Aspirations and weather shocks: Evidence from rural Zambia

Parlasca, Martin C.; Martini, Christina A.; Köster, Maximilian; Ibañez, Marcela Article — Published Version Aspirations and weather shocks: Evidence from rural Zambia Agricultural Economics Provided in Cooperation with: John Wiley & Sons Suggested Citation: Parlasca, Martin C.; Martini, Christina A.; Köster, Maximilian; Ibañez, Marcela (2024) : Aspirations and weather shocks: Evidence from rural Zambia, Agricultural Economics, ISSN 1574-0862, Wiley Periodicals, Inc., Hoboken, NJ, Vol. 55, Iss. 6, pp. 985-999, https://doi.org/10.1111/agec.12858 This Version is available at: https://hdl.handle.net/10419/313702 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Received: 11 May 2023 Revised: 4 July 2024 Accepted: 9 July 2024 DOI: 10.1111/agec.12858 ORIGINAL ARTICLE Aspirations and weather shocks: Evidence from rural Zambia Martin C. Parlasca1Christina A. Martini2Maximilian Köster3 Marcela Ibañez3 1Center for Development Research (ZEF), University of Bonn, Bonn, Germany 2German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, Leipzig University, Leipzig, Germany 3Göttingen University, Göttingen, Germany Correspondence Martin Christoph Parlasca, Center for Development Research (ZEF), University of Bonn, Germany. Email: [email protected] Funding information German Research Foundation, Grant/Award Number: CRC-TRR 228; “FutureRuralAfrica”, Project-ID, Grant/Award Number: 328966760; German Research Foundation, Grant/Award Number: FZT118,202548816 Abstract Aspirations defined as future-oriented desires or ambitions, can determine agricultural investments and rural development. Aspirations are shaped by people’s social, cultural, and physical environment and can be affected by external factors such as natural disasters. This article addresses the question of how weather shocks can influence individual and community aspirations. Using primary panel data from two survey rounds before and during a major drought in Zambia, we show that such extreme weather events can be associated with adverse impacts on individual aspirations. Further exploratory analyses suggest that aspirations towards assets that are particularly vulnerable to droughts are affected most. We do not find any significant effects of drought on community aspirations. KEYWORDS Africa, climate change, drought, water scarcity JEL CLASSIFICATION Q12, Q54, D84, O13, I31 1 INTRODUCTION Aspirations, which can be defined as forward-looking desires or ambitions to achieve a specific goal, are vital for future oriented behavior and are thus important ingredients for economic development (Beaman et al., 2012; Dalton et al., 2016;Ray,2006; Serneels & Dercon, 2021). If aspirations are too low, people are more likely to miss out on profitable investments in income generating activities, assets, education, or health. In agricultural contexts, aspirations have also surfaced as an influential force in several decisions directly related to farming, including fertilizer use, investments in livestock, technology adoption, farm renovations, or climate risk mitigation strategies (Cecchi This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2024 The Author(s). Agricultural Economics published by Wiley Periodicals LLC on behalf of International Association of Agricultural Economists. et al., 2022; Knapp et al., 2021; Mausch et al., 2018;TabeOjong et al., 2023). Aspirations can, therefore, represent critical internal constraints to personal and community growth and are behavioral drivers of many decisions on and off the farm (DeJaeghere et al., 2022;Genicot&Ray, 2020; Lybbert & Wydick, 2018; Nandi & Nedumaran, 2021). At the same time, increasing climate variation and extreme weather events threaten agricultural activities, income generation, and well-being of farmers all over the world. Especially in developing countries, weather shocks, including droughts, cause substantial damage to livelihoods that have been painstakingly built up over many years. As a result, prolonged drought periods have been shown to be associated with higher emotional distress and Agricultural Economics. 