Policy Research Working Paper 10823 Weather, Water, and Work Climatic Water Variability and Labor Market Outcomes in Sub-Saharan Africa Amjad M. Khan Landry Kuate Roland Pongou Fan Zhang Water Global Practice June 2024 Policy Research Working Paper 10823 Abstract Vulnerability to climate change and water scarcity is increas- average, and wet periods with an abundance of soil moisture ing globally. How this affects individual employment (not flooding) increase employment by 4 percentage points. outcomes is still not well understood. Using survey data col- The negative effects of dry shocks are larger in rural, poorer, lected from approximately half a million individuals across and agriculture-dependent areas and for individuals who Sub-Saharan Africa over from 2005 to 2018, this paper hold low-skilled jobs or work as farmers. Moreover, the examines the causal relationship between water availabil- paper finds that the burden of dry shocks disproportion- ity and labor market outcomes. It combines georeferenced ately falls on women, while the benefits of wet shocks accrue household survey data with a drought index that captures more to men. The presence of irrigation infrastructure and the exogenous effects of both rainfall and temperature on the historical evolution of local livelihood strategies—his- water availability. The findings suggest that extremely dry torical mode of subsistence—partly mediate the impacts periods decrease employment by 2.5 percentage points on of water shocks. This paper is a product of the Water Global Practice. 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://www.worldbank.org/prwp. The authors may be contacted at akhan31@worldbank.org, fzhang1@worldbank.org, lkuat097@uottawa.ca, and rpongou@uottawa.ca. 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 Weather, Water, and Work: Climatic Water Variability and Labor Market Outcomes in Sub-Saharan Africa ∗ Amjad M. Khan Landry Kuate† Roland Pongou‡ Fan Zhang§ JEL Codes: O12; O55; Q50. Keywords:: Climate change; Water scarcity; SPEI; Employment outcomes; Poverty traps; Sub-Saharan Africa. ∗ World Bank Group, Washington D.C., akhan31@worldbank.org † Department of Economics, University of Ottawa, lkuat097@uottawa.ca ‡ Department of Economics, University of Ottawa, rpongou@uottawa.ca § World Bank Group, Washington D.C., fzhang1@worldbank.org 1 Introduction Climate change will have significant and heterogeneous impacts on water availability throughout the world. In higher latitudes, historical data and climate models indicate that rainfall is likely to increase, while in lower latitudes overall precipitation is projected to decline. Additional warming caused by climate change will also increase overall rainfall variability, inducing longer periods of drought and increasing water stress as the agricultural, energy, and residential sectors extract more water from ground and surface reserves in order to combat increasing exposure to extreme heat and dryness (IPCC (2021)). Although there is a growing literature on the impacts of water variability on aggregate economic outcomes such as GDP (Damania et al. (2017); Russ (2020)), there has been limited work documenting the causal impact of water variability on labor market outcomes. This paper aims to quantitatively analyze the impacts of water shocks on individual employment outcomes in Sub-Saharan Africa. Farming is the livelihood of two-thirds of the population in the region. With a high dependence on rainfed agriculture and limited resources and investment for climate adaptation, this region is particularly susceptible to climate change. Notwithstanding, Sub-Saharan Africa is growing rapidly, with its population projected to double by 2050. This demographic trend puts enormous pressure on the labor market to create enough jobs to absorb the growing workforce. All of these factors underscore the importance of understanding and mitigating the risks brought about by climate-related water shocks on jobs in Sub-Saharan Africa. Our analysis relies on extensive, georeferenced individual socio-demographic and employ- ment information collected by the Demographic and Health Surveys (DHS) Program. The final dataset includes over half a million individuals from 16 countries across Sub-Saharan Africa over the period from 2005 to 2018. By using a series of pooled cross-sections of DHS surveys, this paper analyzes the respondent’s employment status at the time of the survey and exploit a rich econometric specification that controls for a full set of (subnational-level) region and time fixed effects. The georeferencing also allows us to merge the DHS data with other spatial data on cli- mate conditions, water availability, the presence of irrigation infrastructure, and historic ethnic groups. This paper measures water availability using the Standardized Precipitation-Evapotranspiration In- dex (SPEI), which was developed by Vicente-Serrano et al. (2010). The SPEI measures monthly variation in local climatic water balance and captures the influence of both precipi- tation and temperature on soil moisture—an important determinant of agricultural yield. Im- 2 portantly, agriculture is a key sector through which climate change affects the economy in the region. Water availability for agriculture is influenced by both precipitation and temperature because together they determine not only the amount of water available in surface runoff and soil mois- ture, but also the total losses to evapotranspiration from these sources. This paper finds that water availability determines employment opportunities. Our esti- mates show that exposure to a dry shock reduces the likelihood of having a job by 2.5 percentage points while exposure to a wet shock (though not necessarily floods) increases the likelihood of having a job by 4.2 percentage points. A location is exposed to a dry (wet) shock in a given month if the monthly SPEI in this location is 1.5 standard deviations below (above) its long-term mean. These effects are robust to controlling for individual demographic and socioe- conomic characteristics (such as gender, age, education, and place of residence), region (and grid cell) fixed effects, survey-year fixed effects, and the interaction of the latter two variables. The regional average masks considerable variation across people and places. In the second part of the analysis, this paper explores the distributional implications of water shocks by individual attributes and locations. It finds that the negative effect of water variability on employment is concentrated in rural areas, and among individuals holding low-skilled on-farm employment. These findings corroborate the notion that poorer rural populations, who rely more on agriculture and have fewer resources to invest in climate adaptation, tend to be more vulnerable to climate shocks. Analyzing heterogeneity by demographic characteristics, this paper finds that women disproportionately bear the burden of dry shocks, while men reap the benefits of wet shocks. Moreover, dry shocks reduce employment more among older adults. The presence of irrigation also plays an important role. The occurrence of dry and wet shocks impacts employment outcomes specifically in areas that are equipped for irrigation. Areas without any irrigation are not prone to employment impacts from these shocks. This reflects the fact that irrigation presence is co-determined with agricultural presence, and agricultural productivity is the channel through which climate wields an influence on employment outcomes. The availability of irrigation to provide water also brings increased populations and the planting of thirstier crops, amplifying rather than easing exposure to climate risks. Importantly, however, conditional on the presence of irrigation, the intensity of irrigation does not influence the sensitivity of employment to dry and wet shocks. Next, the paper examines how historical modes of subsistence mediate employment response to water shocks. It is well documented that livelihood strategies and institutions adapt over time, often over multiple generations, to deal with historic climate conditions as well as variabil- 3 ity (Agrawal, 2008; Giuliano and Nunn, 2021). Ethnic groups with differing historical livelihood strategies might demonstrate different adaptive capacity in the face of changing climate con- ditions. Societies that have historically relied on subsistence agriculture might have developed resilience capabilities that allow them to better address climatic challenges and reap the bene- fits of positive shocks. To test this hypothesis, the analysis combines data from the Ethno- graphic Atlas database developed by Murdock (Murdock, 1967) with our primary dataset. The paper analyzes how employment responses to water shocks vary for individuals living in ethnic homelands that have historically relied on (i) agriculture; (ii) hunting, fishing and gathering; or (iii) pastoralism (herding) for their livelihoods. This paper finds that relative to non-agricultural societies, regions where agriculture-focused ethnic groups historically resided experience a significant increase in employment in response to wet shocks, and experience worse effects from dry shocks. Homelands of ethnic groups that historically relied on hunting, fishing and gathering are also found to benefit from wet shocks but are less impacted by dry shocks (relative to other groups). Pastoral ethnic groups experience the largest negative effect from dry shocks while reaping no benefits from wet shocks. These findings suggest that local norms and institutions in historically agricultural regions have adapted well to allow workers to benefit from surplus agricultural productivity during periods of excess water but are also prone to the adverse impacts of dry periods on agricultural productivity. Pastoral ethnic groups have not developed a similar adaptation ability. These findings are consistent with studies showing that societies that experienced the Neolithic Revolution earlier have developed civilization and statehood earlier, and are wealthier and more educated today than pastoral societies (see, e.g., Diamond (1997); Hibbs Jr and Olsson (2004); Putterman (2008); Michalopoulos et al. (2019); Galor and Özak (2016)).1 Interestingly, hunter-fisher- gather regions have better adapted to avoid being worse off during dry periods than regions that historically relied on agriculture for livelihood. This finding is possibly driven by fishing- focused ethnic groups. Our results are consistent with Dalgaard et al. (2020), who find that societies that historically relied on fishing or enjoyed an abundance of marine resources are more developed today, and that these societies differ from purely agrarian societies in terms of culture, personality traits, and institutions. The paper also finds that jobs that are less vulnerable to water variability are more frequently found in historically hunter-fisher-gather ethnic homelands today, and that these regions have more diversified economies than historically agrarian regions. 