WPS8167 Policy Research Working Paper 8167 Tracing Back the Weather Origins of Human Welfare Evidence from Mozambique Javier E. Baez German Caruso Chiyu Niu Poverty and Equity Global Practice Group August 2017 Policy Research Working Paper 8167 Abstract Mozambique is among the African countries most exposed exhibit lower consumption and are more prone to be poor. to weather-related hazards. Using detailed gridded precip- In disentangling the mechanisms at play, this paper presents itation data for individuals’ birth-year and birth-district, suggestive evidence of variation in agricultural output, food this study investigates the effects of extreme rainfall anom- security, and subsequent detrimental effects on human cap- alies around the time of birth on long-run well-being. The ital accumulation as important drivers behind the impacts. results show that the socioeconomic outcomes of adults The study concludes that policy efforts aimed at accelerating are influenced by weather shocks that occur early in life. poverty reduction in Mozambique will have to consider Individuals exposed to floods while in utero or during the the inability of rural households to shield the well-being of first year of life are less likely to participate in the labor children from the consequences of extreme weather shocks. market. Consequently, the households that they are heading This paper is a product of the Poverty and Equity Global Practice Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at jbaez@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Tracing Back the Weather Origins of Human Welfare: Evidence from Mozambique Javier E. Baez German Caruso Chiyu Niu * JEL Classification: I3, J2, O1 Keywords: Natural disasters, long-term human welfare, poverty, human capital. __________________ * Baez: Poverty and Equity Global Practice, Africa Region, World Bank, and Research Fellow, Institute of the Study of Labor (IZA) (jbaez@worldbank.org); Caruso: Economist, Poverty and Equity Global Practice, Latin America and Caribbean Region, World Bank (gcaruso@worldbank.org); Niu: Department of Agricultural and Consumer Economics, University of Illinois We received valuable feedback from seminar participants at the World Bank (Washington D.C.) and USAID (Maputo, Mozambique). We are thankful to Siobhan Murray (World Bank) for producing the spatially and crop explicit water requirement satisfaction index for Mozambique. We also acknowledge superb research assistance from Alessia Thiebaud (World Bank). 1. Introduction Risks, both systemic and idiosyncratic, are part of normal life, especially in developing countries. When they materialize, households can experience considerable hardship, including the destruction of assets, deterioration of health, and loss of employment, income and consumption. Many of the risks frequently faced by households, especially the poor and vulnerable in rural areas, are inherent to agriculture. This is particularly problematic for most countries in Sub-Saharan Africa (SSA), where agriculture constitutes the main economic sector that supports the livelihoods of the vast majority of the population. Climate hazards are increasingly becoming the most common risk faced by rural farmers. Droughts, floods, sea rise and extreme temperatures are projected to become more intense and frequent owing to global warming and to the emergence of new weather patterns. The consequences of these trends are not trivial. Recent estimates indicate that climate change could force more than 100 million people into extreme poverty by 2030. Global crop yields are expected to drop by as much as 5 percent in 2030, even after accounting for higher technology adoption and innovation (Hallegatte et al. 2016). This will likely prompt distressing effects through potentially higher food prices and changes in rural incomes. Climate change is also expected to exacerbate health risks such as malaria and diarrhea. A growing body of research shows that weather shocks can influence key development outcomes at the household level. Surveys of the literature typically conclude that extreme weather has significant detrimental damages on nutrition, education and health. In coping with adverse events, households may dispose of productive assets, such as livestock and land, further undermining their income generating processes, consumption and other dimensions of welfare (Baez et al. 2010). Beyond their contemporaneous effects, the consequences of climate-related shocks can be long-lasting. For instance, Alderman et al. (2006) find that exposure of individuals to droughts in Zimbabwe between 1982 and 1984 during childhood explains that they were 3.4 centimeters shorter and earned 14 percent less by adulthood. Similarly, Maccini and Yang (2009) find that the health, education, and socioeconomic outcomes of adult Indonesian women were sensitive to weather shocks experienced in early years of life. Women exposed to higher levels of 2 rainfall (relative to the local norm) were 0.57 centimeters taller, completed 0.22 more schooling grades, and lived in households scoring 0.12 standard deviations higher on an asset index. This paper investigates empirically the effects of extreme rainfall anomalies around the time of birth on long-run individual well-being in Mozambique. The country provides a relevant context to examine whether environmental conditions at the beginning of life influence outcomes by adulthood. Mozambique is particularly prone to experiencing a wide range of natural hazards and is also heavily reliant on agriculture. The country’s combined risk to different natural disasters is the highest in the Southeastern Africa region (Index for Risk Management, 2016). In addition, agriculture, comprised mostly of rain-fed crops, small-scale farms and subsistence farming, constitutes the main pillar of the Mozambican economy, representing a key source of livelihood for more than three-quarters of the population. The