2024;55:985–999. wileyonlinelibrary.com/journal/agec 985 986 PARLASCA et al. anxiety among affected populations (Abunyewah et al., 2024; Crandon et al., 2022; Meierrieks, 2021; Zamani et al., 2006). This article sets out to investigate if prolonged periods of rainfall scarcity cause farmers to lower their aspirations. We thereby connect two separate strands of literature. On the one side, we build on the existing research on the psychological consequences of droughts. This strand of literature comes primarily from medical research in highincome countries and shows that climate change can erode mental health, cause anxiety and is even associated with crime and increased suicide rates (Daghagh Yazd et al., 2019;Dean&Stain,2010;Freund,2023;Heilmannetal., 2021; Hua et al., 2023; Jones, 2022; Mullins & White, 2019; Obradovich et al., 2018;Vinsetal.,2015). Farmers are particularly vulnerable to such effects due to their strong dependence on rainfall. On the other side, we also build on a small, but growing area of economic research linking people’s aspirations with exposure to natural disasters. Adverse effects of natural disasters on aspirations have recently been documented for floods in Pakistan (Kosec & Mo, 2017) and an invasive species in Kenya (Tabe-Ojong et al., 2021). In both of these examples, natural disasters lowered income and wealth, both in the present and future expectations, and thereby led to a reduction in the goals that people perceive as achievable. Motivated by these two, so far disconnected, strands of literature, we examine how the aspirations of farmers in Zambia were affected by a major drought that struck Southern Africa in late 2018 and early 2019. As in many sub-Saharan countries, droughts represent critical hazards to Zambia’s agrarian economy. The country’s average temperatures and climate variability have increased over the last three decades with severe consequences. The 2015/16 El Niño, for example, has been associated with a decrease in economic growth and negatively affected the productivity of rice and wheat (Alfani et al., 2021; Callahan & Mankin, 2023). The 2019 drought, analyzed in this article, was one of the worst and driest periods in Zambia since several decades (Hulsman et al., 2021). This article makes two main contributions to the literature. First, it is the first to analyze the critical relationship between drought and aspirations. Droughts are slow-onset and systemic shocks that affect approximately 55 million people each year, thereby representing the greatest threat to livestock and crops in almost all parts of the world (UNCCD, 2022). Even though the physical damages of droughts are well documented, adverse effects on aspirations, ambitions, and hope are under-researched. A better understanding of aspiration formation in this context is much needed, for designing policies and programs that try to safeguard vulnerable populations against such events and to understand the potential long-term consequences of droughts beyond the loss of physical assets. The second contribution of the article is a methodological differentiation between farmers’ aspirations for themselves (“individual aspirations”), as well as farmers’ aspirations for their communities (“community aspirations”) (see Martini et al., 2021). Most of the research on aspiration formation focuses on individual aspirations. We motivate the inclusion of community aspirations in this article by the fact that droughts are not idiosyncratic shocks that only affect specific households in a community and leave others unaffected, but rather systemic shocks that influence an entire community, region or even country at once. While extreme weather events can lead to increased competition for resources within a community and, therefore, increasing conflict (Hsiang et al., 2013; Papaioannou & Haas, 2017), extreme weather events can potentially also stifle compassion or a sense of belonging among members of a community, and thereby strengthen their bonds within the community (Fleming et al., 2014;Stephane,2021). Our results show that droughts reduce individual aspirations, particularly among those directly affected by drought-related losses. Aspirations related to arable farming and cow ownership are significantly more affected by drought compared to goat ownership or children’s education. However, no statistically significant effects of the drought on aspirations for the community were found. The remainder of this article is structured as follows. Section 2presents our conceptual framework, followed by a description of the data collection and empirical approach in Section 3. In Section 4, we present and discuss relevant outcomes. We conclude by summarizing the most important findings and their implications for future research and policies in Section 5. 2 CONCEPTUAL FRAMEWORK BasedonBernardetal.