1 The paper also finds that individuals living in historically agricultural ethnic homelands and homelands that derived most of their subsistence from hunting, fishing, and gathering are wealthier and more educated today. 4 Our paper contributes to the literature in several ways. Firstly, this article uses a measure of climatic water balance that is arguably useful for informing adaptation strategies in the face of climate change projection, that is the SPEI index. Instead of relying on rainfall shocks alone like much of the past research, the SPEI accounts for both temperature and precipitation variations. Secondly, unlike the previous studies which primarily focus on the effect of change in precipitation on the aggregate economic outcomes such as GDP (Dell et al., 2012; Brown et al., 2013), or on secondary outcomes such as conflict and migration that arise as a result of economic losses from climate variability (Harari and Ferrara, 2018; Henderson et al., 2017), this paper focuses on quantifying impacts on individual employment outcomes due to variation in water scarcity, an important intermediate economic outcome, and often a policy target of decision makers. It also documents the distributional impacts of water shocks, showing that their consequences are not evenly distributed across individuals and locations. Third, this paper highlights the economic consequences of water scarcity conditional on individual and regional characteristics, including historic modes of production and infrastructure availability. In doing so, the paper captures the role of historic adaptation patterns in determining the response to changes in weather conditions. The remainder of the paper proceeds as follows: Section 2 presents our conceptual frame- work; Section 3 describes the data and outlines the empirical strategy; Section 4 presents the main results. Section 5 describes analysis of the heterogeneous effects. Section 6 examines how irrigation infrastructure and institutions associated with traditional modes of subsistence influence adaptation to water variability. Section 7 concludes with a discussion of the findings. 2 Conceptual Framework There are several potential channels through which availability of water resources and water- related infrastructure can affect labor market outcomes. Firstly, positive impacts of water avail- ability on agricultural activity are well documented (Damania, 2020; Lesk et al., 2016). Better access to water sources including surface and groundwater improves yields (Sekhri, 2014) and agricultural wages (Mueller and Quisumbing, 2011), and narrows the gender wage gap (Maha- jan, 2017). It has also been shown that the adverse impacts of water scarcity on livelihoods can induce migration out of agricultural areas as rural populations seek to smooth out consumption risks (Henderson et al., 2017; Zaveri et al., 2020; Kleemans and Magruder, 2018). Large enough reductions in agricultural productivity caused by poor weather conditions can also give rise 5 to conflicts as the opportunity cost of engaging in conflict falls (Couttenier and Soubeyran, 2014; Harari and Ferrara, 2018; Hsiang et al., 2013; Acemoglu et al., 2020). Unemployment is an important economic outcome through which agricultural livelihoods are impacted, but few studies have examined the impacts of water shocks on jobs. This study attempts to fill this gap. Secondly, climate variation can also influence reliable access to adequate quantities of safe water for domestic and industrial use. This can impact industry and firm productivity (Islam and Hyland, 2019; Islam, 2019). In addition, water scarcity can have adverse impacts on labor productivity due to the adverse effects on nutrition and health of adults and children (Pruss- Ustun et al., 2008; Pongou et al., 2006; Hunter et al., 2010; Alsan and Goldin, 2019). Much of the previous literature focuses on the impacts on human capital accumulation of children, i.e. childhood exposure to adverse shocks during formative years can led to lower productivity when they grow into adults (Maccini and Yang, 2009; Hyland and Russ, 2019; Damania, 2020). But contemporaneous productivity losses can also arise during times of water scarcity if household members had to forgo productive working hours to fetch water from farther distances or to provide care for those who are sick due to inadequate water availability (Koolwal and Van de Walle, 2013). To examine such links, our study documents the impacts of dry shocks on both children and adults. Economic and infrastructure development play an important role in determining the impact of climatic changes. Specifically, investments in irrigation can decouple the response of agriculture from local climate variability (Schlenker et al., 2005); and investment in water supply infrastructure can reduce exposure to water shortages through the conveyance of water from storage farther away (McDonald et al., 2014). In general, less developed areas are more suscepti- ble to climate-induced variation in water availability due to lack of investment in infrastructure. This paper examines the heterogeneous effects of water shocks by baseline economic and infrastructure development status. 3 Data and Identification Strategy This section describes the various sources of data compiled for the analyses and presents some descriptive statistics from our dataset. 6 3.1 Data Description The study relies on two main sources of data: (i) pooled data from the DHS datasets from almost all African countries, which provides data on individual labor outcomes as well as individual and household characteristics; and (ii) the SPEI database with information on monthly climatic water balance at the global scale with a 0.5 degrees spatial resolution. As mentioned, the SPEI departs from other climate databases in that it combines data on temperature with data on precipitation to measure aridity and monitor soil moisture and drought conditions. Demographic and Health Surveys The DHS provide information on respondents’ employment status, our main outcome of inter- est. These are repeated cross-section surveys collected in most developing and middle-income countries since the mid-1980s. The surveys are representative at the national and sub-national level and are comparable across countries and years for most variables. The DHS Program also provides georeferenced data (longitude and latitude) in selected surveys, making it possible to combine these surveys with a wide range of spatial data from other sources. For this analysis, the paper uses data on over half a million individuals from 16 countries across Sub-Saharan Africa over the period from 2005 to 2018. The surveys use a two-stage sampling technique, selecting clusters (or census enumeration zones) at the first stage and households at the second stage. The data also provide information on demographic and socioeconomic status (e.g., age, education, occupation, etc.), which is collected for all household members. In addition, selected sub-samples of respondents in the DHS data provide information on a range of other significant variables of interest, including health status. Lastly, the DHS data are collected over a year and the surveys report the month of the interview, allowing us to match with the corresponding climate variables for the given month. Since our main outcome of interest is the employment status at the time of the interview, this allows us to more accurately estimate the impacts of contemporaneous effect of variation in climatic water balance on employment. Global Standardised Precipitation-Evapotranspiration Index The paper uses the SPEI to measure water availability in a given location. The index was derived using the CRUTS3.0 dataset at a spatial resolution of 0.5 degree. The SPEI is a widely used drought index which captures the effect of both rainfall and temperature on the climatic water balance. Note that the SPEI is a normalized variable, expressed in units of standard deviation from the average climate conditions experienced within a given grid cell over the 7 entire period of data available, i.e. from 1901 to 2018. To measure impacts of unusually dry and wet periods, we