results show that long-term socioeconomic outcomes in Mozambique are sensitive to weather shocks that occur at the earliest stages of life. Precipitation anomalies around the time of birth (measured as standard deviations away from historical average rainfall for each geographical grid) play an important role in explaining lower human capital accumulation, lower labor force participation and lower expenditures in adulthood. More specifically, individuals who were in areas affected by a rainfall shock while in utero or during the first year of life later exhibited a rate of participation in the labor market that was 6% lower than the control mean. In addition, households headed by individuals who were hit by floods while in utero and during early childhood record 14 percent lower consumption per capita than the median consumption of the comparison group. Consequently, floods affecting individuals early in life are found to significantly increase the likelihood that their households will fall below the poverty line (by 9 percentage points, or around 18%). The long-lasting impacts are arguably driven by the negative impacts of extreme weather conditions on agricultural output and, in turn, by their adverse effects on income, food security, nutrition and human capital accumulation. We attempt to disentangle the mechanisms underlying the persistence of extreme weather anomalies. As expected, crop performance for some of the main food staples grown in Mozambique varies largely with water availability conditions. Disturbances in rainfall patterns lead to lower agricultural yields, which then limit physical and economic access 3 to food of affected households, weakening the nutritional status of individuals, particularly children. The results show that shortly after the shocks, children affected by weather anomalies in early life are about 0.5 to 0.8 standard deviations smaller than children who were not. In tandem with the negative effects on nutrition, or partly as a result of them, school attainment is also undermined. Affected children are less likely to attend school regularly, and are shown to accumulate 0.2 fewer years of schooling than other children, on average. This is equivalent to a drop in the school attainment of affected children of 6%, relative to the control mean. Overall, affected individuals are 1.4 percentage points (2.8%) less likely to accumulate some education than unaffected individuals. These effects seem to persist until adulthood: suggestive evidence shows that affected women are shown to become around 0.5-0.7 centimeters shorter than control women and have lower labor productivity. This report is structured as follows. Section 2 describes the socioeconomic and weather risk context of Mozambique. Section 3 presents the empirical strategy and data sources used in the paper. In turn, Section 4 presents and discusses the main results of the analysis. Finally, Section 5 concludes by summarizing the findings and discussing policy recommendations. 2. Country background Socioeconomic context Over the past two decades, Mozambique has achieved rapid economic growth in a context of relative macroeconomic stability. After independence from Portugal in 1975, the country entered a prolonged period of civil war. However, since the end of the conflict in 1992, Mozambique became one of the fastest growing economies in SSA. Between 1993 and 2014, the country’s economy grew at an average of 7.9 percent annually, thanks to responsible macroeconomic and structural policies, implemented in a politically stable post-war environment, and to large flows of foreign investment and donor support. While the contribution of mega-projects to GDP growth has been increasing, the agricultural sector remains the largest sector of the Mozambican economy. The primary sector accounts for a quarter of the country’s GDP and employs the vast majority of the workforce, mostly in low- productivity, self-employment activities. Most farmers practice rain-fed, low-technology 4 subsistence agriculture in small-scale farms, ranging from 0.5 to 1.5 hectares. The overwhelming majority of farmers lack formal land titles. The movement of labor out of agriculture has been slow, and is largely attributable to low rural-urban migration. The pace of sustained and high economic growth has not translated into fast poverty reduction. After the end of the civil war in 1993, Mozambique ranked third among the poorest countries in the world. Since then, the country has seen a moderate decline in poverty rates. However, poverty remains unevenly distributed and concentrated in rural areas. Along the same lines, Mozambique continues to perform poorly across a wide range of human development indicators. In 2014, Mozambique was ranked 178 out of 198 in the Human Development Index. The coverage and quality of public service provision remain low, especially in rural areas. Climate risk profile Globally and regionally, Mozambique is among the countries with the largest exposure to multiple natural hazards as it is extensively and increasingly subject to cyclones, floods and droughts and secondary hazards arising from these events (EM-DAT, INFORM, 2016). Long droughts are recurrent and stem from a combination of low levels of precipitation and the overgrazing and overuse of agricultural lands. They are experienced in 7 out of 10 years in the Southern regions, and in 4 out of 10 years in the Central regions. Drought risk is projected to increase over the coming years, both in terms of higher frequency and longer duration of droughts. Mozambique is also particularly exposed to cyclones. The Mozambican coastline borders one of the most active basins of tropical cyclones, the Southwest Indian Ocean, which alone produces about 10% of all cyclones worldwide. Each year, on average, Mozambique is hit by one tropical storm or cyclone, and by three or four additional tropical disturbances (UN-Habitat, 2015). The cyclone season goes from October to April, with the most intense storms usually taking place between February and April. Tropical cyclones have produced devastating effects in the country, with five tropical cyclones (of category 1 to 4) making landfall between 2000 and 2008. The coastal region, home to over 60 percent of Mozambicans, is the most heavily affected and often experiences widespread destruction of infrastructure and population displacement because of cyclones (GFDRR, 2012). 