(2011), we define aspirations as desires or ambitions to achieve a specific goal. Aspirations differ from related concepts such as wishes, beliefs, or hope because (i) aspirations are future-oriented, meaning that they represent goals that can be achieved in the future but not immediately, and (ii) aspirations are motivators, meaning that an individual can and wants to put effort into achieving the goal which distinguishes aspirations, for example, from fatalism or having a mere wish (Bernard et al., 2011). Aspirations relate to a specific domain. In the past, these domains mostly involved individual aspects such as own wealth, social status, health, or education. In this article, however, we also consider people’s aspirations regarding their community. We employ the community aspiration index developed by Martini et al. (2021). To arrive at relevant dimensions PARLASCA et al. 987 and indicators for community aspirations, Martini et al. (2021) considered community well-being as a set of freedoms individuals can enjoy in their village and hence construct their community aspirations dimensions based on Sen’s conceptualization of development (Sen, 1999). Accordingly, community aspirations relate to social opportunities, economic facilities, protective security, political freedoms and transparency. The indicators for each dimension reflect the degree to which individuals aspire their community to have access to public goods such as education or police protection, but also to live in a community characterized by political freedoms and an improved standard of living. Because community welfare has a public good character, it is crucial to consider the aspirations that each individual has for the community rather than the aspirations of one single individual as the community leader. From a theoretical perspective, aspirations are assumed to be shaped by a person’s “cognitive neighborhood” (Genicot & Ray, 2020). By observing others, people are able to imagine what they could do and what they or their community could achieve (Appadurai, 2004; Genicot & Ray, 2017). Aspirations are, therefore, not only determined by a person’s past aspirations but also strongly depend on ongoing and anticipated social, economic, and cultural influences (Dilley et al., 2021; Genicot & Ray, 2017). In addition to the social component of aspiration formation, aspirations can be shaped by external influences, such as environmental shocks (Nandi & Nedumaran, 2021). When such external circumstances impede the achievement of specific goals, individuals may revise downwards the goals that seem attainable in the future. In one of the first studies to investigate the relationship between natural disasters and aspirations, Kosec and Mo (2017)analyze the effects of a flood in Pakistan on individual aspirations. The authors show that, due to the flood, aspirations fell by a similar magnitude to what would be experienced if household spending were slashed by 50 percent. Tabe-O et al. (2021) further show that farmers’ aspirations in Kenya are negatively affected by ecological stressors such as an invasive plant. Unlike previous studies, Tabe-O et al. (2021) also consider different dimensions of individual aspirations, revealing that the negative association is driven by reductions in aspirations towards future income and assets. The effect of a natural shock on individual aspirations can have both physical and mental components (Kosec &Mo,2017). Physically, natural shocks can have negative impacts on capital assets, people’s livelihood, and create economic barriers, such as worse infrastructure and market frictions, reduced access to education, markets, and health care. Thereby, people may perceive certain goals as more difficult to achieve which leads them to aspire less. Since natural shocks can cause people to see their hard work and investments being damaged or wasted, disasters can also affect aspirations through a mental component. People may perceive that they are losing control over their lives without clear paths forward, making them less willing to invest and, therefore, lowering their aspirations. Furthermore, Kosec and Mo (2017) hypothesize that a shock could also impact individuals’ community negatively, leading to conflicts and instability, which would reduce aspirations. In this article, we therefore aim to test the following two main hypotheses empirically: Hypothesis I: Droughts are negatively associated with people’s individual aspirations. Hypothesis II: Droughts are negatively associated with people’s community aspirations. Based on this, we expect that community aspirations are also affected negatively. However, it is possible that the effects differ across community aspiration dimensions. Effects of a drought on a community aspiration dimension that, for example, reflects social support within a village community may be particularly strong since such dimensions could suffer from political frictions or imbalances the most. For dimensions that look at more tangible, security-related aspects, such as the degree of police presence, it can be hypothesized that the potential effects of adroughtaresmallerbecausepeopleaspiretohavemore safety against future shocks. However, since research on community aspirations is so far extremely thin, it is not yet clear whether such differences actually exist. 