identify grid cell-month-year observations where the SPEI is two standard deviations above and below its long term mean, respectively. The spatial distribution of the frequency of extreme wet and dry shocks is shown in figures 1 and 2. Figure 3 shows the share of grid cells that experienced dry and wet shocks over the entire region in a given year. Notice that the 2010 spike in dry shocks is linked to an unusually strong La Nina event which pulled moisture away from East Africa and led to large-scale drought in the Sahelian belt and Eastern Africa during that year.2 Additional Data Sources The geolocated dataset is matched into first subnational level administrative boundaries ob- tained for the Global Administrative Areas (GADM). This allows us to exploit a rich set of region fixed effects in our econometric specification to account for the location-specific char- acteristics that are shared by households surveyed in a particular administrative area. Data on the presence of irrigation are included to examine whether access to irrigation in a region affects responses of local labor markets to water shocks. To measure the share of each grid cell that is irrigated at the start of the analysis period, the paper uses gridded data on the average area equipped for irrigation between 1990 and 2005 from the UN Food and Agriculture Organization (FAO). The construction of this data set is described in Siebert et al. (2007). To conduct the analysis, these raw data are aggregated to 0.5x 0.5 degree by taking the average percentage of area irrigated in all smaller cells that fall within a larger grid cell. Figure 4 shows the distribution of areas equipped for irrigation in our sample. Additional datasets are included to examine the role of adaptation in mediating the impacts of water shock. The paper uses data on the precolonial mode of subsistence by ethnic home- land to examine whether the impacts of water shocks vary conditional on a region’s historical activities. To do so, information on individual food production strategy by ethnicity from the Ethnographic Atlas, originally compiled by Murdock (1967) are collected. The analysis identi- fies if an ethnic groups subsistence was reliant on either: (i) agriculture; (ii) hunting, gathering and fishing; or (iii) herding and animal husbandry (or pastoralism). Figure 5 illustrates the spatial distribution of the reliance on the three modes of subsistence across different ethnic homelands in sub-Saharan Africa.3 2 For discussion of the droughts and famine in 2010, see these articles from NASA and New York Times. 3 The data are originally in the form of GPS points. The analysis generates Voronoi polygons around these points and assign gridcells to the ethnic homelands of the Voronoi polygon their centroid falls within. 8 Spatial Matching of the Data Using collected data on the geographical coordinates, the analysis merges the DHS dataset with SPEI from 1980 to 2018. The merging consists of assigning to each individual the weather conditions of the grid cell in which their DHS cluster coordinates are located. This paper also uses similar spatial matching to identify the administrative region in which individuals resided at the time (year and month of interview) of the survey. Our identification strategy, as described in greater detail in the next section, exploits within region variation in the exposure of households to (wet or dry) water shocks. The estimated im- pacts of water shocks on labor outcomes compare "treated" households (those that experienced a wet or dry water shock) in clusters B and C with those in the "control" group (those that did not experience a water shock) in cluster D, averaged across the full sample of the dataset. The SPEI varies at the grid cell level, and is already normalized by the long-term average within a grid cell. Treatment thus represents exposure of households to unusually wet or dry conditions compared to their local historical climate. 3.2 Summary Statistics Our main sample consists of 580,592 individuals. The sample covers period from 2005 to 2018 for 16 countries of Sub-Sahara Africa. Table 1 summarizes the key variables of interest. In our sample, 64 % of individuals have a job. Of those who are employed, 17% have a high-skilled jobs, and about 23% are farming employees. The average age of individuals in the sample is 29 years, with relatively highest share in age group 15-24 (39%), following by people in 25-34 (29%). Individuals aged 45 years and above account only for 10 % of the sample. Regarding education, the average individual has 4.3 years of education, with the majority (roughly 60%) completing the secondary and about 31% with no education. Most individuals (61%) live in rural areas. Individuals who were surveyed experienced, on average, weather conditions resulting in an SPEI equal to −0.120, which reflects that on average they experienced more dry shocks than wet shocks. Table 2 reports the sample distribution regarding the ancestral mode of subsistence. Agri- culture ranks as the predominant mode of food production with 59% of the historical activities. Then husbandry represents 22%; and herding 19%. It is worth noting that communities used to rely on different activities in the same period. The paper uses this information to examine the adaptation over time in local activities and how these determine responses to short-run 9 fluctuations in weather conditions captured by the SPEI. 3.3 Identification Strategy Our baseline model is estimated using ordinary least squares, and is specified as follows. Labor_outcomesijt = β1 × 1(speiijt < −1.5) + β2 × 1(speiijt > 1.5) + θ′ Xijt + γj + δt + ηm + ϵijt (1) Where Labor_outcomesijt denotes the employment outcome of individual i in adminis- trative unit j in year t. In our main results, this is a dummy representing an individual’s employment status at the time of the survey. The specification also includes Xi a vector of individual-specific control variables for age, education and type of region (urban or rural). γj andδt are the grid-cell and year fixed effects. γj controls for time-invariant location-specific characteristics that may be correlated with SPEI deviations, while δt allows us to control for region-wide year-specific shocks such as a continent-wide drought. Additionally, the main equa- tion includes ηm as a month-of-year fixed effect to account for the seasonality of employment outcomes. The error term, ϵijt , represents unobserved idiosyncratic shocks. Standard errors at the administrative subdivision level, to account for serial correlation in the error term as well as to account for correlation between individuals live in the same administrative subdivision. Our main coefficients of interest are β1 and β2 , which can be interpreted as the change in the likelihood of having a job in response to a wet or dry shock. In line with related literature, we define positive and negative deviations of the SPEI above and below a set threshold to estimate the impact of wet and dry shocks respectively. Specifically, a dry (wet) shock is defined as a 1.5 standard deviation decrease (increase) in the aggregate water balance over a given month compared to a long-term mean for that month. We use monthly SPEI measure to measure overall water availability, for instance, as stored in surface reservoirs and soil moisture. The analysis uses alternative functional forms to test the robustness of our results.It directly employs the continuous measure of the SPEI, and also tests the sensitivity to varying thresholds. Additionally, to differentiate impacts of more extreme events from less extreme ones, we estimate 10 the following version of equation (1) using multiple cutoffs: Labor_outcomesijt = α1 × 1(speiijt < −2) + β1 × 1(−2 < speiijt < −1) + α2 × 1(speiijt > 2) + β2 × 1(1 < speiijt < 2) + θ′ Xijt + γj + δt + ηm + ϵijt (2) 4 Main Results 4.1 Estimating Impacts on Employment The main estimation results are presented in Table 3. Column 1 shows the results with no controls added. In the next three columns the specification includes a set of controls and examines the sensitivity of results to sequentially adding gridcell, year, and month-of-year fixed effects. Our estimates suggest that exposure to extreme wet and dry conditions weather conditions lead to differing effects of employment, conditional on whether it is a dry or wet shock. While dry shocks, or droughts, reduce the likelihood of having a job, wet shock increases the employment. In term of magnitude, a 1.5 standard-deviation decrease in water availability – a dry shock as measured by the SPEI – reduces the employment rate by 2.5 percentage points; by contrast, a 1.5 standard-deviation increase in water availability – a wet shock as measured by the SPEI – increases the employment rate by 4.0 percentage points. Table 4 shows the results from utilizing alternative functional forms of the SPEI measure. Columns 1 and 3 estimate the impact of wet and dry shocks using a cutoff thresholds of 1 and 2 respectively. Both cases find similar results as before, suggesting that dry shocks are associated with lower employment. The results on wet shocks are associated with higher employment, but the coefficients are smaller in magnitude and not statistically significant. Column 3 presents the results of estimating Equation 2. The direction of the coefficients is still similar to our main findings, while the magnitudes suggest that milder shocks (SPEI variation between 1 and 2 standard deviations) are associated with lower levels of employment change, while more extreme shocks result in higher changes in employment. The coefficient of extreme dry shocks in columns 3 and 4 is of much larger magnitude, suggesting non-linear effects of water availability on employment. Lastly, Column 5 shows the result of directly using the continuous value of the SPEI as a regressor. This gives a positive and statistically significant coefficient, again signalling that increased water availability is associated with higher employment levels. The qualitative impacts of water shocks on employment are consistent with global evidence 11 on the impacts of water shocks on GDP. Russ (2020), for instance focuses on the impact of deviations in surface runoff and finds negative water shocks reduce GDP growth while positive water shocks have the opposite impact. Our results are in line with these findings, except that we focus on the impacts on employment outcomes in particular, with local climatic water balance as our main explanatory variable. To account for variation in runoff patterns, we also control for slope of the location. 