5 Floods are also a recurring event, which result from the high winds and heavy rain associated with cyclones, but also from a combination of excess rainfall, upstream discharges from major river basins, and poor drainage infrastructure. Floods generally occur every two or three years, mostly during the rainy season and along the nine major international river systems that cross Mozambique, or across the low-lying, densely-populated coastal areas (GFDRR, 2012). Particularly severe floods tend to occur once every 15 to 20 years (GFDRR, 2012), depending on rainfall levels both in and outside of Mozambique, where the main rivers originate (UN-Habitat, 2015). The costs of weather shocks are particularly high given Mozambique’s heavy reliance on the agricultural sector for providing livelihoods to the majority of its population. Almost all production (97%) comes from rain-fed agriculture, which is particularly vulnerable to extreme weather. A 2009 estimate of drought and flood costs places average annual losses of maize and sorghum at 9 percent and 7 percent of each crop, respectively. Further losses of around 20 percent of crops are also estimated to occur once every ten years. Climate shocks also impose costs on buildings and physical infrastructure. It has been estimated that an average of 100km of roads and 33,000 households are impacted by flooding every year in Mozambique. There is also substantial spatial variation in the frequency and intensity of natural hazards across the country. For instance, the southern and inland regions are more prone to suffer droughts than the northern and coastal regions. Conversely, areas most affected by cyclones are generally located along the coastline. Areas along the major international river systems are more susceptible to floods. 3. Research Design and Data Identification Strategy The objective of this paper is to estimate the causal effect of exposure to rainfall shocks in early childhood on long-term socioeconomic outcomes. To attain identification, the analysis exploits variation in rainfall across geographic areas and over time. The spatial difference compares the outcomes of adults that were in districts exposed to extreme rainfall anomalies around the time of birth (i.e. in utero or during the first year of life) against outcomes of individuals in the same age cohort that resided in districts not affected by weather shocks. The time difference compares the outcomes of individuals from slightly older or younger cohorts in affected and 6 unaffected districts, which provide two additional control groups. Reduced-form impacts of the shocks were estimated using the following general empirical relationship between the long-term outcome Y of individual i in district d and at time t, and the shock indicators: =∑ ( , ∗ Shock , in utero + , ∗ Shock ∗ 1st year of life ) + District FE + Cohort FE+ ′ + (1) Where the parameters of interest are , and , , which measure the effects of either extreme droughts (s = 1) or extreme floods (s = 2) on the outcomes of interest, whether they affected the child while in utero (indicated by the sub-index u) or during the first year of life (indicated by the sub-index m). The empirical model also includes controls for district and cohort fixed effects, as well as gender and other individual-level covariates (e.g. school attainment), which are captured in the term .1 Finally, the term stands for an individual level zero-mean error term. Standard errors are clustered at the same level of the treatment, namely by district of birth-year. The results of all the empirical models are presented both for the whole sample of adults in the surveys of the census, irrespective of the area of birth (urban or rural), and solely for individuals born in rural districts. Data This type of analysis requires extensive data to match indicators of long-run individual well- being with rainfall anomalies that occur around the time of birth. For the weather information, we use extensive spatial and temporal gridded monthly precipitation data for global land areas for the period 1901-2014, with a resolution of 0.5 x 0.5 degrees (50 x 50 km). This data set is known as CRU TS and is managed by the Climate Research Unit at University of East Anglia. The data are constructed using records from more than 2,400 meteorological stations across the world’s land areas. Station anomalies are interpolated into grid cells combined with local climatology data to obtain absolute monthly values. The variables included in the data set in addition to precipitation are mean temperature, diurnal temperature range, wet-day frequency, vapor pressure and cloud cover (Harris, et al. 2014). For the empirical analysis, all rainfall and temperature variables are 1The empirical models were also run without controlling for the school attainment of the individual since this indicator is endogenous to the weather shocks. 