3 MATERIALS AND METHODS 3.1 Study region To test our hypotheses, we use two rounds of survey data collected by the authors in the Southern Province in Zambia. The region has different micro-climates and is hence characterized by diverse agroecological conditions. Only few villages in the area have direct access to boreholes, ponds, or dams. Most households rely on rain-fed small-scale agriculture and are particularly vulnerable to variations in precipitation (Marcantonio, 2020). Interand intra-annual precipitation in the region varied substantially over the last 50 years (Hamududu & Ngoma, 2020). Even though the Southern Province has experienced several dry conditions, dry spells and unpredictable weather events over the past decades, the drought with El Niño-like conditions in 2018/2019 marked a low point as one of the driest seasons (Hulsman et al., 2021). The 988 PARLASCA et al. FIGURE 1 Study location. production of maize, which is the most important crop in the region, decreased by about 24% at the national level compared to the 5-year average (FAO, 2020). The fact that previous milder droughts already weakened households’ conditions further exacerbated the situation upto the point that approximately 192,000 people (10% of the population) in the province were in a crisis situation and 54,000 people (3% of the population) even in an emergency situation from October 2018 to March 2019 (ACAPS, 2019). 3.2 Survey and measurement of key variables We collected two rounds of survey data in 37 villages (see Figure 1.) The first round of data collection from August to September 2018 covered 643 households. For the second round in August/ September of 2019, we were able to collect follow-up survey data from 495 individuals. Although we made many attempts in the field to track and interview as many households as possible, attrition is still, unfortunately, relatively high. The attrition is not entirely unexpected. Seasonal and permanent migration are common coping and adaptation strategies to the effects of climate change in Zambia and other countries in sub-Saharan Africa (Nawrotzki & DeWaard, 2018). Farming households often migrate to more suitable farming areas or to alternative sources of income generation such as mining sites (Simatele & Simatele, 2015), which has played a large role in the difficulties of re-interviewing farmers. To understand possible biases that could arise from this, we therefore pay special attention to the issue during the analysis. For the measurements of individual aspirations, we follow the most common approach used in recent quantitative research to build an individual aspirations index (e.g., Bernard et al., 2014;Kosec&Mo,2017). Even though other ways to measure aspirations based on indirect and direct methods are able to yield decent quality data for single dimensions, to look at multiple dimensions at once, a summary index is currently the most-used option (Nandi & Nedumaran, 2021). Our index consists of four dimensions: a farmer’s number of cows, their number of goats, their plot size, and their children’s education. These aspects have been shown to closely resemble relevant individual aspiration dimensions for many African farmers (Rao et al., 2020). For each dimension, farmers reported their current status, their aspired status in 10 years, and their expected status in 10 years. Each dimension is first winsorized at a 95% level and then standardized using the 2018 baseline sample mean and standard deviation. The standardized measures are aggregated with equal weights. This gives us the individual aspirations index 𝐴𝑖: 𝐴𝑖= 4 ∑ 𝑑=1 (𝑎𝑖 𝑑−𝜇 𝑠 𝑑 𝜎𝑠 𝑑)(1) where 𝑎𝑖 𝑑is individual 𝑖’s aspiration in dimension 𝑑. Further, 𝜎𝑠 𝑑is the sample standard deviation and 𝜇𝑠 𝑑the sample mean for an aspiration in dimension 𝑑. For the community PARLASCA et al. 989 aspirations index, we follow Martini et al. (2021)andbuild an equal-weighted index consisting of different dimensions: number of community meetings, share of people in the community with suitable housing, school distance, security, share of households in the community receiving financial support, and financial contributions to village. We also winsorize each dimension at a 95 % level and standardize each dimension using the 2018 sample mean and standard deviation. We then aggregate all dimensions with equal weights to one index. The community aspirations index 𝐴𝑐can be formalized as: 𝐴𝑐= 6 ∑ 𝑑=1 (𝑎𝑖 𝑑−𝜇 𝑠 𝑑 𝜎𝑠 𝑑),(2) with 𝑎𝑖 𝑑for individual 𝑖’s aspiration in dimension 𝑑. Further, 𝜎𝑠 𝑑is the sample standard deviation and 𝜇𝑠 𝑑the sample mean for an aspiration in dimension 𝑑. For the rainfall measurements, we use TAMSAT (Tropical Applications of Meteorology using Satellite and ground-based observations) data, which offers rainfall estimates for Zambia with a spatial resolution of .0375◦latitudeby.0375 ◦longitude, covering about 4 km2. Rainfall data from TAMSAT have shown to be highly correlated with rain gauge data in