5 Heterogeneity of Impacts This section further examines the heterogeneous labor impacts of water shocks by character- istics that vary at the individual and regional levels. It begins by examining the variations in impacts across gender and age, two key attributes that are important for understanding the distributional implications of impacts. It then examines the role of local development, as mea- sured by the income level of the country in which individuals reside and by whether individuals live in urban or rural locations. 5.1 Individual Attributes Age group To analyze the differential employment effects of extreme weather conditions by age, the analysis categorizes individuals into specific age groups: 15-24, 25-34, 35-44 and 45-66 years old. The employment effect of exposure to weather shocks might differ by age, for instance, because of differences in occupational choices made by individuals in different age groups. Table 5 reports the estimated coefficients associated with the two different age groups. Our findings suggest that all adult workers over the age of 25 are affected by dry shocks, with impacts rising slightly with age. Wet shocks only impact the youngest age groups. This possibly reflects the need to employ younger workers during wet periods for processing agricultural surplus. Overall, the results are consistent with the main finding of this paper, that is, a positive weather shock in- creases the likelihood of having a job and a negative shock reduces it. The coefficient estimates are statistically significant. Importantly, the fact that the effect of water variability most significantly affects the labor outcomes of adult workers has clear implications for the welfare of individuals typically not in the labor force—the dependent population (that is, children and older adults). Owing to high fertility rates, the dependency ratio—the ratio of individuals not in the labor force (individuals 12 aged 0 to 14, and 65 and over) to individuals in the labor force (individuals aged 15 to 64)—is the highest in Sub-Saharan Africa. This means that the working population in this region is subject to high financial stress. Our findings indicate that water scarcity is likely to significantly add to this pressure. This has negative externalities on the well-being of the dependent population, most notably children. These negative effects, however, are likely to vary across and within countries given differential fertility rates and pace of the demographic transition (Canning et al. (2022), Okoye and Pongou (2022)). Gender The paper analyzes the differential effect of water shocks by gender. There are several potential reasons why the labor impact of water may be different for men and women. One reason could be differences in occupational choice. If women are more likely than men to be involved in activities that depend on water (e.g., agriculture), then any water shock (be it wet or dry) will impact them more. Also, if water scarcity leads to poor health outcomes, and if the care- giving burden of sick household members is greater for women as it has been argued (Koolwal and Van de Walle, 2013), then dry shocks are more likely to affect the labor productivity of women. Moreover, a recent report by WaterAid (2021) argues that the provision of a very basic amenity such as a community water pump or a well can free up the equivalent of 77 million working days per year that women and girls currently spend collecting water. Upgrading to a tap in every house would multiply these benefits, releasing 122 million working days’ worth of time annually. It is believed that this would significantly reduce unpaid domestic work and increase women’s and girls’ educational and livelihood options (Hyland and Russ, 2019). Finally, differences in biology could explain gender differences in the labor impact of dry shocks. It is well documented that males are biologically more fragile than females (Pongou (2013, 2015); Pongou et al. (2017)), which may imply that a dry shock can have a greater negative effect on the health and hence the economic productivity of males. Results of the analysis of the differential effects of water shocks on employment by gender are reported in Table 6. It suggests that, the negative effect of a dry shock appear to be greater for women than for men, while the benefits of a wet shock are greater for men. Indeed, a dry shock decreases the likelihood of having a job by 4.5 percentage points for women, and by 2.2 percentage points for men, with the latter effect not being statistically different from zero. In contrast, a wet shock increases the likelihood of having a job by 3.8 percentage points for women and by 4.1 percentage points for men. The finding that water scarcity hurts women to a greater 13 degree is consistent with the view that women are more likely to engage in agricultural activities, collect water for household use, and care for household members who are sick for reasons that may be linked to inadequate water availability. The finding that water availability benefits men more may reflect gender differences in crop production. It is possible that women are more involved in the production of thirstier crops, and that even wet shocks do not necessarily provide a sufficient amount of water for these crops to grow properly. These results also possibly reflect the need to employ men during wet periods for harvesting certain crops. These are hypotheses that can be explored further in future research. Place of Residence To understand how the effects vary by place of residence, the analysis of individuals living in rural and urban areas is conducted. The results are reported in columns (3) and (4) of table 6. The analysis finds that negative impacts of dry conditions are experienced only by individuals who live in rural areas. These results are consistent with prior expectations, given that rural economies are less diversified and heavily depend on agricultural production, thus prone to the occurrence of droughts. Also, rural areas are less likely to have adequate investments in infras- tructure to buffer them from extreme variability in water availability. Meanwhile, the results in column(4) show that employment in urban area tends to increase during both dry and wet shocks. While puzzling at first, this could reflect changes in rural-urban migration dynamics arising from low and high agricultural productivity induced during dry and wet periods, respec- tively. The results on dry shocks are consistent with findings in the previous literature, where lower agricultural productivity due to low water availability induces reallocation of labor to off-farm and urban employment opportunities (Blakeslee et al., 2020; Henderson et al., 2017). 5.2 Effects on Type of Employment To better understand the effects of exposure to weather shocks on the composition of employ- ment types, the sample is restricted to individuals who reported being employed and use the type of activity they engaged in as the dependent variable. Table 7 reports the point estimates of the impacts of SPEI shocks on the type of employment outcomes, conditional on the indi- vidual being employed. That is, the estimates in this table are obtained after restricting to the subsample of individuals who had a job then estimating the baseline equation expressed in equation 1 with the labour outcomes now reflecting whether an individual is employed: (i) in a low-skill job; (ii) as a hired worker on a farm; (iii) as a self-employed farmer. We conduct this 14 analysis separately for rural and urban populations, as farm employment outcomes are only relevant for the rural subsample. The results suggest that dry shocks reduce the share of low-skill workers in rural areas. Among these, a large impact negative impact is found for individuals employed as on-farm workers by other farmers, while no significant impact is observed on the share of individuals self-employed in farming activities. This reflects how low yields during dry conditions make it costly to employ on-farm workers. In contrast, during wet shocks, there is no significant effect observed on the type of employment in rural areas. Meanwhile, in urban areas, wet shocks are found to increase the share of low-skill workers. This is possible if agricultural surplus during the wet season induces a short-term boost to aggregate demand across an area, firms hire low-skill workers to meet higher output targets. 