes, 2020; Block et al., 2022). Sierra Leone is a low-income country in SSA and its rural communities are largely dependent on subsistence agriculture. Weather patterns relative to norm for planting and harvest seasons affect agricultural yields, and the nutrition available to rural mothers and their children, both during gestation and after birth. In this study, we use rainfall shocks during the pre-birth agricultural seasons and water density metrics to capture shocks during gestation. The rationale for this approach is that rural households are prone to harvest shocks while the extent of impact is unpredictable for urban households. In general, agricultural yields associated with rainfall patterns are a plausible measure of food security in rural Sierra Leone. This pathway requires a lag, separating planting from the harvesting season, to accurately capture household exposure to food insecurity. The shock covers harvest in rural areas for all births within the same locality over the same period of time. On the other hand, the water density and temperature pathways match climatic conditions to accumulated levels during the specific gestation period of each child. The extreme nature of the climatic conditions from these alternative pathways can be regarded as disease environment for extreme positive rainfall (flood) and water scarcity for extreme negative rainfall (droughts) (Almond et al., 2012; Rocha and Soares, 2015). Similarly, above average temperature levels (known as heatwaves) reflect stressor factors during the gestation period (Schetter, 2011; Torche, 2018). Sierra Leone's wet (planting) season runs from April to October each year (Ngegba et al., 2018) and the major crops include cereals and paddy rice. Other main grains are maize, millet and sorghum with mainly subsistence agricultural practices in rural areas. These crops require consistent rainfall hence the need to cultivate them during the wet season. Crop harvests across Sierra Leone commence in October each year following the planting season from April. Rural households' lean season starts just before the harvests when households may have depleted food preserved from the previous cycle. Agricultural extension programs are provided for farmers in this period to support sustainable agricultural practice (Ngegba et al., 2018). We construct drought shocks for the wet season using precipitation levels in the agricultural cycles for rural Sierra Leone. 2.2. Background theory and relevant literature Foetal origin hypothesis underpins the health impacts of exposure to shocks during gestation and also recognises household resource variability among other factors. Many studies in economics have explored this hypothesis empirically (Deschênes et al., 2009; Almond and Currie, 2011) to provide a framework to understand factors affecting important development milestones from childhood to adulthood. To tackle endogeneity concerns, many studies explore natural events providing exogenous variation to estimate causal impacts. Developing effective policy requires an investigation of comprehensive gestation shocks (Deschênes et al., 2009; Andalón et al., 2016) and early life events (Comfort, 2016; Lee and Li, 2021; Balietti et al., 2022; Freudenreich et al., 2022). These are not mutually exclusive events but most studies include each separately for specific policy guidelines. The main objective of this study is to understand if impacts from each of these weather shocks persist through early childhood. Evaluating the impact of these simultaneous shock components on birth and anthropometric health outcomes helps to understand the potential disproportionate impacts of these foetal shocks on children's development pathways. A large body of literature has investigated the impacts of early life weather events focusing on diverse pathways. Transmission of weather events includes malnutrition (Meng and Qian, 2009; Majid, 2015; Block et al., 2022) and disease environment and stress (Deschênes et al., 2009; Andalón et al., 2016; Lee and Li, 2021). Kudamatsu et al. (2012) provide additional context for how extreme weather patterns relate to disease environment and undernutrition and how this increases the infant mortality rate in Africa. Many studies focus on nutrition as the main channel of transmission between weather patterns and birth outcomes or anthropometric health. This is also the focus of the foetal origin hypothesis presented by Almond and Currie (2011). Transient weather shocks affecting seasonal cycles during gestation may distort foetal growth thereby triggering intergenerational effects through their effects on early life health outcomes. The impact of food security on health outcomes is likely to be greater in rural areas where households predominantly depend on rain-fed agricultural practice. While this is quite important, extreme weather events may also be triggered by other supplementary mechanisms beyond scarce economic resources in low-income countries. For example, a disease environment is often triggered during floods, because of unsuitable infrastructure. This can come from dirty water pools that breed insects such as mosquitos during rainy seasons, or a muddy environment that leads to dirty water flows into streams—the main drinking water source for rural households. This shows that both environment and infrastructure play interconnected roles in intergenerational health transmission during gestation. Another mechanism underlying the rainfall pattern during gestation is water access, which may have a direct impact on health outcomes, mostly for rural households that have barriers to potable water during drought seasons (Daghagh Yazd et al., 2020), and in other cases urban water through other mechanisms (Desbureaux and Rodella, 2019). On the other hand, extreme temperatures—heatwave and cold—during gestation can influence early life health outcomes. We broadly categorise weather pathways into nutrition, water density (water scarcity and disease environment) and heatwave. Figure A1 illustrates the selected pathways of weather shocks2. 3. Data 3.1. Birth and anthropometric outcomes We use data from the DHS for Sierra Leone. The Sierra Leone DHS follows the usual DHS two-stage cluster-level sampling procedure for national representative datasets. The first stage involves sampling of clusters within districts, while the second stage includes sampling of the household units within selected clusters. The clusters are designated enumeration areas. The Sierra Leone DHS provides linkage geographic coordinates at the enumeration area level. The DHS reports the geolocation of each enumeration area adjusted by 5 km distance across rural areas to ensure confidentiality of the survey locations and anonymity of the study participants. We use birth weight data linked to the birth records and anthropometric data for children from the Sierra Leone DHS for 2008 and 2013. The birth dates span 11 years from 2003 to 2013. We also use household level demographic variables and those specifically relating to women aged 15–49 years. Birth records include vital statistics such as gender, nature of birth, birth order, information on single or multiple births and other childbirth conditions. In addition to birth weight, we use corresponding HAZ for the sample of children in this study. To strictly maintain a cohort of children in our analysis we drop recent births with no documented record of HAZ and children with HAZ but whose birth weights are missing. We follow the literature to use a gestation period of 38–40 weeks (Rocha and Soares, 2015; Abiona and Ajefu, 2022). The outcome variables are the officially documented birth weight during childbirth (measured in kilogrammes) and the World Health Organisation's standardised HAZ measures from the survey data. We also follow the literature to construct an indicator variable for an occasion of low birth weight (LBW), with 1 for weight less than 2500 g at birth and 0 otherwise. HAZ less than −2 standard deviation units is categorised as stunting in the same manner. 3.2. Weather data To measure weather conditions, we use data from the Center for Climatic Research, University of Delaware. We extract rainfall and temperature datasets from weather stations reporting the terrestrial precipitation and air temperature for 1900–2017 (version 5.01). This archive provides estimates of monthly precipitation and temperature on a 0.5° by 0.5° grid for terrestrial areas across the globe. The rainfall and temperature estimates are based on climatologically aided interpolation of available weather station information made available by Matsuura and Willmott (2017). We use the GPS coordinates provided for each DHS enumeration area across Sierra Leone for the 2008 and 2013 surveys to estimate corresponding weather patterns for baseline weather data required to compute shocks3. To address concerns on selective migration or location sorting across enumeration areas, we restrict our analysis to women surveyed in their permanent place of residence only. Figure 1 presents the distribution of survey clusters in the two survey waves. The administrative borders reflect the 153 chiefdoms across Sierra Leone while the dots represent the sampled clusters—enumeration areas—within each chiefdom for the two waves. Figure 1 Open in new tabDownload slide Distribution of Survey Clusters in Sierra Leone DHS Data 3.3. Shock pathways 3.3.1. Agricultural cycles For empirical analysis, we compute the seasonal mean and standard deviation movements for the historical rainfall period within the same enumeration area and define seasonal droughts using standard deviation movements around the long-term average. These shocks cover all enumeration areas (henceforth designated as localities in this paper) across Sierra Leone between 2002 and 2013 using the methodology in equation (1): $$ \begin{align} {\mathrm{Seasonal}\ \mathrm{drought}}_{ly-1}=1\ if\ {\mathrm{Rainfall}}_{ly-1}<\left(\overline{{\mathrm{Rainfall}}_l}-\left[2\ast \left({\mathrm{Rainfall}}_{SD;l}\right)\right]\ \right),\mathrm{and}\ \mathrm{zero}\ \mathrm{otherwise}, \end{align}$$(1) Where |${\mathrm{Rainfall}}_{ly-1}$| indicates the precipitation level for the reference agricultural season relating to the year of birth of each child within a locality l, and |$\overline{{\mathrm{Rainfall}}_l}$| (⁠|${\mathrm{Rainfall}}_{SD;l}$|⁠) is the average (standard deviation) historical precipitation for a locality in the previous planting seasons covering 30 years. Thus, |${\mathrm{Seasonal}\ \mathrm{drought}}_{ly-1}$| represents an agricultural shock indicator. Figure 2 shows how births across months of the year are calibrated with agricultural cycles for our analysis. Figure 2 Open in new tabDownload slide Matching Agricultural Cycle Across Months of Childbirth—2008 3.3.2. Water density: disease environment and water scarcity The disease environment and water scarcity pathways more directly capture an individual child's exposure to weather events during gestation. We classify gestation exposure to the disease environment and water scarcity as rainfall levels related directly to drought and flood events, and not agricultural practice. The drought measure here is completely different from the drought approach in equation (1) due to the different periodic aggregation of the rainfall pattern but with the same threshold of 2 standard deviation movements. Both the drought and flood shocks capture variation in 9-month