Zambia (Libanda et al., 2020), which underlines the suitability for these data for our context. Since we do not have GPS coordinates of each individual farm, we use the coordinates of headmen interviews to measure rainfall at a village level. Since rainfall is likely to be very highly correlated within villages, we do not expect large losses of variation from this aggregation. Following a standard approach in the literature (Hendrix & Salehyan, 2012; Hidalgo et al., 2010;Kosec &Mo,2017), we estimate the absolute rainfall deviation from the 30-year mean for each village using a two-step procedure. In the first step, we take monthly rainfall totals for each of the five rainy season months (NovemberMarch) and standardize them using the 30-year mean (𝜇𝑣𝑚) and standard deviation (𝜎𝑣𝑚) of rainfall for village 𝑣in month 𝑚. We then take a sum over these five values. This can be formalized as: 𝑥𝑣,2019 = 5 ∑ 𝑚=1 (𝑥𝑣𝑚,2019 −𝜇 𝑣𝑚 𝜎𝑣𝑚 ),(3) In the second step, we standardize these deviations by the 30-year season mean (𝜇𝑣) and standard deviation (𝜎𝑣) of rainfall in village 𝑣. Lastly, since droughts represent a negative deviation of rainfall, we use the absolute value of the standardized deviation in rainfall to ease interpretation of the variable. A higher value of absolute rainfall deviation thus reflects a more intense drought situation. Our primary measure, the absolute value of rainfall deviations from the 30-year village season mean 𝑧𝑎𝑏𝑠, therefore, follows the formula: 𝑧𝑎𝑏𝑠 =|||| 𝑥𝑣,2019 −𝜇 𝑣 𝜎𝑣||||(4) Even though rainfall deviation is the most commonly used indicator for drought, we also consider alternative drought indicators, namely, the square of rainfall deviation as well as the Agricultural Stress Index (ASI) for cropland. The ASI is a tool that combines the Vegetation Health Index in the time and space dimension. Firstly, it calculates the intensity and duration of dry periods during the crop cycle at each location, using crop coefficients to account for how different crops respond to water stress. Secondly, it determines the spatial extent of drought by identifying the percentage of arable land with Vegetation Health Index values below a critical threshold of 35 percent (Kogan, 1995). Administrative areas are then classified based on the percentage of affected land. 3.3 Empirical approach Motivated by our interest in the effects of droughts on farmers’ aspirations, our empirical approach aims to isolate the impact of rainfall deviation on individual and community aspiration indices using multivariate models that control for potential confounders. Potential threats of bias coming from systematic measurement error or reverse causality are very low in our case, given the exogenous nature of rainfall and the rainfall measurement. To account for time-invariant unobserved heterogeneity among households, we apply a linear model with individual fixed effects, which can be expressed as follows: 𝑦𝑖𝑡 =𝛼 0+𝛽 ′ 1𝑅𝑖𝑡 +𝛽 ′ 2𝑋𝑖𝑡 +𝜔 𝑖+𝜖 𝑖𝑡,(5) where 𝑦𝑖𝑡 is either the individual aspiration index or the community aspiration index. Ris continuous variable capturing drought intensity. Xis a time variant set of individual characteristics. The parameter of interest is β1,since a negative and statistically significant coefficient would indicate that higher drought intensity is associated with lowered aspirations. Unobserved time invariant heterogeneity is captured through a fixed effect ωi.𝜖𝑖𝑡 is the individual error term. We calculate robust standard errors clustered at the village level. Since the data only cover two rounds and the variation of rainfall deviation in a given year across villages is quite low, our main specification does not include a dummy variable for the year 2019.Ide- 990 PARLASCA et al. ally, such a year dummy could be included to account for any other general changes that occurred in the region between 2018 and 2019 which may have affected aspirations apart from the drought, such as socio-economic developments. To address this possibility, we include two village level variables that measure structural changes in the region, namely, road quality as well as the share of households in each village with decent housing. However, we also show the results of a secondary specification, in which we include a year dummy. In this secondary specification, the rainfall deviation parameter does not anymore capture the large variation in rainfall between the year 2018 and 2019, but rather the much smaller variation across villages. To improve the precision of our estimates, we also adjust for a vector of time variant socio-economic characteristics X, namely, herd size, plot size, children’s education, and membership in health and savings groups. However, if some of these covariates lie on the causal pathways through which the drought affects aspirations, their inclusion