6 Infrastructure, Institutions and Climate Adaptation Our estimates of the economic impacts of water shocks depend on the adaptation measures that are employed as societies evolve in response to climate conditions they face in their envi- ronments. This section looks into the role played by two main forms of adaptation that can help manage the risks associated climate variability: (i) investments in infrastructure capacity, as evidenced by the presence (or absence) of irrigation infrastructure; and (ii) the develop- ment of behaviors and practices embodied in local institutions, i.e., the development of cultural institutions associated with different historical mode of subsistence. 6.1 Interactions with Irrigation Infrastructure This subsection examines how the presence of irrigation infrastructure determines the impacts of climate- induced variability in the water balance. For each grid cell, the percentage of area equipped for irrigation (AEI) is measured and uses to examine impacts for regions with and without any irrigation present. Additionally, within areas that have at least some irrigation, an interaction term with the variable AEI (both in level and the logarithmic form) is introduced into the regression, allowing the impact of water shock variables to vary with intensity of irrigation. Table 8 shows the results. Column 1 shows that dry and wet shocks do not have a significant impact on employment in regions where there is no irrigation present. The absence of irrigation in these regions may reflect a small role of agriculture in generating employment, thus making 15 employment robust to climate variability. Irrigation presence also brings increased populations and can lead to maladaptation in the form of the planting of thirsty crops. Column 2, on the other hand, shows statistically significant negative impacts of dry shocks and positive impacts of wet shocks for areas with irrigation – echoing our main results. This suggests that the agricultural dependence of such regions on irrigation makes employment sensitive to climate variability. Meanwhile, the interaction terms in columns 3 and 4 suggest that conditional on the presence of at least some irrigation, being equipped with more or less irrigation infrastructure does not influence the impact of dry and wet shocks on employment. Our results suggest that employment in areas with irrigation infrastructure is more vulner- able to the negative impacts of droughts than those with no irrigation infrastructure at all. These results are consistent with the Jevons paradox, whereby the availability of a technology (in this case, irrigation) that enhances the use efficiency of a natural resource (in this case, wa- ter) does not necessarily lead to a reduction in reliance on that resource. Instead, dependency on the resources may in fact increase as a behavioral response to the new technologies. This phenomenon has been well-documented in the case of irrigation, the presence of which induces shifts in cropping patterns toward thirstier crops; this leads employment to rely more heavily on water availability and thus increases exposure to the negative impacts of shocks to water availability (Damania et al., 2017). It is noteworthy, however, that once irrigation is provided (i.e. on the extensive margin), the degree to which an area is equipped for irrigation (i.e. on the intensive margin) does not have any significant impact on sensitivity to the occurrence of water shocks. This speaks to the importance institutions, agricultural choices and water management practices that accompany higher levels of irrigation in determining vulnerability to climate, once irrigation is available in a region. 6.2 Interaction with Pre-colonial Modes of Subsistence This subsection now examines how the effect of water shocks on employment is mediated by the pre-colonial mode of subsistence. This analysis aims to highlight possible adaptation to climate variability through the evolution of local cultural institutions and livelihood strategies. Both formal and informal cultural institutions evolve in the long term (across multiple generations) to account for environmental constraints (Giuliano and Nunn (2021)). For instance, histori- cally, herding (or pastoralism) is a common livelihood strategy for individuals in regions where climatic and environmental conditions make it less feasible to practice sedentary agriculture or to rely on hunting, fishing and gathering. Since the Industrial Revolution, many of these 16 regions have invested in agricultural development but, in the absence of appropriate institu- tional arrangements, agricultural employment in these regions may not be resilient to climate shocks. Additionally, climatic changes over recent decades may also have changed the adequacy of pre-colonial institutions that persist. Thus, it is expected the impact of weather shocks to be contingent on pre-colonial modes of subsistence. To examine this, data from ethnographic atlas are used to identify the primary mode of subsistence local ethnic groups relied on in the pre-industrial era. For each individual, a dummy variable is defined to indicate if they live in a region where ethnic groups historically relied on: (i) agricultural; (ii) hunting, fishing and gathering; or (iii) pastoralism or, in other words, herding. A location is considered as relying on a given mode of subsistence in the pre- industrial era if the share of the population that relied on that mode of subsistence is greater than the sample median. The climate shock variables are interacted with these dummy variables to examine how the impacts on employment vary with the predominant pre-colonial modes of subsistence. Table 9 reports the coefficient estimates of this analysis. In each column, the omitted group is the rest of the sample that does not belong to a homeland with the corresponding historic mode of subsistence. So, for instance, in Column 1, the interaction term shows the relative impact of water shocks on employment of individuals living in regions which historically relied on agriculture against those that did not. The results confirm the heterogeneity of impacts by historic mode of production. The coefficient on the wet and dry shock variables for the base category in Column 1 imply that non-agricultural ethnic homelands experience no impact of wet or dry shocks on employ- ment. The interaction coefficient suggests that, compared to this base category, agricultural homelands experience negative (positive) impacts during dry (wet) shocks, respectively. This difference is large in magnitude and statistically significant. The negative impact of droughts is to be expected if there is persistence in the agricultural- reliance of the economies in these regions till today. Meanwhile, local norms and institutions in historically agricultural regions have also adapted to allow workers to benefit from surplus agricultural productivity during periods of excess water, for example through the development of traditional flood-based agri- cultural systems widely practiced in riverine regions of Africa such as the Niger, Senegal and Volta river basins. On the other hand, Column 2 suggests that regions where the primary mode of subsistence was hunting, fishing and gathering are better off during the occurrence of both wet and dry shocks when compared to other regions. The impact of dry and wet shocks for the hunter-fisher- gatherer regions is statistically different, and the magnitude is large. In the case of dry shocks, it 17 is in the opposite direction from the base group and more than cancels out the negative impact of drought. For wet shocks, the interaction term is positive as well. This suggests that these regions have adapted to allow workers to benefit from increased agricultural productivity during wet spells, while they are also less exposed to negative employment impacts during dry periods. Lastly, in Column 3, we identify homelands where ethnic groups historically relied on herding or pastoralism. The coefficients on the interaction terms suggest that pastoral homelands are no worse off during wet periods, but periods of dry shock bring statistically significant and large reductions in employment in these regions, as compared to non-pastoral regions. This could reflect the evolution of cultural and institutional arrangements that allow farmers to leave local on-farm work during times of drought to participate in off-farm activities, such as pastoral herding. This would generate a reduction in employment for such regions, since such herding activity often entails transhumant pastoralism, where workers temporarily migrate out of their home regions in search of greener pastures for their livestock to graze on and are often not recorded as formally employed in their original locality. Additionally, it is worth noting that there is a positive correlation between irrigation pres- ence and pastoral homelands. This reflects the fact that irrigation investments are frequently required to sustain agriculture in arid regions, where pastoral systems were most likely to be adopted historically. Thus, the results on the interaction between pastoral regions and dry shocks presented in Column 3 should be interpreted in conjunction with the negative interac- tion between irrigation presence and dry shocks observed in previous results from Table 8. This suggests that the negative impacts of irrigation presence during dry shocks may be driven by, or could be be driving, the negative impact of dry shocks in pastoral homelands. Importantly, our findings that employment in ethnic groups that historically derived a large share of subsistence from agriculture benefits more from wet shocks than non-agricultural ethnic groups are consistent with studies showing that these societies are economically and institu- tionally more developed today. Several studies have shown that societies that transitioned to agriculture earlier have developed civilization, statehood, and science earlier (Diamond (1997); Hibbs Jr and Olsson (2004); Putterman (2008)), and they are also more collectivist (Olsson and Paik (2016)). Michalopoulos et al. (2019) also show that the descendants of agricultur- alists are more educated, cooperative, honest, and wealthier today than the descendants of pastoralists. Based on a natural experiment, Galor and Özak (2016) also find that exposure to pre-industrial agro-climatic characteristics which were conducive to agricultural investment has a positive effect on technological adoption, education, and savings today. 