gestational period rainfall as follows: $$ \begin{equation} {\textrm{Gestation}\ \textrm{drought}}_{lym}=1\ if\ \sum_{m=-8}^0{\mathrm{R}}_{lm}<\left(\overline{{\mathrm{R}}_l}-\left[2\ast \left({\mathrm{R}}_{SD;l}\right)\right]\right), \ \textrm{and}\ \textrm{zero}\ \textrm{otherwise}, \end{equation}$$(2) $$ \begin{equation} \textrm{Gestation}\ {\textrm{flood}}_{lym}=1\ if\ \sum_{m=-8}^0{\mathrm{R}}_{lm}>\left(\overline{{\mathrm{R}}_l}>{\mathrm{R}}_{SD;l}\right),\ \textrm{and}\ \textrm{zero}\ \textrm{otherwise}, \end{equation}$$(3) where |${\mathrm{R}}_{lm}$| indicates the monthly rainfall during the gestation period specific to each child. We compute accumulated rainfall patterns for each child within locality l from the month of conception (m = −8) to delivery (m = 0). |$\overline{{\mathrm{R}}_l}$| is the average historical rainfall level for child-specific gestational period for each locality in the past 30 years covering a similar period. |${\mathrm{R}}_{SD;l}$| represents the associated standard deviation which depicts the volatility of rainfall levels. We follow Rocha and Soares (2015) by using progressive monthly accumulation of low rainfall levels used to capture water scarcity for Brazilian municipalities while extending this to flood. This is similar to the approach used in Carrillo (2020). Note that weather data distribution inconsistencies explain the asymmetric drought and flood composition in our analysis. For seasonal rainfall, the flood shock thresholds (above 1 or 2 standard deviation movements from the historical average) are unavailable across localities. This implies that we are unable to include an indicator for flood in our seasonal shock regressions. On the other hand, the distribution of the gestation shock thresholds differs for the drought and flood compositions where drought exists for both 1 and 2 standard deviation movements below; but we have up to 1 standard deviation above the norm across localities for the flood component. We chose the extreme drought shock while complementing that with the available gestation flood shock for our empirical context. This particularly enables us to cover shocks for the diverse nature of crops planted by smallholder farmers in rural Sierra Leone. 3.3.3. Temperature shocks We now examine the role of weather-related stress on health outcomes by focusing on gestational temperature variation. We follow the standard literature (Molina and Saldarriaga, 2017) by defining temperature variability as the deviation in the air temperature from each locality's historical mean as follows: $$ \begin{equation} {\mathrm{Temperature}\ \mathrm{shock}}_{lym}=\kern0.5em \left[\frac{1}{9}\ \sum_{m=-8}^0\left({\mathrm{temp}}_{lym}-\overline{{\mathrm{temp}}_l}\right)\ \right]/{SD}_l, \end{equation}$$(4) for a child born in locality |$l$| in year y and month m, where m = {−8; −7;...; 0}. The variable |${\mathrm{temp}}_{lym}$| is the monthly air temperature in the corresponding locality for the |$m$|th month preceding a child's month of birth, |$\overline{{\mathrm{temp}}_l}$| is the local historical moving temperature average for similar periodic progression in the past 30 years, and |${SD}_l$|is the standard deviation of the locality's observed temperature over the same period. While |${\mathrm{Temperature}\ \mathrm{shock}}_{lym}$| captures the temperature exposure for each child during gestation, we are interested in the propensity of extreme temperature measures—both cold spells and heatwaves. We capture this in equation (4) by separating the shock variable into absolute negative and positive deviations, respectively, around the historical norm. In the results section, we also complement negative deviation estimations with trimester-level heatwave exposure indicators4. This depicts the propensity of the exposure to heatwaves at the embryonic, foetal and perinatal stages of gestation indicating the first, second and third trimesters, respectively. Summary statistics of the outcome variables and non-standardised weather parameters are reported in Table 1. The number of observations for 2008 and 2013 varies significantly, reflecting a baseline distribution of samples over the two surveys. Average children's age is 24 months in 2008 and 24.6 months in 2013, and 47.6% and 45.6% of the children are males in 2008 and 2013, respectively. Average birth weight is approximately 3.3 kg across surveys but the incidence of low birth weight is 18.7% for 2008 but just 10% in 2013. This pattern is contrary to a relatively smaller difference in stunting rates of 38.6% in 2008 and 41.2% in 2013. Also, the gendered distribution of birth weight and HAZ in Figure 3 shows more even distribution of these variables for girls. About 12.3% (42.3) of boys and 15.1% (35.7) of girls have low birth weight (stunting). The patterns are consistent with the theoretical proposition of stronger girl foetuses relative to boys due to considerably lower deterioration in anthropometric health. Table 1 Summary Statistics . Survey year . . 2008 . 2013 . Variables Mean Std dev Mean Std dev  No. of observations 422 NA 1,335 NA  Age of child (months) 24.009 15.815 24.589 16.407  Gender of child (male) 0.476 0.500 0.456 0.498 Health outcomes, transition variables  Birth weight (kg) 3.314 1.049 3.258 0.664  Low birth weight (indicator) 0.187 0.391 0.104 0.306  HAZ (standard deviation) −1.433 2.086 −1.505 1.913  Stunting (indicator) 0.386 0.487 0.412 0.492 Weather variables  Total yearly rainfall measure (millimetre) 1603 504 2301 435  Rainfall standard deviation (millimetre) 390 87 438 57  Total yearly temperature (⁠|${}^{\circ}\mathrm{C}$|⁠) 186 3 186 4 . Survey year . . 2008 . 2013 . Variables Mean Std dev Mean Std dev  No. of observations 422 NA 1,335 NA  Age of child (months) 24.009 15.815 24.589 16.407  Gender of child (male) 0.476 0.500 0.456 0.498 Health outcomes, transition variables  Birth weight (kg) 3.314 1.049 3.258 0.664  Low birth weight (indicator) 0.187 0.391 0.104 0.306  HAZ (standard deviation) −1.433 2.086 −1.505 1.913  Stunting (indicator) 0.386 0.487 0.412 0.492 Weather variables  Total yearly rainfall measure (millimetre) 1603 504 2301 435  Rainfall standard deviation (millimetre) 390 87 438 57  Total yearly temperature (⁠|${}^{\circ}\mathrm{C}$|⁠) 186 3 186 4 Notes: Summary statistics are reported for 1,757 observations for childbirth records spanning 2003–2008 from the 2008 DHS survey and 2008–2013 from the 2013 DHS survey. Sample includes rural localities only. Low birth weight indicator is designated 1 if birth weight measure is less than 2.5 kg—and 0 otherwise; stunting is regarded as 1 if HAZ is below −2 standard deviations—and 0 otherwise. The sample of observations is captured from women with surviving babies only and restricted to a cohort of children with corresponding measures of birth weight and HAZ. Open in new tab Table 1 Summary Statistics . Survey year . . 2008 . 2013 . Variables Mean Std dev Mean Std dev  No. of observations 422 NA 1,335 NA  Age of child (months) 24.009 15.815 24.589 16.407  Gender of child (male) 0.476 0.500 0.456 0.498 Health outcomes, transition variables  Birth weight (kg) 3.314 1.049 3.258 0.664  Low birth weight (indicator) 0.187 0.391 0.104 0.306  HAZ (standard deviation) −1.433 2.086 −1.505 1.913  Stunting (indicator) 0.386 0.487 0.412 0.492 Weather variables  Total yearly rainfall measure (millimetre) 1603 504 2301 435  Rainfall standard deviation (millimetre) 390 87 438 57  Total yearly temperature (⁠|${}^{\circ}\mathrm{C}$|⁠) 186 3 186 4 . Survey year . . 2008 . 2013 . Variables Mean Std dev Mean Std dev  No. of observations 422 NA 1,335 NA  Age of child (months) 24.009 15.815 24.589 16.407  Gender of child (male) 0.476 0.500 0.456 0.498 Health outcomes, transition variables  Birth weight (kg) 3.314 1.049 3.258 0.664  Low birth weight (indicator) 0.187 0.391 0.104 0.306  HAZ (standard deviation) −1.433 2.086 −1.505 1.913  Stunting (indicator) 0.386 0.487 0.412 0.492 Weather variables  Total yearly rainfall measure (millimetre) 1603 504 2301 435  Rainfall standard deviation (millimetre) 390 87 438 57  Total yearly temperature (⁠|${}^{\circ}\mathrm{C}$|⁠) 186 3 186 4 Notes: Summary statistics are reported for 1,757 observations for childbirth records spanning 2003–2008 from the 2008 DHS survey and 2008–2013 from the 2013 DHS survey. Sample includes rural localities only. Low birth weight indicator is designated 1 if birth weight measure is less than 2.5 kg—and 0 otherwise; stunting is regarded as 1 if HAZ is below −2 standard deviations—and 0 otherwise. The sample of observations is captured from women with surviving babies only and restricted to a cohort of children with corresponding measures of birth weight and HAZ. Open in new tab Figure 3 Open in new tabDownload slide Density of Birth Weight and Age-Standardised Height Measures 4. Empirical methods We model econometric equations using exogenous variation in weather patterns for our identification strategy. Our main empirical approach focuses on modelling weather across pathways to draw inference on the impacts of gestational shocks on birth and early life health outcomes. The baseline equations for the seasonal drought, water density (disease environment and water scarcity) and temperature pathways are specified as follows: Seasonal low precipitation level model $$ \begin{equation} {\mathrm{Health}\ \mathrm{outcome}}_{ilym}={\alpha}_{lm}+{\varnothing}_{ym}+{\beta}_1{\mathrm{Seasonal}\ \mathrm{drought}}_{ly-1}+{X}_i^{\prime }\ {\theta}_x+{Z}_{ct}^{\prime }\ {\theta}_z+{\varepsilon}_{ilym} \end{equation}$$(5) Gestational low and high precipitation levels model $$ \begin{align} &{\mathrm{Health}\ \mathrm{outcome}}_{ilym}={\alpha}_{lm}+{\varnothing}_{ym}+{\mu}_1{\mathrm{Gestation}\ \mathrm{drought}}_{lym}+{\mu}_2\mathrm{Gestation}\ {\mathrm{flood}}_{lym} +{X}_i^{\prime }\ {\theta}_x+{Z}_{ct}^{\prime }\ {\theta}_z+{\varepsilon}_{ilym} \end{align}$$(6) Temperature variation model $$ \begin{equation} {\mathrm{Health}\ \mathrm{outcome}}_{ilym}={\alpha}_{lm}+{\varnothing}_{ym}+\kern0.5em {\delta}_1{\mathrm{Temperature}\ \mathrm{shock}}_{lym}+{X}_i^{\prime }\ {\theta}_x+{Z}_{ct}^{\prime }\ {\theta}_z+{\varepsilon}_{ilym} \end{equation}$$(7) Where |${\mathrm{Health}\ \mathrm{outcome}}_{ilym}$| represents health variables—namely birth weight and HAZ—for an observation i in localitylfor a child born in year y and month m using the cohort of children identified in the data section. We estimate the same set of models for low birth weight and stunting indicators to capture intergenerational health transition impacts of the shocks at the threshold level. |${\alpha}_{lm}$| is the locality-of-birth by month-of-birth fixed effects and |${\varnothing}_{ym}$| is the year-of-birth by month-of-birth fixed effects. |${\beta}_1$| is the parameter of interest in equation (5). This parameter measures the impact of exposure to locality harvests from low rainfall. Agricultural season droughts in the lead up to a child's birth are an important determinant of crop harvests and the nutritional intake of the mother during the child's gestation. The baseline impacts for the disease environment and water scarcity pathways are derived from parameters |${\mu}_1$| and |${\mu}_2$| in equation (6), while equation (7) examines the impact of temperature variation using parameter |${\delta}_1$|⁠. In the regression analysis we include a vector of individual level controls, denoted as X in equations (5)–(7), for the child, mothers and