7 weighted by population at the district level and are obtained by overlaying the gridded CRU TS climate data with population and district-level grid maps. Data on demographics and socioeconomics at the individual, household and district levels are drawn from three main sources. The first one is the 2008-09 income and expenditure household survey (known as IOF, for its acronym in Portuguese) collected by the National Institute of Statistics of Mozambique. The IOF survey is representative at the province level and, in addition to providing detailed data on expenditures (including self-consumption) and incomes, it contains a wealth of information on sociodemographic characteristics, education, health, children’s anthropometrics, labor market indicators and housing conditions. The second source of data is the 2011 Demographic and Health Survey (DHS), which provides a range of variables in aspects related to population, health and nutrition. Lastly, we used a 10% random sample of the 2007 population census to obtain additional information on human capital indicators. Rainfall shocks Mozambique has a tropical climate that is characterized by two seasons. The wet (rainy) season goes from October to March, and is characterized by heavier rainfall between December and March. Most cyclones are also experienced over the same period, especially in the coastal areas. The dry season extends from April to September, and is particularly pronounced in the south of the country, making it more prone to drought. Due to Mozambique’s location in the tropical zone, temperatures do not fluctuate much within and across seasons. Rainfall shocks are prevalent across the country. We define them as district-level annual cumulative rainfall events between 1950 and 2014 that were two standard deviations above (for floods) or below (for droughts) the historical annual mean for the corresponding district. The annual historical mean is calculated over the 1901-2014 period for each of the 128 districts comprising the 11 provinces of Mozambique. It is observed that extreme precipitation (whether in deficit or in excess) occurs with relative frequency. There is wide geographical variation across the country; provinces such as Niassa, Tete, Inhambane and Manica record the largest number of rainfall anomalies (Figure 1). Major disasters such as the floods that occurred in 2000, which were caused by Cyclone Eline and Cyclone Judah and affected mostly districts in the Inhambane and Gaza provinces, are well captured by the rainfall shock variable (Figure 2). 8 4. Results Long-term effects We first investigate whether individuals that were exposed to severe droughts or floods while in utero or during their first year of life are less likely to participate in the labor market later in life. The 2008-09 IOF collects several variables that can be used to construct measures of participation in paid and unpaid activities. We constructed two indicators. The first one (Labor Participation 1) measures engagement or desire to work in remunerated activities, whereas the second one (Labor Participation 2) adds participation in non-remunerated activities –a form of labor engagement that is quite prevalent among workers in the country– to the first definition. The results indicate that excess of rainfall while in utero had a negative effect on the probability to participate in the labor market in adulthood (Table 1). Affected individuals exhibit on average a participation rate that is 6% compared to the control mean. This finding is consistent across subsamples (all versus rural) and is also robust to the inclusion of spatial (i.e. district) fixed effects. Since the 2008-09 IOF survey did not collect the exact month of birth (but only the year) of the individuals interviewed, the window used to define the in utero and first year of life periods are a noisy measure of the actual stage of life. Arguably, this could explain why the effects of floods are seen only for individuals classified as being in utero. As for the droughts, they are not found to influence labor force participation, regardless of whether they were experienced while in utero or in the first year of life. We also assess the effects of extreme weather on the well-being of individuals investigating the relationship between droughts and floods and expenditures per capita, as measured by the IOF 2008-09. The results indicate that extreme floods are associated with lower expenditures per capita and, consequently, higher likelihood of households with affected individuals to be poor (Table 2). Seeking to pinpoint the direct effects on household well-being, the subsample for this part of the analysis consists of individuals that are household heads. In line with the results on labor participation, households headed by individuals that were hit by floods while in utero and during early childhood record lower consumption per capita, approximately 14% lower than the median consumption of the comparison group. The results are statistically weaker for the rural subsample. Therefore, lower consumption per capita due to floods that affected individuals early in life 9 increases the likelihood that the households they head in adulthood fall below the poverty line (9 percentage points or around 18%). This evidence suggests that effects of uninsured weather shocks that occurred decades ago show strong persistence over time and are still felt by affected individuals and their families to this day. Why negative effects? The burden placed by rainfall shocks on agricultural output is identified as the most plausible channel driving the effects on socioeconomic outcomes in the long run. Back in the 1970s, virtually all Mozambicans (nearly 98 percent) lived in rural areas. At the time, agriculture –even more so than today— represented the main source of their livelihoods. Arguably, agriculture is therefore the main mechanism at play behind the observed negative impacts. In fact, the shocks examined in this paper are expected to strongly affect the water balance of soil by translating into too much or too little water. Severe droughts or floods are expected to exert a negative impact on yields and, in turn, likely on household consumption, food security and income. We use the Water Requirement Satisfaction Index (WRSI) to examine the relationship between water conditions and agricultural yields in Mozambique. The WRSI is a measure employed by the Famine Early Warning System project (FEWS NET) to predict harvest outcomes and to identify potential