18 Our finding that societies that historically relied on hunting, gathering and fishing have better adapted to avoid being worse off during dry periods than other societies is possibly driven by fishing-focused ethnic groups. Dalgaard et al. (2020) find that societies that derived subsistence from fishing or enjoyed an abundance of marine resources are more developed today, and that these societies differ from purely agrarian societies in terms of culture, personality traits, and institutions. Additional explanations are provided for the higher resilience of historically agricultural so- cieties and societies that historically relied on hunting, gathering and fishing to negative water shocks. Their greater resiliency could be explained by the structure of their economy. Table 10 finds that jobs that are more frequently found in these societies are also those less likely to be affected by water shocks. For example, results suggest that these societies are less dependent on agriculture today, which is a finding consistent with the idea that individuals in these societies are more educated and live in wealthier households. 4 In addition, services jobs, sales jobs, and domestic jobs are found more frequently found in societies where subsistence was derived primarily from fishing, hunting, and gathering. These jobs are exactly those that suffer less from water variability or may even gain from dry shocks (see Table A.2). Economic diversification may also explain why historically agricultural societies and societies that historically relied on hunting, gathering and fishing are more resilient to negative water shocks. Table A.3 finds that exposure to economic (or employment) diversification (calculated using the Herfindahl index on the different employment categories listed in Table 10) for working individuals is greater in these societies compared to pastoral societies. In addition to fostering economic growth, economic diversification increases the opportunity to move out of a sector experiencing a negative employment shock to find a job in a different sector, which minimizes the overall unemployment rate in the economy. All these findings suggest that societies that historically relied on agriculture and those that relied on hunting, fishing and gathering are better able to adapt to, or mitigate, climatic shocks, explaining the fact that employment in those societies is not affected by drought. 4 In Appendix Table A.2, we estimate the effect of the historical mode of subsistence on individual education and household wealth, and find that individuals living in historically agricultural homelands and those living in homelands that historically relied on fishing, hunting, and gathering are more educated and wealthier today than individuals living in historically pastoral homelands. These findings complement those of Michalopoulos et al. (2019) as the latter study instead examines the impact of ancestral economic culture. 19 7 Conclusion This paper analyzes the average and heterogeneous impacts of water variability on employment outcomes in Sub-Saharan Africa. Our main finding demonstrates the causal connection between water and jobs opportunities. While a 1.5 standard-deviation dry shock reduces the probability of holding a job by 2.5 percentage points; a wet shock of the same magnitude raises the prob- ability of holding a job by more than 4 percentage points. The effects of water availability on employment depends on the nature of one’s occupation. The negative impacts of water scarcity primarily affect low-skilled on-farm workers in the agriculture sector. By contrast, agricultural workers who are self-employed are not affected. The findings suggest that low-skilled agricul- tural workers in rural areas are most vulnerable to the impacts of climate change, and that they should be a main target of climate adaptation policies and interventions. The findings from the heterogeneity analyses indicate that the adverse impacts of a dry shock primarily impact individuals between the ages of 25 and 65, while the positive effects of a wet shocks are found to be concentrated among younger workers. Our analysis also reveals gender disparities in the labor effects of water shocks. Specifically, it is observed that adverse impacts of unfavorable weather shocks have a greater magnitude on females, while the advantages of favorable weather shocks are similar for males and females. Overall, our results show that the poorest bear the heaviest burden of water shocks. The greater level of impacts that water shocks have on poor and rural populations largely stems from their heavy dependence on the agriculture sector for employment and lack of adaptive capacities for climate change. The impacts are concentrated among adults of prime working and reproductive age (ages 18 to 45), leaving both these adults and their dependents vulnerable to climate risks. Water scarcity may consequently negatively impact health outcomes of not only adults, but also of their children – potentially giving rise to intergenerational poverty traps that lock populations and regions into a vicious cycle of poverty. The paper also examines the effects of infrastructure development in mediating the impacts of water shocks. The effect of wet and dry shocks is found to be concentrated in areas equipped with at least some irrigation, reflecting vulnerability to climate induced by agricultural reliance. In this case, the availability of irrigation to provide water also brings increased populations and could have prompted the planting of thirstier crops, amplifying rather than easing exposure to climate risks. This observation aligns with the Jevons paradox, wherein efficiency improvement can paradoxically result in increased resource consumption instead of anticipated decreases. However, the intensity of irrigation conditional on irrigation presence does not influence the impacts of 20 dry and wet shocks. Given the large degree of heterogeneity in the characteristics of irriga- tion schemes, an important direction for future research is to document additional evidence on how climate vulnerability is influenced by the institutions, crop choices and water management practices that accompany irrigation. This paper also shows that the historical evolution of livelihood strategies affects the employment effect of water scarcity. Historically agricultural societies have been more sensitive to both wet and dry shocks, as compared to non-agricultural ethnic groups. Also, societies that historically relied on fishing, hunting, and gathering are not affected by dry shocks and benefit from wet shocks. Moreover, pastoral societies are more ad- versely affected by dry shocks than non-pastoral ethnic groups. The findings thus suggest that regions that were homelands of pastoral ethnic groups have not developed livelihood strategies and institutions to allow them to benefit more from wet shocks and respond to the economic challenges posed by changing climate conditions. These findings are consistent with studies showing that societies that experienced the Neolithic Revolution earlier are more institution- ally and economically developed today. Indeed, the analysis shows that individuals living in historically agricultural ethnic homelands or in locations that derived subsistence from fishing, hunting, and gathering are more educated, wealthier, and more exposed to economic diversifi- cation in the present. Our findings provide critical information for the design of policies and investments for man- aging water variability. The results suggest the key roles that infrastruc- ture investment and institutional adaptation play in mediating the impacts of, and building resilience to, water risks associated with climate change. Importantly, while irrigation invest- ments generate significant employment opportunities, they must be accompanied by improved management practices and better monitoring systems for irrigated agriculture to better address the labor-related impacts of water scarcity. The findings also provide key information for the design of disaster risk- management strategies, such as social protections and insurance pro- grams aimed at mitigating the deprivations associated with water risks. Investment aimed at protecting adult and child health can help avert the adverse effects of inadequate water access on human capital and labor productivity. Further, our analysis suggests that policies designed to address the economic challenges of water scarcity should be context specific and that they should also account for differences in occupations, historical social and economic institutions, and livelihood practices. 