their partners. We control for the gender of the child to capture differential outcomes for boys and girls while including age-in-months fixed effects to control for non-linear HAZ responses to shocks across age in months (Rabassa et al., 2014). Mothers' demographic characteristics include age during survey, age at first birth, and indicators for education category and marital status. Partners' covariates consist of demographic characteristics including the individual's age group, education and occupation categories. Z is an array of household level controls including household size, age of household head and an indicator variable for the gender of the head of household. In these models, the error term (⁠|${\varepsilon}_{ilym}$|⁠) accounts for the unobserved time-variant locality characteristics. We estimate equations (5)–(7) by including locality-of-birth by month-of-birth fixed effects (also referred to as birth cohort fixed effects in this paper) to capture unobserved heterogeneity across localities. Note that the variables in the error term are orthogonal to a child's exposure to shock during gestation, which leads to unbiassed estimates of rainfall and temperature shocks on the outcome variables. Our main assumption is an identically and independently distributed (iid) error term across localities, which is correlated within each locality due to spatial correlation of rainfall and temperature patterns for households in the same locality. Hence, standard errors are clustered at the locality level. Our empirical strategy primarily focuses on the temporal weather variation that comes from the locality-of-birth by month-of-birth fixed effects. This means that our models compare children born in the same locality and same month but in different years. This approach is better than the spatial identification variation and avoids natural location selection concerns that may arise from the data. In the regression analysis, we estimate each pathway separately. For equation (5), the identification strategy exploits seasonal rainfall variation and locality concentration of births. This specification uses birth cohort fixed effects to identify impacts of weather shocks on health outcomes. Exogenous variation in the agricultural cycle rainfall across localities is important for the causal interpretation of estimates in this model. Our data shows evidence of established widespread locality-level drought distribution for both cross-regional differences and temporal rainfall variation over a 12-year period that is deemed sufficient to identify the exogenous impact of harvests5. For direct gestational shocks encompassing disease environment, water scarcity and temperature, identification is based on rainfall and temperature variation across individual-level gestational period. This methodology provides a more extensive variation in the weather patterns plausible for specific impacts of weather shocks. 5. Results 5.1. Main results 5.1.1. Gestational low precipitation levels Table 2 presents coefficient estimates of impacts of seasonal drought shock on birth weight and anthropometric outcomes from regressions of equation (5). These are treatment effect estimates to test the impact of gestational nutrition on health outcomes. These results capture locality rainfall variations underlying agricultural harvests, and how they affect birth outcomes and anthropometric health in Panels A and B. We mainly provide Table 2 results from alternative models for equation (5) where we exclude the year-of-birth by month-of-birth fixed effects in columns (1) and (3) while reporting fully-specified models in columns (2) and (4). Each regression model includes all controls, month-of-birth fixed effect and locality-of-birth by month-of-birth fixed effects. In Panel A, baseline estimates6 for low birth weight show a positive correlation of 8.3 percentage points with seasonal drought shock. We also estimate that seasonal drought shock incidence is associated with a decline in the HAZ by 0.47 standard deviations (Panel B). This result is reinforced by the estimated increase of stunting by 15.9 percentage points. Estimated coefficients are statistically significant at traditional levels across Panels A and B. Table 2 Impact of Seasonal Drought Shock on Health Outcomes in Rural Sierra Leone . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Seasonal drought −0.086 −0.084 0.070 0.083* (0.098) (0.099) (0.047) (0.046)  Constant 3.112*** 3.111*** 0.141* 0.118 (0.180) (0.178) (0.075) (0.079)  R-squared 0.371 0.378 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Seasonal drought −0.438* −0.473** 0.144** 0.159*** (0.236) (0.240) (0.060) (0.060)  Constant −0.553 −0.535 0.259** 0.204* (0.437) (0.443) (0.117) (0.119)  R-squared 0.357 0.362 0.292 0.302  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Seasonal drought −0.086 −0.084 0.070 0.083* (0.098) (0.099) (0.047) (0.046)  Constant 3.112*** 3.111*** 0.141* 0.118 (0.180) (0.178) (0.075) (0.079)  R-squared 0.371 0.378 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Seasonal drought −0.438* −0.473** 0.144** 0.159*** (0.236) (0.240) (0.060) (0.060)  Constant −0.553 −0.535 0.259** 0.204* (0.437) (0.443) (0.117) (0.119)  R-squared 0.357 0.362 0.292 0.302  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 Notes: Table 2 presents coefficient estimates of regressions of seasonal drought shock on multiple outcomes for rural surviving births in Sierra Leone between 2003 and 2013. Panels A and B present estimated results for child health outcomes: birth weight (kg) and anthropometric health—using age-standardised height z-scores—HAZ (sd). Low birth weight and stunting indicators are used as supplementary outcomes. These measure proportion of child observations below 2.5 kg for birth weight and below −2 standard deviations for HAZ. Explanatory drought variable follows equation (1) for seasonal drought representing 2 standard deviations below the norm. Each column for each panel presents a separate regression. Columns (1) and (3) present estimated coefficients from regressions including all controls and locality-of-birth by month-of-birth fixed effects while columns (2) and (4) additionally include year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. Robust standard errors, clustered at the locality level, are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab Table 2 Impact of Seasonal Drought Shock on Health Outcomes in Rural Sierra Leone . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Seasonal drought −0.086 −0.084 0.070 0.083* (0.098) (0.099) (0.047) (0.046)  Constant 3.112*** 3.111*** 0.141* 0.118 (0.180) (0.178) (0.075) (0.079)  R-squared 0.371 0.378 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Seasonal drought −0.438* −0.473** 0.144** 0.159*** (0.236) (0.240) (0.060) (0.060)  Constant −0.553 −0.535 0.259** 0.204* (0.437) (0.443) (0.117) (0.119)  R-squared 0.357 0.362 0.292 0.302  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Seasonal drought −0.086 −0.084 0.070 0.083* (0.098) (0.099) (0.047) (0.046)  Constant 3.112*** 3.111*** 0.141* 0.118 (0.180) (0.178) (0.075) (0.079)  R-squared 0.371 0.378 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Seasonal drought −0.438* −0.473** 0.144** 0.159*** (0.236) (0.240) (0.060) (0.060)  Constant −0.553 −0.535 0.259** 0.204* (0.437) (0.443) (0.117) (0.119)  R-squared 0.357 0.362 0.292 0.302  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 Notes: Table 2 presents coefficient estimates of regressions of seasonal drought shock on multiple outcomes for rural surviving births in Sierra Leone between 2003 and 2013. Panels A and B present estimated results for child health outcomes: birth weight (kg) and anthropometric health—using age-standardised height z-scores—HAZ (sd). Low birth weight and stunting indicators are used as supplementary outcomes. These measure proportion of child observations below 2.5 kg for birth weight and below −2 standard deviations for HAZ. Explanatory drought variable follows equation (1) for seasonal drought representing 2 standard deviations below the norm. Each column for each panel presents a separate regression. Columns (1) and (3) present estimated coefficients from regressions including all controls and locality-of-birth by month-of-birth fixed effects while columns (2) and (4) additionally include year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. Robust standard errors, clustered at the locality level, are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab The results show marginal statistical precision for estimated results for birth outcomes (precisely low birth weight)7. Lack of statistical power for treatment effects for birth weight may be due to the small sample composition captured due to the data restriction imposed while linking it to HAZ. In particular, the weak birth weight findings may be partially driven by the imprecise birth weight measures with the documented uneven birth weight patterns using our sample (Figure 3). However, these results do not necessarily contradict evidence documenting the impact of nutrition during gestation on foetal development stages in related studies (Burlando, 2014; Andalón et al., 2016). Robust precision and persistence of estimates across anthropometric health outcomes can be attributable to limited implication of the above restriction on the composition of the reference HAZ sample. Using coefficient estimates from the baseline results to evaluate the impacts of these estimates, an incidence of seasonal drought shock is associated with a 67% increase in the likelihood of low birth weight; a 32% decrease in HAZ; and a 39% increase in the probability of stunting when compared to the mean. Our research design is opposed to a relationship between dry season shocks and birth and/or anthropometric health. To test this alternative pathway, we run tests for both birth and anthropometric outcomes using the underlying drought variation in rainfall for consecutive out-of-season months8. In Table A1, we present the coefficients for the fully-specified fixed effect models—which includes locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects—with a full set of controls. All coefficients are quite small and statistically insignificant, supporting our initial proposition that estimated impacts for the anthropometric health outcomes are associated with only agricultural shocks. This result provides additional credibility for nutrition as the underlying channel for the estimated results in Table 2. 