food security issues on a seasonal basis. The WRSI is defined as the ratio of seasonal actual crop evapotranspiration to the seasonal crop water requirement, and captures the expected impact of water deficits on harvest at different points in time over the growing season. The WRSI is crop-specific, spatially explicit and identifies the start of the season based on rainfall patterns. It ranges between 0 and 100, with values below 50 showing levels of soil water well below the minimum. The WRSI has some limitations such as not accounting for excess water (e.g. floods), and the coarse resolution of some of its static inputs (e.g.: soil water holding capacity). Weather data and crop growth models reveal the high sensitivity of agricultural output to extreme weather in Mozambique. While data on agricultural yields at the province level is patchy, both spatially and across time, we look at the unconditional correlation between the WRSI for maize and their corresponding yields. The WRSI mean is calculated locally, at the grid cell level, then averaged over the province. The results provide suggestive evidence of a positive relation between water supply during the growing season and crop performance (Figure 3). More 10 specifically, a ten-unit reduction in the cumulative crop water requirement index is shown to reduce maize yields by 0.1 tons per hectare. Figure 3 also illustrates the existence of a positive relationship between yield anomalies (i.e.: deviations from the mean at the provincial level) and WRSI anomalies. Since the WRSI is capped at 100 and is not designed to measure excess water, the points located in the lower right of the graph might be revealing the negative effects of too much rain or flooding. Low values of the WRSI are observed with high frequency in Mozambique, and this may be indicative of recurrent crop failures in the areas in which precipitation levels fall well below the historical mean (Figure 4). Lower crop yields caused by rainfall shocks may reduce household incomes and consumption, and significantly affect parents’ ability to afford nutritional inputs for their young children. When confronted with floods or droughts, households may be forced to cut basic investments in the nutrition and human capital of their children. Hence, it is plausible that birth-year rainfall affects the early nutritional status of children. The underlying mechanism that causes the effects of rainfall shocks to persist over time is arguably their influence on critical endowments of affected individuals, such as their health during a crucial period of physical development (i.e. the nutritional environment in the womb and in the first year of life). Seeking to investigate this possibility in more detail, we employ the same identification strategy used above to explore the contemporaneous (i.e. short-term) impacts of the floods and droughts on the anthropometrics of children, more specifically on the height-for-age z-score, a strong predictor of height in adulthood. The sample is comprised of children aged 0 to 4 that were measured and weighted in the DHS 2011. The results show a strong relationship between rainfall anomalies in the first year of life and the height-for-age z-score (Table 3). For instance, shortly after the shocks occur, affected children are about 0.5 to 0.8 standard deviations (based on the preferred specifications that control for district fixed effects) smaller than the control children, whose mean is -1.89 standard deviations below the World Health Organization international reference group. The results are robust across specifications and subsamples, and are statistically significant in a quantitative sense. In addition, affected children underperform in terms of schooling indicators, and this is possibly mediated by the negative effects of extreme rainfall anomalies on their health as infants. 11 Reduced-form results of empirical models on school participation and attainment show that droughts reduce the likelihood of school-aged children to attend school regularly. Partly as result, affected children accumulate 0.2 fewer years of schooling, equivalent to a drop in their school attainment of 6% relative to the control mean (3.3 years) (Table 4). Similarly, data from the 2007 census allow us to distinguish individuals who accumulated some level of education from those who do not have any education (variable “Any Education” in the right panel of Table 4). Performing the analysis on this sample confirms the results. Affected individuals are 1.4 percentage points (2.8%) less likely to accumulate some education. Thinking about the result chain, it is unlikely that rainfall anomalies occurring around the time of birth exerted a direct impact on school enrollment, attendance, progress and attainment of the children because school participation for affected infants began 4 or 5 years after the shocks happened. A more plausible explanation is that the negative effects of droughts and floods on the schooling of affected individuals are mediated by the influence that these shocks exerted on their nutritional status as infants. The long-term human capital and productivity channel The nutritional deprivations suffered in infancy appear to have long-lasting consequences on the physical development of affected individuals. In trying to connect the pathways linking early life health shocks –caused by extreme weather– and wellbeing in adulthood, the analysis looks at the height of adults several decades after they experienced droughts and floods around the time of birth. While comprehensive data on adult anthropometrics are rarely available in household surveys, the DHS constitutes an exception. However, the DHS normally collects anthropometrics data only for women 15 to 40 years old, and not for men. This is the case for the DHS’s implemented in Mozambique. Another limitation of the DHS is the lack of information concerning the place and date of birth of adults in the sample. An attempt to circumvent this is made by restricting the sample to include only districts that exhibit permanent migration rates below 20%. To do so, the 1997 census is used to calculate the share of adults born in the same district of permanent residence (see Figure 5). Reduced-form results on height (measured as a percentage of an international reference group) for this sample are shown in columns 5 and 6 of Table 5.2 The 2The internal validity of these results may be compromised by the fact migration decisions are likely endogenous to weather shocks. However, a district-level regression of migration rates on weather shocks and district fixed effects does not reveal systematic differences in migration patterns between affected and non-affected villages. 