21 8 Figures and Tables Figure 1: Wet shocks between 2005 and 2018 Notes: The figure above shows the number of wet shocks experienced in each gridcell between 2005 and 2018, measured in terms of monthly SPEI deviations of more than 2 standard deviations above it’s long term mean. 22 Figure 2: Dry shocks between 2005 and 2018 Notes: The figure above shows the number of dry shocks experienced in each gridcell between 2005 and 2018, measured in terms of monthly SPEI deviations of more than 2 standard deviations below it’s long term mean. 23 Figure 3: Percentage of people exposed to dry and wet shocks Notes: The figure above shows the trend of percentage of gridcell-month observations exposed to dry and wet shocks in each year. We define a dry (wet) shock as corresponding to a monthly SPEI value lower than −2 (greater than 2). 24 Figure 4: Percent of area equipped for irrigation from FAO, circa 2000 Notes: The figure above shows the spatial distribution of irrigation presence around the year 2000, as estimated by FAO. 25 Figure 5: Pre-colonial mode of dependence, from data on ethnic homelands 26 Notes: The figure above shows the spatial distribution of three main modes of production among pre-colonial ethnic homelands on the African continent, derived from data on ethnic homelands. 27 Table 1: Summary Statistics Statistic N Mean St. Dev. Labor outcomes current employment 580,592 0.641 0.4779 if employed: low skill 536,743 0.834 0.371 farming self-employment 536,743 0.069 0.253 farming employee 536,743 0.230 0.421 Water availability SPEI (in levels) 580,592 −0.120 0.785 wet shock (SPEI > 1.5) 580,592 0.271 0.162 wet shock (SPEI > 2) 580,592 0.010 0.101 wet shock (SPEI > 1) 580,592 0.077 0.267 wet shock (1 < SPEI < 2) 580,592 0.066 0.249 dry shock (SPEI < −1.5) 580,592 0.040 0.197 dry shock (SPEI < −2) 580,592 0.006 0.083 dry shock (SPEI < −1) 580,592 0.132 0.339 dry shock (−2 1.5) 0.0476*** 0.0433*** 0.0562** 0.0430* 0.0426 (0.0166) (0.0117) (0.0245) (0.0260) (0.0266) Controls N Y Y Y Y Gridcell FE N N Y Y Y Year FE N N N Y Y Month FE N N N N Y Observations 580,592 572,907 572,907 572,907 572,907 R2 0.022 0.133 0.231 0.234 0.234 Note: ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01. This table reports a set of estimates building on the baseline specification expressed in Equation 1. Our main regressor are dummies indicating wet and dry shocks, considering deviation of the SPEI above and below a threshold of 1.5 and -1.5, respectively. Column (5) is our preferred specification in which controls including slope, individual age and years of education and for grid cell, year and month-of-year fixed effect are added. Standard errors clustered at the grid cell-month-year level are reported in parenthesis. 30 Table 4: Robustness Checks — Alternative Functional Forms Dependent variable: current employment (1) (2) (3) (4) (5) Alternative Cutoffs: SPEI<-1 -0.0172** (0.00773) SPEI>1 0.00742 (0.0142) SPEI<-1.5 -0.0251* (0.0135) SPEI>1.5 0.0426 (0.0266) SPEI<-2 -0.0693* -0.0708* (0.0385) (0.0387) -22 0.0235 0.0200 (0.0394) (0.0397) Continuous measure of SPEI: SPEI 0.00801* (0.00434) Observations 572,907 572,907 572,907 572,907 572,907 R2 0.234 0.234 0.234 0.234 0.234 Note: ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01. This table reports a set of estimates obtained using our preferred specification in column 5 of Table 3 and varying the functional form of the the SPEI, our main regressor of interest. Our preferred specification controls for slope, individual age and years of education and for interaction between administration subdivision and year fixed effect. Standard errors clustered at the grid cell-month-year level are reported in parenthesis. 31 Table 5: Heterogeneity by individual characteristics —- Effects by age group Dependent variable: current employment Age Group: 15 - 24 25 - 34 35 - 44 45 and above (1) (2) (3) (4) Dry shock (SPEI < −1.5) -0.00332 -0.0319** -0.0400*** -0.0558*** (0.0182) (0.0146) (0.0152) (0.0154) Wet shock (SPEI > 1.5) 0.0643* 0.0374 0.0256 0.0430 (0.0341) (0.0279) (0.0310) (0.0310) Controls Yes Yes Yes Yes Full set of FEs Yes Yes Yes Yes Observations 227,183 173,653 119,491 60,040 R2 0.167 0.200 0.221 0.224 Note: ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01. This table reports the coefficient estimates of the impacts of extreme weather conditions (wet and drought shock) on the probability of having a job for subsamples of different age groups. Individuals aged 15-24, 25-34, 35-33 and more than 45 years old are considered. The standard errors are reported in brackets. Standard errors clustered at the grid cell-month-year level are reported in parenthesis. 32 Table 6: Heterogeneity by gender and region Dependent variable: current employment Females Males Rural Urban (1) (2) (3) (4) Dry shock (SPEI < −1.5) -0.0450*** -0.0227 -0.0749*** 0.0506*** (0.0170) (0.0139) (0.0173) (0.0156) Wet shock (SPEI > 1.5) 0.0387 0.0413** -0.0177 0.0710*** (0.0409) (0.0195) (0.0292) (0.0246) Controls Yes Yes Yes Yes Full set of FEs Yes Yes Yes Yes Observations 416,785 156,113 351,180 221,726 R2 0.227 0.279 0.258 0.234 Note: ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01. This table reports the coefficient estimates of the impacts of extreme weather conditions (wet and drought shock) on the probability of having a job conditional on restricted samples based on age group and gender. Two age group, 18-45 and individuals aged more than 45 years old are considered. Standard errors clustered at the grid cell-month-year level are reported in parenthesis. 33 Table 7: Effects on employment composition (restricting to employed individuals only) Dependent variable: Rural Urban Hired Farm Self-Employed Low-skill Worker Farmer Low-skill (1) (2) (3) (4) Dry shock (SPEI < −1.5) -0.0465*** -0.146*** 0.0201 -0.0250 (0.0117) (0.0155) (0.0213) (0.0215) Wet shock (SPEI > 1.5) -0.0104 -0.0247 -0.0874 0.0538** (0.0196) (0.0290) (0.0599) (0.0254) Controls Yes Yes Yes Yes Full set of FEs Yes Yes Yes Yes Observations 224,233 224,233 224,233 131,354 R2 0.171 0.508 0.409 0.128 Note: ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01. This table reports the coefficient estimates on the type of employment outcomes attaines by individuals for rural and urban areas, using our preferred specification but restricting to individuals who reported being employed. These outcomes include whether an individual (i) has a low-skill job; (ii) is employed are hired labor on a farm;(iii) is self-employed in farming activities. Standard errors clustered at the grid cell- month-year level are reported in parenthesis. 34 Table 8: Differential effects of SPEI shocks by level of irrigation Dependent variable: current employment Gridcell Area Equipped for Irrigation (AEI): AEI = 0 AEI>0 (1) (2) (3) (4) Dry shock -0.0144 -0.0323** -0.0298* -0.0251 (0.0226) (0.0164) (0.0179) (0.0219) ... × % Area Equipped for Irrigation -0.00115 (0.00477) ... × ln(% Area Equipped for Irrigation) -0.0131 (0.0246) Wet shock 0.00446 0.0770** 0.0586 0.0725 (0.0399) (0.0318) (0.0396) (0.0476) ... × % Area Equipped for Irrigation -0.00115 (0.0257) ... × ln(% Area Equipped for Irrigation) 0.0110 (0.0767) Controls Yes Yes Yes Yes Full set of FEs Yes Yes Yes Yes Observations 137,441 435,466 435,466 435,466 R2 0.242 0.232 0.232 0.232 Note: ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01. This table reports the how the effects of the occurrence of extreme weather conditions (wet and drought shock) vary by the the presence of irrigation. Column 1 shows the effects in regions without irrigation, column 2 - 4 shows the impacts in regions where irrigation is present. Standard errors clustered at the gridcell-month-year level are reported in parenthesis. 35 Table 9: Effects of SPEI shocks by Pre-colonial mode of subsistence Dependent variable: current employment cutoffs=2 (1) (2) (3) Dry shock (SPEI < −1.5) -0.00454 -0.0949*** 0.0379** (0.0155) (0.0186) (0.0153) ... × Dummy_agri -0.0655** (0.0281) ... × Dummy_fish_hunt_gat 0.152*** (0.0231) ... × Dummy_pastoral -0.128*** (0.0238) Wet shock (SPEI > 1.5) 0.0259 -0.00693 0.0382 (0.0308) (0.0304) (0.0282) ... × Dummy_agri 0.0956* (0.0503) ... × Dummy_fish_hunt_gat 0.0853* (0.0460) ... × Dummy_pastoral 0.0217 (0.0556) Controls Yes Yes Yes Full set of FEs Yes Yes Yes Observations 570,873 570,873 570,873 R2 0.234 0.235 0.235 Note: ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01. This table reports results on how the impact of wet and dry shocks varies with the historical mode of subsistence of the pre-colonial ethnic group that resided in a region. Standard errors clustered at the grid cell-month-year level are reported in parenthesis. 36 Table 10: Occupation type and historic modes of production (1) (2) (3) (4) (5) manager services sales agri domestic agriculture −0.0039*** −0.00189*** 0.0519*** −0.01676*** −0.00074 (0.0009) (0.00014) (0.0016) (0.00147) (0.00099) fish/hunt/gath −0.0035*** 0.0024** 0.0277*** −0.0267*** 0.00502*** (0.0007) (0.0001) (0.0012) (0.0010) (0.00072) Controls Yes Yes Yes Yes Yes Full set of FEs Yes Yes Yes Yes Yes Observations 473,586 473,586 473,586 473,586 473,586 R2 0.1081 0.0206 0.055 0.1053 0.0131 Note: ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01. This table reports a set of estimates of the relationship between occupation and historic modes of production. It is now considered whether an indi- vidual works in a specific position or sector, including manager, services, sales, agriculture or domestic. All specifications control for DHSCLUST features including, population density, and whether it is in rural or urban areas and for interaction between administration subdivision and year fixed effect. Standard errors clustered at the administrative subdivision level are reported in parenthesis. 