5.1.2. Gestational low and high precipitation levels Our theoretical framework recognises the water density pathway by linking this to the gestation period for various dimensions of early life impacts. While other studies have examined these direct exposures relating to environmental factors such as water pollution and vegetation on a range of child health outcomes (Brainerd and Menon, 2014; Mulmi et al., 2016), this paper is the first to investigate weather-related gestation outcomes for the transition of birth to anthropometric health. Potential disaggregated pathways for water in this framework include potable water accessibility, disease environment, and infrastructure. Access to potable water is an important direct pathway of rainfall to birth outcomes, which may persist to short-term. This is particularly important in rural Sierra Leone where households rely on streams, rivers and untreated wells for domestic water. On one hand, we capture access to water directly tied to the gestational months by measuring the foetal exposure to drought during conception. On the other hand, we use a flood indicator to measure the impact of disease environment in the same setting. We combine both water access and disease environment as explanatory variables of interest in the same model using equation (6). We report results of water density pathways in Table 3. While there are no notable results for birth outcomes, both gestation drought and flood shocks demonstrate results consistent with the expected pattern for anthropometric health. The drought shock shows a strong and persistent impact on HAZ, reinforcing the role of water scarcity during gestation. Our results also uphold evidence for the disease environment for stunting. While it is difficult to interpret these results as disproportionate impacts due to the asymmetric composition of the drought and flood shocks from equations (2) and (3), there is suggestive evidence that the resultant drought shock may constitute other stress factors related to water scarcity for rural women in Sierra Leone. In Panel B column (2), the estimated impact of drought shock on HAZ is a reduction of 0.61 standard deviations. Similarly, estimated impacts for stunting in column (4) show an increase of 17.5 and 9.7 percentage points for drought and flood respectively. These results translate to a decrease in HAZ of 41% association with gestational drought—with a corresponding increase in the likelihood of stunting by 43% for gestational drought shocks and 24% for flood shocks. Table 3 Impact of gestation drought and flood shocks on health outcomes in rural Sierra Leone . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Gestation drought −0.070 −0.083 0.037 0.062 (0.104) (0.109) (0.050) (0.049)  Gestation flood 0.064 0.067 0.037* 0.033 (0.047) (0.050) (0.022) (0.022)  Constant 3.100*** 3.113*** 0.132* 0.114 (0.181) (0.178) (0.075) (0.079)  R-squared 0.371 0.379 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Gestation drought −0.598*** −0.606** 0.167*** 0.175*** (0.225) (0.235) (0.057) (0.060)  Gestation flood −0.145 −0.149 0.109*** 0.097*** (0.126) (0.137) (0.033) (0.035)  Constant −0.548 −0.558 0.259** 0.224* (0.436) (0.445) (0.115) (0.120)  R-squared 0.360 0.364 0.298 0.306  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Gestation drought −0.070 −0.083 0.037 0.062 (0.104) (0.109) (0.050) (0.049)  Gestation flood 0.064 0.067 0.037* 0.033 (0.047) (0.050) (0.022) (0.022)  Constant 3.100*** 3.113*** 0.132* 0.114 (0.181) (0.178) (0.075) (0.079)  R-squared 0.371 0.379 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Gestation drought −0.598*** −0.606** 0.167*** 0.175*** (0.225) (0.235) (0.057) (0.060)  Gestation flood −0.145 −0.149 0.109*** 0.097*** (0.126) (0.137) (0.033) (0.035)  Constant −0.548 −0.558 0.259** 0.224* (0.436) (0.445) (0.115) (0.120)  R-squared 0.360 0.364 0.298 0.306  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 Notes: Table 3 presents coefficient estimates of regressions of gestation drought and flood shocks on multiple outcomes for rural surviving births in Sierra Leone between 2003 and 2013. Panels A and B present estimated results for child health outcomes: birth weight (kg) and anthropometric health—using age-standardised height z-scores—HAZ (sd). Low birth weight and stunting indicators are used as supplementary outcomes. These measure proportion of child observations below 2.5 kg for birth weight and below −2 standard deviations for HAZ. Explanatory variables follow equations (2) and (3) for gestation shocks representing below 2 standard deviations of the norm for gestational drought and above 1 standard deviation of the norm for gestational flood, respectively. Each column for each panel presents a separate regression. Columns (1) and (3) present estimated coefficients from regressions including all controls and locality-of-birth by month-of-birth fixed effects while columns (2) and (4) additionally include year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. Robust standard errors, clustered at the locality level, are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab Table 3 Impact of gestation drought and flood shocks on health outcomes in rural Sierra Leone . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Gestation drought −0.070 −0.083 0.037 0.062 (0.104) (0.109) (0.050) (0.049)  Gestation flood 0.064 0.067 0.037* 0.033 (0.047) (0.050) (0.022) (0.022)  Constant 3.100*** 3.113*** 0.132* 0.114 (0.181) (0.178) (0.075) (0.079)  R-squared 0.371 0.379 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Gestation drought −0.598*** −0.606** 0.167*** 0.175*** (0.225) (0.235) (0.057) (0.060)  Gestation flood −0.145 −0.149 0.109*** 0.097*** (0.126) (0.137) (0.033) (0.035)  Constant −0.548 −0.558 0.259** 0.224* (0.436) (0.445) (0.115) (0.120)  R-squared 0.360 0.364 0.298 0.306  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 . Dependent variables . Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator)  Gestation drought −0.070 −0.083 0.037 0.062 (0.104) (0.109) (0.050) (0.049)  Gestation flood 0.064 0.067 0.037* 0.033 (0.047) (0.050) (0.022) (0.022)  Constant 3.100*** 3.113*** 0.132* 0.114 (0.181) (0.178) (0.075) (0.079)  R-squared 0.371 0.379 0.283 0.295  Mean dependent variable (Birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (sd) Stunting (indicator)  Gestation drought −0.598*** −0.606** 0.167*** 0.175*** (0.225) (0.235) (0.057) (0.060)  Gestation flood −0.145 −0.149 0.109*** 0.097*** (0.126) (0.137) (0.033) (0.035)  Constant −0.548 −0.558 0.259** 0.224* (0.436) (0.445) (0.115) (0.120)  R-squared 0.360 0.364 0.298 0.306  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ Observations 1,677 1,677 1,677 1,677 Notes: Table 3 presents coefficient estimates of regressions of gestation drought and flood shocks on multiple outcomes for rural surviving births in Sierra Leone between 2003 and 2013. Panels A and B present estimated results for child health outcomes: birth weight (kg) and anthropometric health—using age-standardised height z-scores—HAZ (sd). Low birth weight and stunting indicators are used as supplementary outcomes. These measure proportion of child observations below 2.5 kg for birth weight and below −2 standard deviations for HAZ. Explanatory variables follow equations (2) and (3) for gestation shocks representing below 2 standard deviations of the norm for gestational drought and above 1 standard deviation of the norm for gestational flood, respectively. Each column for each panel presents a separate regression. Columns (1) and (3) present estimated coefficients from regressions including all controls and locality-of-birth by month-of-birth fixed effects while columns (2) and (4) additionally include year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. Robust standard errors, clustered at the locality level, are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab 5.1.3. Gestational temperature We also estimate the impact of gestation temperature shock using the deviation measure in equation (4). Results of the regression of health outcomes on temperature—equation 6—are reported in Table 4. Similar to the pattern shown for rainfall shocks, the estimated impacts for birth outcomes lack statistical power. However, the estimated patterns are considerably strong and statistically significant for anthropometric outcomes. Coefficient estimates are also consistent with a priori theoretical expectation especially with the distribution of impacts across negative and positive temperature shocks (columns (3) and (4)). Negative (positive) temperature deviation is associated with an average increase (decrease) of 0.80 (0.25) standard deviations in HAZ and an average decrease (increase) in stunting of 32.4 (1.8) percentage points. All coefficient estimates are statistically significant except for estimated positive temperature for stunting. The results generally indicate that there is a positive association between linear temperature deviation and HAZ while the negative deviation increase is disproportionate relative to the observed decline from the positive temperature deviation. This asymmetric pattern from disaggregated temperature could be explained by the lack of extreme positive temperature shocks (heatwaves) observed in the data wherein the negative temperature shocks are also mild. In general, the negative temperature measure reflects favourable gestational temperature weather conditions. We complement our findings with an alternative heatwave indicator for average trimester-level stress factors. Table A2 shows that second trimester heatwave is strongly associated with both birth and anthropometric outcomes as opposed to the divergent impacts in Table 4. Table 4 Impact of gestational temperature shocks on health outcomes in rural Sierra Leone Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator) Birth weight (kg) Low birth weight (indicator)  Linear temperature deviation −0.006 0.016 (0.020) (0.010)  Negative temperature deviation −0.044 −0.024 (0.098) (0.048)  Positive temperature deviation −0.015 0.014 (0.028) (0.014)  Constant 3.113*** 0.134* 3.122*** 0.135* (0.181) (0.076) (0.180) (0.078)  R-squared 0.370 0.283 0.370 0.283 Mean dependent variable (birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (SD) Stunting (indicator) HAZ (SD) Stunting (indicator)  Linear temperature deviation −0.334*** 0.066*** (0.061) (0.017)  Negative temperature deviation 0.801*** −0.324*** (0.288) (0.079)  Positive temperature deviation −0.248*** 0.018 (0.083) (0.022)  Constant −1.191*** 0.403*** −1.281*** 0.453*** (0.434) (0.116) (0.433) (0.114)  R-squared 0.344 0.278 0.345 0.284  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Observations 1,677 1,677 1,677 1,677 Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator) Birth weight (kg) Low birth weight (indicator)  Linear temperature deviation −0.006 0.016 (0.020) (0.010)  Negative temperature deviation −0.044 −0.024 (0.098) (0.048)  Positive temperature deviation −0.015 0.014 (0.028) (0.014)  Constant 3.113*** 0.134* 3.122*** 0.135* (0.181) (0.076) (0.180) (0.078)  R-squared 0.370 0.283 0.370 0.283 Mean dependent variable (birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (SD) Stunting (indicator) HAZ (SD) Stunting (indicator)  Linear temperature deviation −0.334*** 0.066*** (0.061) (0.017)  Negative temperature deviation 0.801*** −0.324*** (0.288) (0.079)  Positive temperature deviation −0.248*** 0.018 (0.083) (0.022)  Constant −1.191*** 0.403*** −1.281*** 0.453*** (0.434) (0.116) (0.433) (0.114)  R-squared 0.344 0.278 0.345 0.284  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Observations 1,677 1,677 1,677 1,677 Notes: Table 4 presents coefficient estimates of regressions of gestation temperature shocks on multiple outcomes for rural surviving births in Sierra Leone between 2003 and 2013. Panels A and B present estimated results for child health outcomes: birth weight (kg) and anthropometric health—using age standardised height z-scores—HAZ (sd). Low birth weight and stunting indicators are used as supplementary outcomes. These measure proportion of child observations below 2.5 kg for birth weight and below −2 standard deviations for HAZ. Explanatory heatwave variable follows equation (4) for gestational temperature reference relative to historical timeline of the gestation period. Each column for each panel presents a separate regression. Columns (1) and (2) present estimated coefficients for linear deviation from historical air temperature shock while columns (3) and (4) disaggregate deviation to negative for absolute values for deviation lower than norm and positive for deviations above norm. All regressions include all controls, locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. Robust standard errors clustered at the locality level are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab Table 4 Impact of gestational temperature shocks on health outcomes in rural Sierra Leone Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator) Birth weight (kg) Low birth weight (indicator)  Linear temperature deviation −0.006 0.016 (0.020) (0.010)  Negative temperature deviation −0.044 −0.024 (0.098) (0.048)  Positive temperature deviation −0.015 0.014 (0.028) (0.014)  Constant 3.113*** 0.134* 3.122*** 0.135* (0.181) (0.076) (0.180) (0.078)  R-squared 0.370 0.283 0.370 0.283 Mean dependent variable (birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (SD) Stunting (indicator) HAZ (SD) Stunting (indicator)  Linear temperature deviation −0.334*** 0.066*** (0.061) (0.017)  Negative temperature deviation 0.801*** −0.324*** (0.288) (0.079)  Positive temperature deviation −0.248*** 0.018 (0.083) (0.022)  Constant −1.191*** 0.403*** −1.281*** 0.453*** (0.434) (0.116) (0.433) (0.114)  R-squared 0.344 0.278 0.345 0.284  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Observations 1,677 1,677 1,677 1,677 Variables . (1) . (2) . (3) . (4) . Panel A: birth outcomes Birth weight (kg) Low birth weight (indicator) Birth weight (kg) Low birth weight (indicator)  Linear temperature deviation −0.006 0.016 (0.020) (0.010)  Negative temperature deviation −0.044 −0.024 (0.098) (0.048)  Positive temperature deviation −0.015 0.014 (0.028) (0.014)  Constant 3.113*** 0.134* 3.122*** 0.135* (0.181) (0.076) (0.180) (0.078)  R-squared 0.370 0.283 0.370 0.283 Mean dependent variable (birth weight) 3.272 0.124 Panel B: anthropometric health HAZ (SD) Stunting (indicator) HAZ (SD) Stunting (indicator)  Linear temperature deviation −0.334*** 0.066*** (0.061) (0.017)  Negative temperature deviation 0.801*** −0.324*** (0.288) (0.079)  Positive temperature deviation −0.248*** 0.018 (0.083) (0.022)  Constant −1.191*** 0.403*** −1.281*** 0.453*** (0.434) (0.116) (0.433) (0.114)  R-squared 0.344 0.278 0.345 0.284  Mean dependent variable (HAZ) −1.488 0.406 Controls ✓ ✓ ✓ ✓ Locality-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Year-of-birth X month-of-birth FE ✓ ✓ ✓ ✓ Observations 1,677 1,677 1,677 1,677 Notes: Table 4 presents coefficient estimates of regressions of gestation temperature shocks on multiple outcomes for rural surviving births in Sierra Leone between 2003 and 2013. Panels A and B present estimated results for child health outcomes: birth weight (kg) and anthropometric health—using age standardised height z-scores—HAZ (sd). Low birth weight and stunting indicators are used as supplementary outcomes. These measure proportion of child observations below 2.5 kg for birth weight and below −2 standard deviations for HAZ. Explanatory heatwave variable follows equation (4) for gestational temperature reference relative to historical timeline of the gestation period. Each column for each panel presents a separate regression. Columns (1) and (2) present estimated coefficients for linear deviation from historical air temperature shock while columns (3) and (4) disaggregate deviation to negative for absolute values for deviation lower than norm and positive for deviations above norm. All regressions include all controls, locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. Robust standard errors clustered at the locality level are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab 5.2. Heterogeneous impacts by gender We estimate heterogeneous impacts of seasonal and gestational rainfall shocks across boys and girls to examine the gender dimension of our treatment effects. Table 5 presents fully specified models for birth outcomes (birth weight and low birth weight) from seasonal shock (Panel A) and gestation shocks (Panel B) for a comparable sample of observations across boys and girls. Table 6 presents similar heterogeneous results for HAZ. In Table 5, we document striking results where previously muted coefficient estimates are statistically significant when we estimate the impacts of shocks on birth weight for boys. Here, the incidence of seasonal drought shock is associated with a decrease in birth weight of 412 g, which corresponds to an increase in the incidence of low birth weight by 12.2 percentage points. Gestation drought shock shows an association of a decline in birth weight of 391 g while the estimated impact of flood shock is an increase of 132 grams—not statistically significant. Coefficient estimates for birth weight of girls are much smaller in magnitude and not precisely estimated (columns (3) and (4)). These pronounced asymmetric impacts for boys are consistent with a priori theoretical evidence of relatively weaker boy foetuses with similar exposure to adverse shocks when compared to girls. Surprisingly, Table 6 reports a concentration of treatment effects for girls when focussing on girls. Here, coefficient estimates for boys are generally smaller compared to estimated effects for girls in a similar context and mostly statistically insignificant9. Results for girls in columns (3) and (4) are also much bigger than the baseline coefficients for the total sample of observations reported in Tables 2 and 3 respectively. The results also reflect asymmetric impacts of gestation drought and flood shocks previously documented in the baseline results more prominently10. Table 5 Gendered impact of weather shocks on birth outcomes . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Seasonal drought −0.412*** 0.122** −0.076 0.129 (0.144) (0.061) (0.143) (0.096)  Constant 3.223*** 0.060 3.358*** 0.052 (0.257) (0.103) (0.218) (0.137)  R-squared 0.562 0.458 0.432 0.390  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Panel B: direct gestation Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Gestation drought −0.391** 0.066 0.082 0.067 (0.167) (0.070) (0.176) (0.102)  Gestation flood 0.132 0.003 0.037 0.026 (0.100) (0.035) (0.054) (0.034)  Constant 3.212*** 0.060 3.371*** 0.051 (0.255) (0.103) (0.218) (0.138)  R-squared 0.563 0.455 0.433 0.388  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Observations 703 703 825 825 . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Seasonal drought −0.412*** 0.122** −0.076 0.129 (0.144) (0.061) (0.143) (0.096)  Constant 3.223*** 0.060 3.358*** 0.052 (0.257) (0.103) (0.218) (0.137)  R-squared 0.562 0.458 0.432 0.390  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Panel B: direct gestation Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Gestation drought −0.391** 0.066 0.082 0.067 (0.167) (0.070) (0.176) (0.102)  Gestation flood 0.132 0.003 0.037 0.026 (0.100) (0.035) (0.054) (0.034)  Constant 3.212*** 0.060 3.371*** 0.051 (0.255) (0.103) (0.218) (0.138)  R-squared 0.563 0.455 0.433 0.388  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Observations 703 703 825 825 Notes: Table 5 reports heterogeneous impacts of estimated impacts of rainfall shocks on birth outcomes for boys and girls. Panel A presents results for seasonal drought, while Panel B presents estimated results for gestation drought and flood. These results refer to the baseline results in Columns (2) and (4) for Panel A of Tables 2 and 3. All regressions include all controls, locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. See Tables 2 and 3 for additional notes. Robust standard errors clustered at the locality level are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab Table 5 Gendered impact of weather shocks on birth outcomes . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Seasonal drought −0.412*** 0.122** −0.076 0.129 (0.144) (0.061) (0.143) (0.096)  Constant 3.223*** 0.060 3.358*** 0.052 (0.257) (0.103) (0.218) (0.137)  R-squared 0.562 0.458 0.432 0.390  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Panel B: direct gestation Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Gestation drought −0.391** 0.066 0.082 0.067 (0.167) (0.070) (0.176) (0.102)  Gestation flood 0.132 0.003 0.037 0.026 (0.100) (0.035) (0.054) (0.034)  Constant 3.212*** 0.060 3.371*** 0.051 (0.255) (0.103) (0.218) (0.138)  R-squared 0.563 0.455 0.433 0.388  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Observations 703 703 825 825 . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Seasonal drought −0.412*** 0.122** −0.076 0.129 (0.144) (0.061) (0.143) (0.096)  Constant 3.223*** 0.060 3.358*** 0.052 (0.257) (0.103) (0.218) (0.137)  R-squared 0.562 0.458 0.432 0.390  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Panel B: direct gestation Birth weight (kg) Underweight (indicator) Birth weight (kg) Underweight (indicator)  Gestation drought −0.391** 0.066 0.082 0.067 (0.167) (0.070) (0.176) (0.102)  Gestation flood 0.132 0.003 0.037 0.026 (0.100) (0.035) (0.054) (0.034)  Constant 3.212*** 0.060 3.371*** 0.051 (0.255) (0.103) (0.218) (0.138)  R-squared 0.563 0.455 0.433 0.388  Mean dependent variable (Birth weight) 3.315 0.119 3.234 0.129 Observations 703 703 825 825 Notes: Table 5 reports heterogeneous impacts of estimated impacts of rainfall shocks on birth outcomes for boys and girls. Panel A presents results for seasonal drought, while Panel B presents estimated results for gestation drought and flood. These results refer to the baseline results in Columns (2) and (4) for Panel A of Tables 2 and 3. All regressions include all controls, locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. See Tables 2 and 3 for additional notes. Robust standard errors clustered at the locality level are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab Table 6 Gendered impact of weather shocks on anthropometric health outcomes . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Seasonal drought 0.317 0.068 −1.244*** 0.245*** (0.414) (0.115) (0.361) (0.094)  Constant −0.550 0.305 −1.304** 0.396** (0.814) (0.193) (0.597) (0.190)  R-squared 0.402 0.384 0.484 0.440  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Panel B: direct gestation HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Gestation drought −0.120 0.146 −1.104*** 0.220** (0.426) (0.114) (0.360) (0.088)  Gestation flood −0.150 0.056 −0.041 0.091* (0.258) (0.071) (0.192) (0.051)  Constant −0.565 0.310 −1.279** 0.442** (0.813) (0.193) (0.600) (0.189)  R-squared 0.402 0.387 0.482 0.442  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Observations 703 703 825 825 . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Seasonal drought 0.317 0.068 −1.244*** 0.245*** (0.414) (0.115) (0.361) (0.094)  Constant −0.550 0.305 −1.304** 0.396** (0.814) (0.193) (0.597) (0.190)  R-squared 0.402 0.384 0.484 0.440  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Panel B: direct gestation HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Gestation drought −0.120 0.146 −1.104*** 0.220** (0.426) (0.114) (0.360) (0.088)  Gestation flood −0.150 0.056 −0.041 0.091* (0.258) (0.071) (0.192) (0.051)  Constant −0.565 0.310 −1.279** 0.442** (0.813) (0.193) (0.600) (0.189)  R-squared 0.402 0.387 0.482 0.442  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Observations 703 703 825 825 Notes: Table 6 reports heterogeneous impacts of estimated impacts of rainfall shocks on birth-to-HAZ and HAZ for boys and girls. Panel A presents results for seasonal drought, while Panel B presents estimated results for gestation drought and flood. These results refer to the baseline results in Columns (2) and (4) for Panels B and C of Tables 2 and 3. All regressions include all controls, locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. See Tables 2 and 3 for additional notes. Robust standard errors clustered at the locality level are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab Table 6 Gendered impact of weather shocks on anthropometric health outcomes . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Seasonal drought 0.317 0.068 −1.244*** 0.245*** (0.414) (0.115) (0.361) (0.094)  Constant −0.550 0.305 −1.304** 0.396** (0.814) (0.193) (0.597) (0.190)  R-squared 0.402 0.384 0.484 0.440  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Panel B: direct gestation HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Gestation drought −0.120 0.146 −1.104*** 0.220** (0.426) (0.114) (0.360) (0.088)  Gestation flood −0.150 0.056 −0.041 0.091* (0.258) (0.071) (0.192) (0.051)  Constant −0.565 0.310 −1.279** 0.442** (0.813) (0.193) (0.600) (0.189)  R-squared 0.402 0.387 0.482 0.442  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Observations 703 703 825 825 . Boys . Girls . Variables . (1) . (2) . (3) . (4) . Panel A: seasonal HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Seasonal drought 0.317 0.068 −1.244*** 0.245*** (0.414) (0.115) (0.361) (0.094)  Constant −0.550 0.305 −1.304** 0.396** (0.814) (0.193) (0.597) (0.190)  R-squared 0.402 0.384 0.484 0.440  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Panel B: direct gestation HAZ (sd) Stunting (indicator) HAZ (sd) Stunting (indicator)  Gestation drought −0.120 0.146 −1.104*** 0.220** (0.426) (0.114) (0.360) (0.088)  Gestation flood −0.150 0.056 −0.041 0.091* (0.258) (0.071) (0.192) (0.051)  Constant −0.565 0.310 −1.279** 0.442** (0.813) (0.193) (0.600) (0.189)  R-squared 0.402 0.387 0.482 0.442  Mean dependent variable (HAZ) −1.624 0.441 −1.371 0.376 Observations 703 703 825 825 Notes: Table 6 reports heterogeneous impacts of estimated impacts of rainfall shocks on birth-to-HAZ and HAZ for boys and girls. Panel A presents results for seasonal drought, while Panel B presents estimated results for gestation drought and flood. These results refer to the baseline results in Columns (2) and (4) for Panels B and C of Tables 2 and 3. All regressions include all controls, locality-of-birth by month-of-birth fixed effects and year-of-birth by month-of-birth fixed effects. See Section 4 for a list of controls used in the regression analysis. See Tables 2 and 3 for additional notes. Robust standard errors clustered at the locality level are reported in parentheses. ***, ** and * represent significance at 1%, 5% and 10% levels, respectively. Open in new tab The pattern shows that the main results are disproportionately driven by the impacts of droughts on girls. This difference by gender for birth to anthropometric health is interesting evidence on the impact of early life shocks on health outcomes. The reversal of shock impacts between birth and short-term periods is quite intriguing. One explanation is that this may be related to potential behavioural responses in allocation of household resources, which coincidentally reduces the impact of shocks on boys in the short-term. We investigate this further by estimating the impact of most recent agricultural shock on food intake across boys and girls. Results in Table A3 demonstrate that rainfall deviation generally increases nutrient intake using food consumption score. More importantly, positive shock increases the nutrient mix for boys in a manner that is not evident for girls while negative shock shows a more prominent adverse effect for girls. Note that estimated impacts for girls are not statistically significant. While similar results have been presented in the literature on aggregate diet composition (Aguilar and Vicarelli, 2022), our results taken together indicate potential discriminatory behaviour against girls. These behavioural responses may be exacerbated by the information passed to parents to reverse boys' stunting classifications due to exposure to gestational shock. This outcome thrives amidst a general disincentive to invest in girl child human capital outcomes in Africa (Iddrisu et al., 2018). 5.3. Endogenous and selection issues One important limitation of the empirical analysis of gestational exposure to weather shocks and birth outcomes is the confounding nature of weather events and the household decision making process. We check for sample selection into childbearing as fertility decisions are endogenous and may be affected by weather conditions. To investigate this, we estimate the impact of recent agricultural shocks on mothers' use of contraceptives. Results show that there is no impact of recent seasonal harvest shocks on the adoption of contraception devices (Table A5). Similarly, we address potential selection bias in estimated coefficients with successful births using data on miscarried pregnancies from the DHS surveys used in this study. The data provides information on women with miscarried pregnancies and provides details such as the month and year of the event. Similar to the main results, we focus on the rural households surveyed in their permanent place of residence during the survey to tackle issues of adaptation and location sorting. We estimate foetal mortality models by conducting regressions of equations (5) and (6) using alternative measures of miscarriage. This includes counts of miscarriage cases and fraction of miscarried pregnancies during a conception year and across localities. The results are reported in Table A4. Estimated coefficients for all categories of outcome and gestational shocks (from precipitation and temperature) are small and generally imprecise. This result shows that the coefficient estimates reported for the main results are not likely to be biassed upwards by potential selection bias arising from miscarried pregnancies linked to adverse weather events11. 6. Discussion and conclusion In this paper, we match the birth outcomes of children to their anthropometric health (HAZ) in the short-term to examine the impact of early life events on their growth trajectory. While we document persistence in the impact of shocks on HAZ outcomes, we observe only marginal effects on birth outcomes. The pathways for the weather shocks include agricultural seasons' rainfall patterns that impact maternal nutrition and food security, droughts which create water scarcity and the asymmetric impact of negative temperature deviation relative to positive deviation. We complement the later by using trimester level heatwave indicator relating to stressor factor during conception. Heterogeneous results show that the impacts of gestational weather shocks which are stronger for boys immediately after birth have been reversed towards girls who show weaker health at a later stage of early life. Supporting analyses show evidence of discriminatory food allocation against girls which explains the gendered-differential reversal during the transition period. Our findings contribute to the research on the impact of weather shocks on health outcomes for low-income countries by examining the persistence and pathways of these effects. Lack of robust evidence on the transition of health outcomes across different stages of life and from different components of weather shocks hinders effective policy. This study is important for the SSA region due to the likelihood of increasing exposure to unpredictable weather events from climate change. Our findings can help provide structured evidence addressing both deficiencies in the research context and policy guidelines for similar settings. This paper provides an understanding of the importance of development milestones using links between birth and anthropometric health outcomes that have not been explored in the literature. The study context is placed within the critical programming period which captures both seasonal and gestational shocks to provide specific guidance for policy interventions. Our results compare to earlier studies on the impacts of early life events in Africa (Thai and Falaris, 2014; Abiona, 2017). Our main results show an approximately 32% (41%) decline in HAZ, corresponding to a 39% (43%) in likelihood of stunting, in response to incidence of seasonal (gestation) droughts across children aged 0 to 59 months. These results are comparable to estimated impacts of drought shocks on HAZ for Malawi where early life droughts lead to a decrease in age-standardised weight z-scores and HAZ of up to 43% and 27%, respectively (Abiona, 2017). Our findings also align with those from other studies in that the impacts of weather events are driven by droughts rather than impacts of floods on early life health. Other settings with similar results for the impacts of weather shocks on child height include Rabassa et al. (2014) for Nigeria, Thai and Falaris (2014) for Vietnam and Groppo and Kraehnert (2016) for Mongolia. This study presents a clear policy framework to consider in designing intervention programs to tackle early life weather shocks. There are indications that extremely low rainfall patterns are more disruptive for early life growth trajectories than excess rainfall patterns. The twist on impacts by gender between birth and anthropometric health outcomes is another dimension to guide policy. While our finding on the reversal of the impacts of shocks from boys to girls in the short-term is not new in the literature, food consumption patterns in favour of boys in periods of abundance consolidate the evidence in this area. Our findings provide additional evidence from the health transition dimension for more equitable household resource allocation decisions. In the absence of more robust evidence it is unclear if the disproportionate allocation of resources is due to deliberate efforts to sideline girls (Sahn and Stifel, 2002; Maitra and Rammohan, 2011). This study shows that timing of shocks relative to the period of conception, gestation, birth or breastfeeding is also an important factor for consideration of policy options. More broadly, the scope of policy aiming to address impacts of environmental shocks may be more targeted with an understanding of the residual impacts of the nature of the shocks especially as it relates to gender balance. Ways to mitigate these impacts include making primary healthcare facilities more accessible to rural households, providing extra support to pregnant or breastfeeding women, educating parents that boys and girls have the same nutritional needs regardless of the birth outcomes. Acknowledgement I wish to express my deep appreciation to UTS Business Research Grant Committee for providing financial support to carry out this research. I am also grateful to workshop attendants at the Centre for Health Economics Research and Evaluation (CHERE), University of Technology Sydney, and seminar participants at the 2021 International Health Economics Association (iHEA) Congress for their comments and suggestions. I am indebted to two anonymous referees who reviewed the paper and provided extensive feedback that helped shape and improve the overall quality of the paper. The survey data used in this article can be obtained from the website of DHS using the following links: https://dhsprogram.com/methodology/survey/survey-display-324.cfm and https://dhsprogram.com/methodology/survey/survey-display-450.cfm. The findings and opinions expressed in this paper are exclusively those of the author. The author is also solely responsible for the content and any errors. Supplementary material Supplementary material is available at Journal of African Economies online. Data availability The data underlying this article are available in the article and in its online supplementary material. Footnotes 1 The DHS data collects extensive contemporaneous HAZ measures (for children aged 0–59 months) in reference to the year and month of survey. It also collects historic record of birth weight for each child at birth as reported by their parents. Note that the gap between the period of birth weight record and HAZ measure is determined by the timeline of the age-in-months of each child at the date of the survey. 