12 results indicate that extreme rainfall events experienced during infancy, particularly droughts for the district fixed effect models, hamper the physical development of women. Affected women are around 0.5-0.7 centimeters shorter than the comparison unaffected women, on average. Suggestive evidence of positive returns to health in Mozambique is found, implying that uninsured weather shocks that worsen nutrition status also lead to lower productivity. The existence of a positive association between improved nutrition and increased productivity is often assumed and has been widely demonstrated in the empirical literature. Good health is particularly important in agricultural settings in developing countries, where the structure of employment requires strong physical development, particularly among men. To contextualize the negative effects of weather shocks on adults’ height, this study provides non-parametrical estimates of the unconditional relationship between height and school attainment and wealth for adult women in Mozambique (Figure 6). The results show a quasi-monotonic relationship between women’s height, human capital, and wealth accumulation. This suggests that the physical development lost due to extreme weather is likely to translate into lower productivity and, ultimately, lower welfare.3 5. Conclusions Mozambique’s economy is dominated by agriculture, and therefore farmers are largely exposed to unpredictable rainfall. In developing countries, agriculture can be a very risky activity: where financial and insurance systems are underdeveloped, farmers’ exposure to risk is higher. Since the agricultural sector constitutes the backbone of the Mozambican economy, accounting for a quarter of GDP and representing the main source of employment for the vast majority of the population, weather shocks tend to have devastating impacts on the livelihoods of Mozambicans. Mozambique is also one of the countries most affected by natural disasters, both on a regional and on a global scale, and this intensifies farmers’ vulnerability. Mozambique is frequently hit by severe droughts, cyclones, and floods. Rainfall data for Mozambique spanning several decades indicate that years with precipitations levels well below or above the historical mean are common 3 The relationship is estimated with much less precision for women above 175cms because of the low number of observations in this height range. 13 and spatially diverse. Farmers’ exposure to these risks is very high and is expected to increase in the future because of climate change. This paper empirically documents the negative effects of extreme rainfall anomalies around the time of birth on long-run individual well-being in Mozambique. Individuals exposed to extreme weather (particularly floods) during infancy are shown to be significantly less likely to participate in the labor market several decades later. More specifically, individuals who were in areas affected by a rainfall shock while in utero or during the first two years of life later exhibited a rate of participation in the labor market that was 6% lower than the control mean. Their households also recorded 14% lower per capita consumption than the median of the comparison households. Lower consumption per capita due to floods experienced early in life is also shown to increase the likelihood that households will fall into poverty by 9 percentage points, or around 18%. Overall, the evidence suggests that the effects of uninsured weather shocks can persist for several decades. Reduced agricultural output and its harmful effects on food security and nutrition are arguably the main mechanisms behind the negative impacts. Agricultural output is highly sensitive to extreme weather in Mozambique. The lower crop yields caused by rainfall shocks negatively affect household consumption, and limit parents’ ability to invest in adequate nutrition for their children at a critical stage of their development. This translates into below-average anthropometric outcomes of affected children, who are shown to be about 0.5 to 0.8 standard deviations smaller than children in the control group. In addition, the nutritional deprivations suffered during infancy have long-lasting consequences that extend to poorer schooling outcomes and lower adult welfare and productivity. School attendance of affected children is lower, and they end up accumulating 0.2 fewer years of schooling than other children, on average. Overall, affected individuals are 1.4 percentage points (2.8%) less likely to reach some level of education than unaffected individuals. These effects persist until adulthood: women affected by shocks in their early life are shown to remain around 0.5 to 0.7 centimeters shorter than other women, on average. Non-parametric evidence also suggests that their productivity is considerably lower than that of unaffected women. Policy actions aimed at accelerating poverty reduction in Mozambique should account for the inability of rural households to shield the well-being of their children from the negative effects of 14 weather shocks. As evidenced by the findings of this paper, the informal means available to most of