37 9 Appendix Table A.1: Countries and DHS Samples used in the anlaysis Country Year 2005-2009 2010-2014 2015-2019 Angola 2015 Benin 2011 2017 Burundi 2011 2018 Cameroon 2014 Comoros 2018 Cote d’ivoire 2008 Ethiopia 2012 Gabon 2014 Guinea 2006, 2009 2012 Kenya 2011 2015 Lesotho 2014 Madagascar 2005 Malawi 2015 Mali 2006 2014 Mozambique 2016 Nigeria 2013 Notes: This table reports the distribution of DHS sample used in the analysis by year and country. 38 Table A.2: Economic diversification and historic modes of production (1) (2) Education Wealth agriculture 0.2282*** 0.0024*** (0.0100) (0.00063) fish/hunt/gath 0.0607*** 0.0202*** (0.0075) (0.00017) Controls Yes Yes Full set of FEs Yes Yes Observations 508,180 508,180 R2 0.7883 0.7251 Note: ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01. This table reports a set of estimates of the relationship between years of education; wealth index; and urbanization; and historic modes of production. All specifications control for DHSCLUST features including, population density, and whether it is in rural or urban areas and for interaction between administration subdivision and year fixed effect. Standard errors clustered at the administrative subdivision level are reported in parenthesis. Table A.3: Index of economic diversification and historic modes of production (1) Index Individual (Working population) agriculture 0.0014*** (0.0006) fish/hunt/gath 0.00310*** (0.00005) Controls Yes Full set of FEs Yes Observations 508,180 R2 0.743 Note: ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01. This table reports a set of estimates of the relationship between index of economic diversification and historic modes of production. We look at how diversified are societies which used to rely on agriculture and fish/hunt/Gath as the modes of production compared to pastoral societies (the reference category). All specifications control for slope, individual age and years of education and for interaction between administration subdivision and year fixed effect. Standard errors clustered at the administrative subdivision level are reported in parenthesis. 39 References Acemoglu, D., Fergusson, L., and Johnson, S. (2020). Population and conflict. The Review of Economic Studies, 87(4):1565–1604. Agrawal, A. (2008). The role of local institutions in adaptation to climate change. Alsan, M. and Goldin, C. (2019). Watersheds in child mortality: The role of effective water and sewerage infrastructure, 1880–1920. Journal of Political Economy, 127(2):586–638. Blakeslee, D., Fishman, R., and Srinivasan, V. (2020). Way down in the hole: Adaptation to long-term water loss in rural india. American Economic Review, 110(1):200–224. Brown, C., Meeks, R., Ghile, Y., and Hunu, K. (2013). Is water security necessary? an empir- ical analysis of the effects of climate hazards on national-level economic growth. Philosoph- ical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 371(2002):20120416. Canning, D., Mabeu, M. C., and Pongou, R. (2022). Colonial origins and fertility: Can the market overcome history? Working Paper No. 2014, Stanford King Center on Global Devde- lopment. Couttenier, M. and Soubeyran, R. (2014). Drought and civil war in sub-saharan africa. The Economic Journal, 124(575):201–244. Dalgaard, C.-J., Knudsen, A. S. B., and Selaya, P. (2020). The bounty of the sea and long-run development. Journal of Economic Growth, 25:259–295. Damania, R. (2020). The economics of water scarcity and variability. Oxford Review of Eco- nomic Policy, 36(1):24–44. Damania, R., Desbureaux, S., Hyland, M., Islam, A., Rodella, A.-S., Russ, J., and Zaveri, E. (2017). Uncharted waters: The new economics of water scarcity and variability. World Bank Publications. Dell, M., Jones, B. F., and Olken, B. A. (2012). Temperature shocks and economic growth: Evidence from the last half century. American Economic Journal: Macroeconomics, 4(3):66– 95. 40 Diamond, J. (1997). Guns, Germs, and Steel: The Fates of Human Societies. & Company, Inc., 1999. New York: WW Norton. Galor, O. and Özak, Ö. (2016). The agricultural origins of time preference. American economic review, 106(10):3064–3103. Giuliano, P. and Nunn, N. (2021). Understanding cultural persistence and change. The Review of Economic Studies, 88(4):1541–1581. Harari, M. and Ferrara, E. L. (2018). Conflict, climate, and cells: a disaggregated analysis. Review of Economics and Statistics, 100(4):594–608. Henderson, J. V., Storeygard, A., and Deichmann, U. (2017). Has climate change driven urbanization in africa? Journal of development economics, 124:60–82. Hibbs Jr, D. A. and Olsson, O. (2004). Geography, biogeography, and why some countries are rich and others are poor. Proceedings of the national Academy of sciences, 101(10):3715–3720. Hsiang, S. M., Burke, M., and Miguel, E. (2013). Quantifying the influence of climate on human conflict. Science, 341(6151):1235367. Hunter, P. R., MacDonald, A. M., and Carter, R. C. (2010). Water supply and health. PLoS medicine, 7(11):e1000361. Hyland, M. and Russ, J. (2019). Water as destiny–the long-term impacts of drought in sub- saharan africa. World Development, 115:30–45. IPCC (2021). Climate change 2021: The physical science basis. Technical report, Intergovern- mental Panel on Climate Change. Islam, A. (2019). The burden of water shortages on informal firms. Land Economics, 95(1):91– 107. Islam, A. and Hyland, M. (2019). The drivers and impacts of water infrastructure reliability–a global analysis of manufacturing firms. Ecological Economics, 163:143–157. Kleemans, M. and Magruder, J. (2018). Labour market responses to immigration: Evidence from internal migration driven by weather shocks. The Economic Journal, 128(613):2032– 2065. 41 Koolwal, G. and Van de Walle, D. (2013). Access to water, women’s work, and child outcomes. Economic Development and Cultural Change, 61(2):369–405. Lesk, C., Rowhani, P., and Ramankutty, N. (2016). Influence of extreme weather disasters on global crop production. Nature, 529(7584):84–87. Maccini, S. and Yang, D. (2009). Under the weather: Health, schooling, and economic conse- quences of early-life rainfall. American Economic Review, 99(3):1006–26. Mahajan, K. (2017). Rainfall shocks and the gender wage gap: Evidence from indian agriculture. World Development, 91:156–172. McDonald, R. I., Weber, K., Padowski, J., Flörke, M., Schneider, C., Green, P. A., Gleeson, T., Eckman, S., Lehner, B., Balk, D., et al. (2014). Water on an urban planet: Urbanization and the reach of urban water infrastructure. Global environmental change, 27:96–105. Michalopoulos, S., Putterman, L., and Weil, D. N. (2019). The influence of ancestral lifeways on individual economic outcomes in sub-saharan africa. Journal of the European Economic Association, 17(4):1186–1231. Mueller, V. and Quisumbing, A. (2011). How resilient are labour markets to natural disasters? the case of the 1998 bangladesh flood. Journal of Development Studies, 47(12):1954–1971. Murdock, G. P. (1967). Ethnographic atlas: a summary. Ethnology, 6(2):109–236. Okoye, D. and Pongou, R. (2022). Missions, fertility transition, and the reversal of fortunes: Evidence from border discontinuities in the emirates of nigeria. Working Paper. Olsson, O. and Paik, C. (2016). Long-run cultural divergence: Evidence from the neolithic revolution. Journal of Development Economics, 122:197–213. Pongou, R. (2013). Why is infant mortality higher in boys than in girls? a new hypothesis based on preconception environment and evidence from a large sample of twins. Demography, 50(2):421–444. Pongou, R. (2015). Sex differences in early-age mortality: The preconception origins hypothesis. Demography, 52(6):2053–2056. Pongou, R., Ezzati, M., and Salomon, J. A. (2006). Household and community socioeconomic and environmental determinants of child nutritional status in cameroon. BMC public health, 6(1):1–19. 42 Pongou, R., Kuate Defo, B., and Tsala Dimbuene, Z. (2017). Excess male infant mortality: The gene-institution interactions. American Economic Review, 107(5):541–545. Pruss-Ustun, A., Organization, W. H., et al. (2008). Safer water, better health: costs, benefits and sustainability of interventions to protect and promote health. World Health Organization. Putterman, L. (2008). Agriculture, diffusion and development: Ripple effects of the neolithic revolution. Economica, 75(300):729–748. Russ, J. (2020). Water runoff and economic activity: The impact of water supply shocks on growth. Journal of Environmental Economics and Management, 101:102322. Schlenker, W., Hanemann, W. M., and Fisher, A. C. (2005). Will us agriculture really benefit from global warming? accounting for irrigation in the hedonic approach. American Economic Review, 95(1):395–406. Sekhri, S. (2014). Wells, water, and welfare: the impact of access to groundwater on rural poverty and conflict. American Economic Journal: Applied Economics, 6(3):76–102. Siebert, S., Doll, P., Feick, S., Hoogeveen, J., and Frenken, K. (2007). Global map of irrigation areas version 4.0. 1. Johann Wolfgang Goethe University, Frankfurt am Main, Germany/Food and Agriculture Organization of the United Nations, Rome, Italy. Vicente-Serrano, S. M., Beguería, S., and López-Moreno, J. I. (2010). A multiscalar drought index sensitive to global warming: the standardized precipitation evapotranspiration index. Journal of climate, 23(7):1696–1718. WaterAid, V. E. (2021). Mission-critical: Invest in water, sanitation and hygiene for a healthy and green economic recovery. Zaveri, E. D., Wrenn, D. H., and Fisher-Vanden, K. (2020). The impact of water access on short- term migration in rural india. Australian Journal of Agricultural and Resource Economics, 64(2):505–532. 43