2 This list focuses on restricted rural pathways based on data availability and does not include exhaustive weather pathways. 3 To achieve this, we match GPS coordinates of each enumeration area to the GPS of the four closest weather stations from UDEL's weather data archive to obtain estimates of rainfall and temperature. This process used inverse weighted distance average of the weather parameters from the four closest weather stations to the GPS of the enumeration area similar to the approach used in other literature that we follow to construct the weather parameters in this study (Rocha and Soares, 2015). This approach may lead to small variation of absolute weather patterns especially for agricultural cycles of neighbouring enumeration areas with limited distribution of weather stations. We partially address this problem by using deviation of weather measures from the historical norm designated as shocks to provide an additional level of variation in this paper (see equations 1–4 below). 4 We depict the trimester-level shocks using the threshold method for depicting warmer temperatures in the literature (Jagnani et al., 2021). This is designated as 1 if |${\mathrm{temp}}_{lym}>25{}^{\circ}\mathrm{C}\ \mathrm{in}\ \mathrm{at}\ \mathrm{least}\ \mathrm{one}\ \mathrm{of}\ \mathrm{the}\ \mathrm{three}\ \mathrm{months}$| of the: first trimester for months |$m=-8\ to-6$|⁠; second trimester for months |$m=-5\ to-3$| and third trimester for months |$m=-2\ to\ 0.$| Note that each trimester is treated separately and each trimester indicator is designated 0 otherwise. 5 We include the agricultural cycle of 2002 to capture shocks for children born in early 2003. Overall, we use large spatial–temporal variation of the seasonal rainfall level when compared to the locality average to identify explanatory variables of interest. A similar pattern exists for the gestation period shocks in equations (2) and (3). 6 We refer to the fully-specified models—columns (2) and (4)—as the baseline results. 7 Nevertheless, the pattern of the coefficient estimates for birth weight and low birth weight are largely consistent with a priori theoretical expectation as seasonal drought is negatively (positively) correlated with birth weight (low birth weight indicator). 8 The out of season period for Sierra Leone (November, December, January, February and March) comprises the dry season months when extensive crop harvesting of the previous planting season takes place. There is minimal crop cultivation during this period compared to the rainy season. 9 A formal t-test for equality of coefficient estimates across boys and girls indicates differential estimates for all cases except one. 10 We also use gender interaction term in an alternative specification for sensitivity analysis of our gendered-differential results (results available in the supplementary file published with this paper). Overall, the results reinforce the main findings that the gender difference in impacts reverses from infancy to young children. 11 The infant mortality literature relates to the selection effect discussed in Section 5.3. Findings from the existing literature suggest that our estimates are likely to be conservative estimates (i.e. underestimate the true effects of weather shocks on child nutrition). References Abhishek C. ( 2010 ) ‘ Supply Shocks and Gender Bias in Child Health Investments: Evidence From the ICDS Programme in India ’, The B.E. Journal of Economic Analysis & Policy , 10 ( 1 ): 1 – 26 . Google Scholar OpenURL Placeholder Text WorldCat Abiona O. ( 2017 ) ‘ Adverse Effects of Early Life Extreme Precipitation Shocks on Short-Term Health and Adulthood Welfare Outcomes ’, Review of Development Economics , 21 ( 4 ): 1229 – 54 . Google Scholar Crossref Search ADS WorldCat Abiona O. ( 2022 ) ‘ Malnutrition Pathway for the Impact of in utero Drought Shock on Child Growth Indicators in Rural Households ’, Environment and Development Economics , 27 ( 1 ): 20 – 39 . Google Scholar Crossref Search ADS WorldCat Abiona O. , Ajefu J. ( 2022 ) ‘ The Impact of Timing of in-utero Drought Shocks on Birth Outcomes in Rural Households: Evidence from Sierra Leone ’, Journal of Population Economics . Forthcoming . https://doi.org/10.1007/s00148-022-00926-w. Google Scholar OpenURL Placeholder Text WorldCat Aguilar A. , Vicarelli M. ( 2022 ) ‘ El Niño and Children: Medium-Term Effects of Early-Life Weather Shocks on Cognitive and Health Outcomes ’, World Development , 150 ( C ): 105690. https://doi.org/10.1016/j.worlddev.2021.105690. Google Scholar OpenURL Placeholder Text WorldCat Almond D. , Currie J. ( 2011 ) ‘ Killing Me Softly: The Fetal Origins Hypothesis ’, Journal of Economic Perspectives , 25 ( 3 ): 153 – 72 . Google Scholar Crossref Search ADS PubMed WorldCat Almond D. , Currie J., Herrmann M. ( 2012 ) ‘ From Infant to Mother: Early Disease Environment and Future Maternal Health ’, Labour Economics , 19 ( 4 ): 475 – 83 . Google Scholar Crossref Search ADS WorldCat Andalón M. , Azevedo J. P., Rodríguez-Castelán C., Sanfelice V., Valderrama-González D. ( 2016 ) ‘ Weather Shocks and Health at Birth in Colombia ’, World Development , 82 : 69 – 82 . Google Scholar Crossref Search ADS WorldCat Baguet M. , Dumas C. ( 2019 ) ‘ How Does Birth Weight Affect Health and Human Capital? A Short- and Long-Term Evaluation ’, Health Economics , 28 ( 5 ): 597 – 617 . Google Scholar Crossref Search ADS PubMed WorldCat Baker R. E. , Anttila-Hughes J. ( 2020 ) ‘ Characterizing the Contribution of High Temperatures to Child Undernourishment in Sub-Saharan Africa ’, Scientific Reports , 10 : 18796 . Google Scholar Crossref Search ADS PubMed WorldCat Balietti A. , Datta S., Veljanoska S. ( 2022 ) ‘ Air Pollution and Child Development in India ’, Journal of Environmental Economics and Management , 113 : 102624. https://doi.org/10.1016/j.jeem.2022.102624. Google Scholar OpenURL Placeholder Text WorldCat Bassino J.-B. , Lagoarde-Segot T., Woitek U. ( 2022 ) ‘ Prenatal Climate Shocks and Adult Height in Developing Countries. Evidence from Japan (1872–1917) ’, Economics & Human Biology , 45 ( 101115 ): 101115. https://doi.org/10.1016/j.ehb.2022.101115. Google Scholar OpenURL Placeholder Text WorldCat Behrman J. R. , Rosenzweig M. R. ( 2004 ) ‘ Returns to Birthweight ’, The Review of Economics and Statistics , 86 ( 2 ): 586 – 601 . Google Scholar Crossref Search ADS WorldCat Black S. E. , Devereux P. J., Salvanes K. G. ( 2007 ) ‘ From the Cradle to the Labor Market? The Effect of Birthweight on Adult Outcomes ’, Quarterly Journal of Economics , 122 ( 1 ): 409 – 39 . Google Scholar Crossref Search ADS WorldCat Block S. , Haile B., You L., Headey D. ( 2022 ) ‘ Heat Shocks, Maize Yields, and Child Height in Tanzania ’, Food Security , 14 : 93 – 109 . Google Scholar Crossref Search ADS WorldCat Brainerd E. , Menon N. ( 2014 ) ‘ Seasonal Effects of Water Quality: The Hidden Costs of the Green Revolution to Infant and Child Health in India ’, Journal of Development Economics , 107 ( C ): 49 – 64 . Google Scholar OpenURL Placeholder Text WorldCat Burlando A. ( 2014 ) ‘ Transitory Shocks and Birth Weights: Evidence from a Blackout in Zanzibar ’, Journal of Development Economics , 108 ( C ): 154 – 68 . Google Scholar OpenURL Placeholder Text WorldCat Carrillo B. ( 2020 ) ‘ Early Rainfall Shocks and Later-Life Outcomes: Evidence from Colombia ’, The World Bank Economic Review , 34 ( 1 ): 179 – 209 . Google Scholar Crossref Search ADS WorldCat Case A. , Angela F., Christina P. ( 2005 ) ‘ The Lasting Impact of Childhood Health and Circumstance ’, Journal of Health Economics , 24 ( 2 ): 365 – 89 . Google Scholar Crossref Search ADS PubMed WorldCat Case A. , Paxson C. ( 2008 ) ‘ Stature and Status: Height, Ability, and Labor Market Outcomes ’, Journal of Political Economy , 116 ( 3 ): 499 – 532 . Google Scholar Crossref Search ADS PubMed WorldCat Chen X. , Tan C. M., Zhang X., Zhang X. ( 2020 ) ‘ The Effects of Prenatal Exposure to Temperature Extremes on Birth Outcomes: The Case of China ’, Journal of Population Economics , 33 ( 4 ): 1263 – 302 . Google Scholar Crossref Search ADS PubMed WorldCat Comfort A. B. ( 2016 ) ‘ Long-Term Effect of in utero Conditions on Maternal Survival Later in Life: Evidence from Sub-Saharan Africa ’, Journal of Population Economics , 29 ( 2 ): 493 – 527 . Google Scholar Crossref Search ADS WorldCat Conti G. , Hanson M., Inskip H., Crozier S., Cooper C., Godfrey K. ( 2020 ) Beyond Birthweight: The Origins of Human Capital, IZA Working Paper 13296. Institute of Labor Economics. Cornwell K. , Inder B. ( 2015 ) ‘ Child Health and Rainfall in Early Life ’, Journal of Development Studies , 51 ( 7 ): 865 – 80 . Google Scholar Crossref Search ADS WorldCat Daghagh Yazd S. , Wheeler S. A., Zuo A. ( 2020 ) ‘ Understanding the Impacts of Water Scarcity and Socio-Economic Demographics on Farmer Mental Health in the Murray-Darling Basin ’, Ecological Economics , 169 ( C ): 106564. https://doi.org/10.1016/j.ecolecon.2019.106564. Google Scholar OpenURL Placeholder Text WorldCat Desbureaux S. , Rodella A.-S. ( 2019 ) ‘ Drought in the City: The Economic Impact of Water Scarcity in Latin American Metropolitan Areas ’, World Development , 114 ( C ): 13 – 27 . Google Scholar OpenURL Placeholder Text WorldCat Deschênes O. , Greenstone M., Guryan J. ( 2009 ) ‘ Climate Change and Birthweight ’, The American Economic Review , 99 ( 2 ): 211 – 7 . Google Scholar Crossref Search ADS PubMed WorldCat Feeny S. , Mishra A., Trinh T.