the population affected by massive crop failures will smooth consumption only partially. In light of this, it appears that the objectives of reducing poverty and decreasing vulnerability to natural disasters can be more easily achieved if they are considered together. In particular, two broad policy areas require special attention. The first area of focus includes policies aiming to strengthen, ex-ante, the resilience of households. These are policies that make the whole agricultural system less sensitive to extreme weather, such as improving disaster preparedness and weather information systems, creating opportunities for income diversification away from agriculture, allowing for higher labor mobility, and promoting wider access to agricultural technologies, such as irrigation and drought- or flood-resistant seeds. A second area of focus encompasses policies aimed at protecting, ex-post, the basic well-being of poor individuals. These include, for example, food security programs and scalable social safety nets that can protect the poor and prevent them from falling deeper into poverty once the weather shocks have happened. 15 References Alderman, H., J.R. Behrman, V. Lavy, and R. Menon (2001). "Child Health and School Enrollment: A Longitudinal Analysis." Journal of Human Resources: 36 (1), 185-205. Alderman, H., J. Hoddinott, and B. Kinsey (2006). "Long Term Consequences of Early Childhood Malnutrition," Oxford Economic Papers: 58 (3), 450-474. Banerjee, A., E. Duflo, G. Postel-Vinay, and T. Watts (2007). "Long Run Health Impacts of Income Shocks: Wine and Phylloxera in 19th Century France." NBER Working Paper No. 12895, January Behrman, J.R. and M. Rosenzweig (2004). "Returns to Birthweight." Review of Economics and Statistics: 86 (2), 586-601. Calderon, M. C. (2007). “High Quality Nutrition in Childhood and Wages in Early Adulthood: A Two-Step Quantile Regression Approach from Guatemalan Workers”, mimeo, University of Pennsylvania. Case, A., A. Fertig and C. Paxson (2005). "The Lasting Impact of Childhood Health and Circumstance." Journal of Health Economics, March. Elo, I. T. and S. H. Preston (1992). "Effects of Early-Life Conditions on Adult Mortality: A Review." Population and Development Review: 13 (1), 77-100. Frankenberg, E., J.P. Smith, and D. Thomas (2003). "Economic Shocks, Wealth, and Welfare." Journal of Human Resources: 38(2), 280-321. Global Facility for Disaster Reduction and Recovery – GFDRR (2012). “Mozambique: Disaster Risk Financing and Insurance Country Note”. GFDRR. Hallegatte, Stephane, Mook Bangalore, Laura Bonzanigo,Marianne Fay, Tamaro Kane, Ulf Narloch, Julie Rozenberg, David Treguer, and Adrien Vogt-Schilb (2016). “Shock Waves: Managing the Impacts of Climate Change on Poverty. Climate Change and Development Series.” World Bank. doi:10.1596/978-1-4648-0673-5. Harris, I., Jones, P.D., Osborn, T.J. and Lister, D.H. (2014), “Updated high-resolution grids of monthly climatic observations – the CRU TS3.10 Dataset”, International Journal of Climatology, 34: 623–642. doi:10.1002/joc.3711 Hoddinott, J. and B. Kinsey (2001). "Child Growth in the Time of Drought." Oxford Bulletin of Economics and Statistics: 63 (4), 409-436. Index for Risk Management - INFORM (2016). “Index for Risk Management: Results 2016”. Inter-Agency Standing Committee Task Team for Preparedness and Resilience and the European Commission. 16 Maccini, Sharon and Dean Yang (2009). "Under the Weather: Health, Schooling, and Economic Consequences of Early-Life Rainfall”. American Economic Review, 99(3): 1006-26. Strauss, J. and D. Thomas (1998). "Health, Nutrition, and Economic Development." Journal of Economic Literature: 36 (2): 766-817. UN-Habitat, UEM (2015). “Safer Schools: Developing Guidelines on School Safety and Resilient School Building Codes”. UN-Habitat, Eduardo Mondlane University, Mozambique, World Bank, and Global Facility for Disaster Reduction and Recovery. World Bank (2013). “World Development Report 2014: Risk and Opportunity—Managing Risk for Development. Washington”. World Bank. doi: 10.1596/978-0-8213-9903–3. World Bank (2016). “Mozambique: A Systematic Country Diagnostic”. World Bank. 17 Figure 1. Rainfall anomalies nationally and across provinces (1950-2014) Source: World Bank staff calculations using CRU-TS data 18 Figure 2. Rainfall Anomalies across Districts Affected by Cyclones Eline and Judah in 2000 Source: World Bank staff calculations using CRU-TS data Figure 3. Relationship between the WRSI and Maize Yields across Provinces in Mozambique 1.6 1.8 1.4 1.6 Maize yield (Tn/ha) 1.2 y = 0.0101x - 0.1478 1 1.4 Yield anomalies 0.8 1.2 0.6 1 0.4 0.7 0.9 1.1 1.3 1.5 0.2 0.8 0 30 50 70 90 0.6 WRSI 0.4 WRSI anomalies Note: WRSI provincial average from posto levels measured on X-axis. Maize yields at the provincial level measure in tons per hectare on the Y-axis Provinces: Cabo Delgado, Gaza, Inhambane, Maputo, Manica, Nampula, Nassa, Sofala, Tete and Zambezia. Years: 2002, 2005, 2006, 2007 and 2008. Source: World Bank staff calculations using FEWS NET and FAO Agromaps 19 Figure 4. The Maize WRSI for Mozambique (2004-2015) Source: World Bank staff calculations using FEWS NET and FAO Agromaps 20 Figure 5. Migration rates by district (Census 1997) Note: the y-axis measures the proportion of people born in a district different from the district of residency Source: World Bank staff calculations using CRU-TS data Figure 6. Relationship between Height and School Attainment and Wealth for Women in Mozambique Note: Locally weighted scatterplot (lowess). Bandwidth = 1.94 for height-educational attainment function and 1.8 for height- wealth index function. Bootstrapped 95% confidence interval shown in gray. Source: World Bank staff calculations using data from the Mozambique DHS 2011 21 Table 1. Effects on Labor Force Participation Labor Participation 1 Labor Participation 2 All Rural All Rural (1) (2) (3) (4) (5) (6) (7) (8) Lack of rain utero 0.025 0.015 0.019 0.014 