-A., Ye L., Zhu A. ( 2021 ) ‘ Early-Life Exposure to Rainfall Shocks and Gender Gaps in Employment: Findings from Vietnam ’, Journal of Economic Behavior and Organization , 183 ( C ): 533 – 54 . Google Scholar OpenURL Placeholder Text WorldCat Freudenreich H. , Aladysheva A., Brück T. ( 2022 ) ‘ Weather Shocks Across Seasons and Child Health: Evidence from a Panel Study in the Kyrgyz Republic ’, World Development , 155 ( 105801 ): 105801. https://doi.org/10.1016/j.worlddev.2021.105801. Google Scholar OpenURL Placeholder Text WorldCat Geruso M. , Spears D. ( 2018 ) ‘ Heat, Humidity, and Infant Mortality in the Developing World ’, NBER Working Paper Series. Working Paper 24870 . http://www.nber.org/papers/w24870. Groppo V. , Kraehnert K. ( 2016 ) ‘ Extreme Weather Events and Child Height: Evidence from Mongolia ’, World Development , 86 : 59 – 78 . Google Scholar Crossref Search ADS WorldCat Hirvonen K. , Sohnesen T. P., Bundervoet T. ( 2020 ) ‘ Impact of Ethiopia's 2015 Drought on Child Undernutrition ’, World Development , 131 ( 104964 ): 104964. https://doi.org/10.1016/j.worlddev.2020.104964. Google Scholar OpenURL Placeholder Text WorldCat Hoddinott J. , Kinsey B. ( 2001 ) ‘ Child Growth in the Time of Drought ’, Oxford Bulletin of Economics and Statistics , 63 ( 4 ): 409 – 36 . Google Scholar Crossref Search ADS WorldCat Iddrisu A. M. , Danquah M., Quartey P., Ohemeng W. ( 2018 ) ‘ Gender Bias in Households’ Educational Expenditures: Does the Stage of Schooling Matter? ’ World Development Perspectives , 10–12 : 15 – 23 . Google Scholar OpenURL Placeholder Text WorldCat Jagnani M. , Barrett C. B., Liu Y., You L. ( 2021 ) ‘ Within-Season Producer Response to Warmer Temperatures: Defensive Investments by Kenyan Farmers ’, The Economic Journal , 131 ( 633 ): 392 – 419 . Google Scholar Crossref Search ADS WorldCat Kaul T. ( 2018 ) ‘ Intra-Household Allocation of Educational Expenses: Gender Discrimination and Investing in the Future ’, World Development , 104 ( C ): 336 – 43 . Google Scholar OpenURL Placeholder Text WorldCat Kudamatsu M. , Persson T., Strömberg D. ( 2012 ) ‘ Weather and Infant Mortality in Africa ’, CEPR Discussion Paper , No. DP9222 . https://ssrn.com/abstract=2210191. Google Scholar OpenURL Placeholder Text WorldCat Lee W.-S. , Li B. G. ( 2021 ) ‘ Extreme Weather and Mortality: Evidence from Two Millennia of Chinese Elites ’, Journal of Health Economics . Forthcoming . https://doi.org/10.1016/j.jhealeco.2020.102401. Google Scholar OpenURL Placeholder Text WorldCat Maitra P. , Rammohan A. ( 2011 ) ‘ The Link between Infant Mortality and Child Nutrition in India: Is There any Evidence of a Gender Bias? ’ Journal of the Asia Pacific Economy , 16 ( 1 ): 81 – 110 . Google Scholar Crossref Search ADS WorldCat Majid M. F. ( 2015 ) ‘ The Persistent Effects of In Utero Nutrition Shocks over the Life Cycle: Evidence from Ramadan Fasting ’, Journal of Development Economics , 117 : 48 – 57 . Google Scholar Crossref Search ADS WorldCat Matsuura K. , Willmott C. ( 2017 ) ‘Terrestrial air temperature and precipitation: 1900–2017’, in Gridded Monthly Time Series ' Version 5.01 . Center for Climatic Research, University of Delaware , Delaware, United States of America . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Meng X. , Qian N. ( 2009 ) The long-term consequences of famine on survivors: evidence from a unique natural experiment using China's great famine. NBER working paper no. 14917 . Cambridge : National Bureau of Economic Research, Inc . Molina O. , Saldarriaga V. ( 2017 ) ‘ The Perils of Climate Change: In utero Exposure to Temperature Variability and Birth Outcomes in the Andean Region ’, Economics & Human Biology , 24 : 111 – 24 . Google Scholar Crossref Search ADS WorldCat Mulmi P. , Block S. A., Shively G. E., Masters W. A. ( 2016 ) ‘ Climatic Conditions and Child Height: Sex-Specific Vulnerability and the Protective Effects of Sanitation and Food Markets in Nepal ’, Economics and Human Biology , 23 ( C ): 63 – 75 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Ngegba M. P. , Moriba P., Kandeh J. B. A., Moiwo J. P., Massaquoi S. B. ( 2018 ) ‘ Assessing Efficiency of Action Aid Sierra Leone (AASL) Extension Services in Sierra Leone ’, International Journal of Agricultural Extension , 6 ( 2 ): 129 – 38 . Google Scholar Crossref Search ADS WorldCat Rabassa M. , Skoufias E., Jacoby H. ( 2014 ) ‘ Weather and Child Health in Rural Nigeria ’, Journal of African Economies , 23 : 464 – 92 . Google Scholar Crossref Search ADS WorldCat Randell H. , Gray C., Grace K. ( 2020 ) ‘ Stunted from the Start: Early Life Weather Conditions and Child Undernutrition in Ethiopia ’, Social Science & Medicine , 261 ( C ): 113234. https://doi.org/10.1016/j.socscimed.2020.113234. Google Scholar OpenURL Placeholder Text WorldCat Rocha R. , Soares R. R. ( 2015 ) ‘ Water Scarcity and Birth Outcomes in the Brazilian Semiarid ’, Journal of Development Economics , 112 : 72 – 91 . Google Scholar Crossref Search ADS WorldCat Rodríguez L. ( 2016 ) ‘ Intrahousehold Inequalities in Child Rights and Well-Being. A Barrier to Progress? ’ World Development , 83 ( C ): 111 – 34 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Sahn D. E. , Stifel D. C. ( 2002 ) ‘ Parental Preferences for Nutrition of Boys and Girls: Evidence from Africa ’, Journal of Development Studies , 39 ( 1 ): 21 – 45 . Google Scholar Crossref Search ADS WorldCat Schetter C. D. ( 2011 ) ‘ Psychological Science on Pregnancy: Stress Processes, Biopsychosocial Models, and Emerging Research Issues ’, Annual Review of Psychology , 62 : 531 – 58 . Google Scholar Crossref Search ADS PubMed WorldCat Thai T. Q. , Falaris E. M. ( 2014 ) ‘ Child Schooling, Child Health, and Rainfall Shocks: Evidence from Rural Vietnam ’, The Journal of Development Studies , 50 ( 7 ): 1025 – 37 . Google Scholar Crossref Search ADS WorldCat Thiede B. , Strube J. ( 2020 ) ‘ Climate Variability and Child Nutrition: Findings from Sub-Saharan Africa ’, Global Environmental Change , 65 ( C ): 102192. https://doi.org/10.1016/j.gloenvcha.2020.102192. Google Scholar OpenURL Placeholder Text WorldCat Torche F. ( 2018 ) ‘ Prenatal Exposure to an Acute Stressor and children's Cognitive Outcomes ’, Demography , 55 ( 5 ): 1611 – s39 . Google Scholar Crossref Search ADS PubMed WorldCat © The Author(s) 2023. Published by Oxford University Press on behalf of the Centre for the Study of African Economies, all rights reserved. For Permissions, please email: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

Journal of African Economies – Oxford University Press

Published: Jan 16, 2023

References

  • Supply Shocks and Gender Bias in Child Health Investments: Evidence From the ICDS Programme in India
  • Adverse Effects of Early Life Extreme Precipitation Shocks on Short-Term Health and Adulthood Welfare Outcomes
  • Malnutrition Pathway for the Impact of in utero Drought Shock on Child Growth Indicators in Rural Households
  • The Impact of Timing of in-utero Drought Shocks on Birth Outcomes in Rural Households: Evidence from Sierra Leone
    Abiona, O.; Ajefu, J.
  • El Niño and Children: Medium-Term Effects of Early-Life Weather Shocks on Cognitive and Health Outcomes
    Aguilar, A.; Vicarelli, M.
  • Killing Me Softly: The Fetal Origins Hypothesis
    Almond, D.; Currie, J.
  • From Infant to Mother: Early Disease Environment and Future Maternal Health
    Almond, D.; Currie, J.; Herrmann, M.
  • Weather Shocks and Health at Birth in Colombia
    Andalón, M.; Azevedo, J. P.; Rodríguez-Castelán, C.; Sanfelice, V.; Valderrama-González, D.
  • How Does Birth Weight Affect Health and Human Capital? A Short- and Long-Term Evaluation
    Baguet, M.; Dumas, C.
  • Characterizing the Contribution of High Temperatures to Child Undernourishment in Sub-Saharan Africa
    Baker, R. E.; Anttila-Hughes, J.
  • Air Pollution and Child Development in India
    Balietti, A.; Datta, S.; Veljanoska, S.
  • Prenatal Climate Shocks and Adult Height in Developing Countries. Evidence from Japan (1872–1917)
    Bassino, J.-B.; Lagoarde-Segot, T.; Woitek, U.
  • Returns to Birthweight
    Behrman, J. R.; Rosenzweig, M. R.
  • From the Cradle to the Labor Market? The Effect of Birthweight on Adult Outcomes
    Black, S. E.; Devereux, P. J.; Salvanes, K. G.
  • Heat Shocks, Maize Yields, and Child Height in Tanzania
    Block, S.; Haile, B.; You, L.; Headey, D.
  • Seasonal Effects of Water Quality: The Hidden Costs of the Green Revolution to Infant and Child Health in India
    Brainerd, E.; Menon, N.
  • Transitory Shocks and Birth Weights: Evidence from a Blackout in Zanzibar
  • Early Rainfall Shocks and Later-Life Outcomes: Evidence from Colombia
  • The Lasting Impact of Childhood Health and Circumstance
    Case, A.; Angela, F.; Christina, P.
  • Stature and Status: Height, Ability, and Labor Market Outcomes
    Case, A.; Paxson, C.
  • The Effects of Prenatal Exposure to Temperature Extremes on Birth Outcomes: The Case of China
    Chen, X.; Tan, C. M.; Zhang, X.; Zhang, X.
  • Long-Term Effect of in utero Conditions on Maternal Survival Later in Life: Evidence from Sub-Saharan Africa
  • Child Health and Rainfall in Early Life
    Cornwell, K.; Inder, B.
  • Understanding the Impacts of Water Scarcity and Socio-Economic Demographics on Farmer Mental Health in the Murray-Darling Basin
    Daghagh, Yazd S.; Wheeler, S. A.; Zuo, A.
  • Drought in the City: The Economic Impact of Water Scarcity in Latin American Metropolitan Areas
    Desbureaux, S.; Rodella, A.-S.
  • Climate Change and Birthweight
    Deschênes, O.; Greenstone, M.; Guryan, J.
  • Early-Life Exposure to Rainfall Shocks and Gender Gaps in Employment: Findings from Vietnam
    Feeny, S.; Mishra, A.; Trinh, T.-A.; Ye, L.; Zhu, A.
  • Weather Shocks Across Seasons and Child Health: Evidence from a Panel Study in the Kyrgyz Republic
    Freudenreich, H.; Aladysheva, A.; Brück, T.
  • Extreme Weather Events and Child Height: Evidence from Mongolia
    Groppo, V.; Kraehnert, K.
  • Impact of Ethiopia's 2015 Drought on Child Undernutrition
    Hirvonen, K.; Sohnesen, T. P.; Bundervoet, T.
  • Child Growth in the Time of Drought
    Hoddinott, J.; Kinsey, B.
  • Gender Bias in Households’ Educational Expenditures: Does the Stage of Schooling Matter?
    Iddrisu, A. M.; Danquah, M.; Quartey, P.; Ohemeng, W.
  • Within-Season Producer Response to Warmer Temperatures: Defensive Investments by Kenyan Farmers
    Jagnani, M.; Barrett, C. B.; Liu, Y.; You, L.
  • Intra-Household Allocation of Educational Expenses: Gender Discrimination and Investing in the Future
  • Weather and Infant Mortality in Africa
    Kudamatsu, M.; Persson, T.; Strömberg, D.
  • Extreme Weather and Mortality: Evidence from Two Millennia of Chinese Elites
    Lee, W.-S.; Li, B. G.
  • The Link between Infant Mortality and Child Nutrition in India: Is There any Evidence of a Gender Bias?
    Maitra, P.; Rammohan, A.
  • The Persistent Effects of In Utero Nutrition Shocks over the Life Cycle: Evidence from Ramadan Fasting
  • Gridded Monthly Time Series
    Matsuura, K.; Willmott, C.
  • The Perils of Climate Change: In utero Exposure to Temperature Variability and Birth Outcomes in the Andean Region
    Molina, O.; Saldarriaga, V.
  • Climatic Conditions and Child Height: Sex-Specific Vulnerability and the Protective Effects of Sanitation and Food Markets in Nepal
    Mulmi, P.; Block, S. A.; Shively, G. E.; Masters, W. A.
  • Assessing Efficiency of Action Aid Sierra Leone (AASL) Extension Services in Sierra Leone
    Ngegba, M. P.; Moriba, P.; Kandeh, J. B. A.; Moiwo, J. P.; Massaquoi, S. B.
  • Weather and Child Health in Rural Nigeria
    Rabassa, M.; Skoufias, E.; Jacoby, H.
  • Stunted from the Start: Early Life Weather Conditions and Child Undernutrition in Ethiopia
    Randell, H.; Gray, C.; Grace, K.
  • Water Scarcity and Birth Outcomes in the Brazilian Semiarid
    Rocha, R.; Soares, R. R.
  • Intrahousehold Inequalities in Child Rights and Well-Being. A Barrier to Progress?
  • Parental Preferences for Nutrition of Boys and Girls: Evidence from Africa
    Sahn, D. E.; Stifel, D. C.
  • Psychological Science on Pregnancy: Stress Processes, Biopsychosocial Models, and Emerging Research Issues
  • Child Schooling, Child Health, and Rainfall Shocks: Evidence from Rural Vietnam
    Thai, T. Q.; Falaris, E. M.
  • Climate Variability and Child Nutrition: Findings from Sub-Saharan Africa
    Thiede, B.; Strube, J.
  • Prenatal Exposure to an Acute Stressor and children's Cognitive Outcomes