0.017 0.010 0.013 0.010 (0.034) (0.031) (0.034) (0.031) (0.034) (0.031) (0.034) (0.031) Lack of rain 1st year 0.044 0.040 0.036 0.036 0.037 0.035 0.031 0.032 (0.029) (0.027) (0.030) (0.027) (0.029) (0.027) (0.029) (0.028) Excess of rain utero -0.030** -0.041** -0.034** -0.040** -0.033** -0.040*** -0.036** -0.040*** (0.015) (0.016) (0.015) (0.016) (0.015) (0.015) (0.015) (0.015) Excess of rain 1st year 0.006 -0.012 0.005 -0.011 0.004 -0.008 0.004 -0.008 (0.015) (0.014) (0.014) (0.014) (0.014) (0.013) (0.013) (0.013) R-squared 0.093 0.139 0.090 0.127 0.100 0.132 0.098 0.126 Control Mean .657 .657 .659 .659 .667 .667 .669 .669 Observations 19,018 19,018 18,593 18,593 19,018 19,018 18,593 18,593 District FE no yes no yes no yes no yes Note: Robust standard errors in parenthesis clustered at the birthyear-district level. Labor Participation 1 = 1 if individual either worked in the past 7 days or did not work but had a job. Labor Participation 2 is equal to Labor Participation 1 but also include individuals that either worked in the past 7 days or had been looking for jobs in the past month. Rainfall shocks are defined as two standard deviations below (drought) or above (floods) the historical mean for the district. *** p<0.01, ** p<0.05, * p<0.1. Source: World Bank staff calculations using IOF-2008/09 Table 2. Effects on Expenditure Per Capita and Probability of Falling Below the Poverty Line Expenditure per capita Probability of being poor All Rural All Rural (1) (2) (3) (4) (5) (6) (7) (8) Lack of rain utero 0.177 -1.370 -0.173 -1.181 0.070 0.020 0.092 0.039 (4.693) (5.389) (4.799) (5.652) (0.101) (0.098) (0.101) (0.099) Lack of rain 1st year -0.245 -1.484 0.221 -2.107 0.069 -0.010 0.066 -0.009 (4.168) (4.895) (4.171) (4.889) (0.083) (0.077) (0.083) (0.077) Excess of rain utero -7.309* -5.017 -6.359* -4.742 -0.005 -0.027 -0.008 -0.028 (3.819) (4.968) (3.794) (4.975) (0.040) (0.038) (0.040) (0.038) Excess of rain 1st year -8.097*** -4.429* -7.045** -3.990 0.099** 0.093** 0.095** 0.092** (2.860) (2.693) (2.809) (2.674) (0.042) (0.039) (0.042) (0.039) R-squared 0.011 0.038 0.011 0.033 0.015 0.084 0.015 0.083 Control Mean 27.744 27.744 27.110 27.110 0.498 0.498 0.500 0.500 Observations 6,321 6,321 6,228 6,228 6,321 6,321 6,228 6,228 District FE no yes no yes no yes no yes Note: Robust standard errors in parenthesis clustered at the birthyear-district level. Rainfall shocks are defined as two standard deviations below (drought) or above (floods) the historical mean for the district. *** p<0.01, ** p<0.05, * p<0.1. Source: World Bank staff calculations using IOF-2008/09 22 Table 3. Effects on Children’s Anthropometrics Height for Age z-score All Rural (1) (2) (3) (4) Lack of rain utero -0.160 0.154 -0.054 0.174 (0.194) (0.176) (0.184) (0.176) Lack of rain 1st year -0.779*** -0.582*** -0.670*** -0.559*** (0.172) (0.165) (0.158) (0.165) Excess of rain utero -0.500 -0.449 -0.409 -0.447 (0.363) (0.286) (0.361) (0.286) Excess of rain 1st year -1.177** -0.818** -1.072** -0.805** (0.515) (0.363) (0.510) (0.362) R-squared 0.026 0.110 0.028 0.104 Control Mean -1.893 -1.893 -1.987 -1.987 Observations 5,620 5,620 5,056 5,056 District FE no yes no yes Note: Robust standard errors in parenthesis clustered at the birthyear-district level. Rainfall shocks are defined as two standard deviations below (drought) or above (floods) the historical mean for the district. *** p<0.01, ** p<0.05, * p<0.1. Source: World Bank staff calculations using IOF-2008/09 Table 4. Effects on School Participation and Attainment School Attendance Education Attainment Any Education All Rural All Rural All Rural (1) (2) (3) (4) (5) (6) Lack of rain utero -0.261*** -0.262*** -0.213 -0.201 -0.015* -0.014* (0.069) (0.069) (0.274) (0.277) (0.008) (0.008) Lack of rain 1st year 0.064 0.064 -0.466* -0.446* 0.011 0.013 (0.041) (0.041) (0.247) (0.252) (0.008) (0.008) Excess of rain utero -0.005 -0.003 0.220 0.220 0.007 0.009 (0.011) (0.011) (0.190) (0.190) (0.006) (0.006) Excess of rain 1st year -0.008 -0.009 0.057 0.047 0.008 0.010 (0.012) (0.012) (0.165) (0.167) (0.008) (0.008) R-squared 0.077 0.078 0.260 0.243 0.197 0.170 Control Mean 0.932 0.931 3.438 3.348 0.532 0.509 Observations 8,304 8,134 14,232 13,871 488,192 457,854 District FE yes yes yes yes yes yes Note: Robust standard errors in parenthesis clustered at the birthyear-district level. Rainfall shocks are defined as two standard deviations below (drought) or above (floods) the historical mean for the district. Any education is binary variable that takes a value 1 if the individual has one or more years of education, and zero otherwise. *** p<0.01, ** p<0.05, * p<0.1. Source: World Bank staff calculations using data from the IOF-2008/09 and the Population Census 2007 23 Table 5. Effects of Extreme Utero and Birth Year Rainfall on Women’s Adult Height Height for Age Adults All Rural Low Migration (1) (2) (3) (4) (5) (6) Lack of rain utero -0.338 -0.467 -0.299 -0.558 -0.333 -0.406 (0.544) (0.580) (0.588) (0.626) (0.437) (0.457) Lack of rain 1st year -0.287 -0.582* -0.205 -0.627* -0.332 -0.610** (0.330) (0.323) (0.341) (0.337) (0.306) (0.304) Excess of rain utero -0.522** -0.328 -0.524** -0.329 -0.483* -0.330 (0.250) (0.207) (0.264) (0.229) (0.259) (0.250) Excess of rain 1st year -0.196 -0.111 -0.175 -0.095 -0.137 -0.109 (0.293) (0.269) (0.306) (0.270) (0.298) (0.294) R-squared 0.012 0.105 0.012 0.105 0.012 0.097 Control Mean 94.772 94.772 94.703 94.703 94.708 94.708 Observations 7,445 7,445 6,954 6,954 7,305 7,097 District FE no yes no yes no yes Note: Robust standard errors in parenthesis clustered at the birthyear-district level. Rainfall shocks are defined as two standard deviations below (drought) or above (floods) the historical mean for the district. Height-for-age variable defined as the percentage of the height-for-age in a reference population.*** p<0.01, ** p<0.05, * p<0.1. Source: World Bank staff calculations using data from the Mozambique DHS 2011 24