into the model could lead to an underor overestimation of the true effect. Similarly, conditioning on socio-economic characteristics also creates possibilities of collider bias. For example, including farm size in the regression could introduce such bias if a reduction in farm size is both a result of the drought as well as of lowered aspirations. We, therefore, also present a model without any control variables. To explore a potential mechanism, we additionally estimate the following model: 𝑦𝑖𝑡 =𝛼 0+𝛾 0(𝑅𝑖𝑡|𝑆𝑖𝑡 =0)+𝛾 1(𝑅𝑖𝑡|𝑆𝑖𝑡 =1)+𝛽 ′ 2𝑋𝑖𝑡 +𝜔 𝑖+𝜖 𝑖𝑡, (6) where 𝑆𝑖𝑡 is an indicator variable that is one if a respondent experienced a loss in livestock or a crop failure 12 months prior to the data collection and zero otherwise. The comparison of size and statistical significance of 𝛾0and 𝛾1helps analyze whether the effect of the drought is more severe for respondents with agricultural losses compared to those who did not report any losses. We also conduct several robustness tests. First, we test how sensitive the rainfall deviation coefficient is with regard to the choice of control variables. To that end, we estimate the model shown in Equation (5) for all possible combinations of control variables. Point estimates and confidence intervals of the main parameter of interest, that is, rainfall deviation, for these alternative specifications are then summarized and illustrated in a specification curve. Furthermore, we estimate the relative degree of selection under proportional selection of observables and unobservables (Oster, 2019). Similar to Ruml and Parlasca (2022), we present the parameter δto indicate the extent of bias stemming from unobservable factors in comparison to observable ones that would be required to nullify the obtained results. Assuming proportional selection, greater values of δindicate more robust results concerning unobserved confounding variables. Oster (2019)suggestsa threshold value of 1. The parameter δis, however, also contingent upon the extent to which unobserved confounders are allowed to explain variability in the regression. This extent, which we denote as 𝑅2 𝑚𝑎𝑥, is calculated as the 𝑅2 of the regression with all control variables included, multiplied by a certain factor larger than one. Oster (2019) suggests a factor of 1.3, while a more conservative factor of 2.2 is also commonly used. We show results for the more conservative value of 𝑅2 𝑚𝑎𝑥. Attrition between the two survey rounds is substantial and requires particular attention. To better understand the potential threat of attrition bias, we follow two approaches. First, we use observable characteristics of the baseline to estimate Inverse Probability Weights giving higher weight to those observations that – based on observable baseline characteristics – were more likely to have attrited between survey rounds. The selection equation is a probit model with the following predictors: age, gender, education, marriage, mobile phone usage, locus of control, individual aspiration index, community aspiration index, rainfall deviation, distance to the next water source, education of the youngest child, herd size, plot size, membership in a health group, membership in a savings group, housing, and road quality in the village, and the time lived in the village. Second, we follow Özler et al. (2021) to estimate treatment effect bounds. To do so, we impute the outcome of missing observations with the round-specific mean plus or minus the round specific standard deviation multiplied by the factors .1, .2, .3, and .4. Since the drought variables are estimated at the village level, we do not need to impute the treatment variables of attriters. 4RESULTS 4.1 Descriptive overview Table 1provides basic summary statistics for the balanced sample of 495 households in 2018 and 2019. About half of all individuals are female and the average age is around 44 years. Most households engage in crop farming, but keeping livestock is also common. The descriptive statistics also demonstrate the severity of the 2019 drought. While the sample average absolute rainfall deviation in 2018 was around .75, it jumped to 6.02 in 2019. Due to the drought, 79% of the sampled farmers reported a loss of livestock or crop failure. Table 1also shows that individual aspirations decreased significantly from 2018 to 2019. The reduction PARLASCA et al. 991 TABLE 1 Summary statistics. 2018 2019 Mean SD Mean SD Difference Female (1 =yes) .51 .51 Age (years) 44.05 14.93 44.10 14.96 Education (years) 7.15 2.77 7.17 2.76 Absolute rainfall deviation .75 .24 6.02 .29 *** Experienced crop failure or livestock loss in the last year due to drought .79 Plot size 5.89 7.60 5.66 6.67 Cows owned 3.51 4.07 6.22 5.36 *** Goats owned 5.12 5.88 5.92 6.65 ** Children’s education (average years) 3.79 2.82 3.18 2.97 A: Aspiration indices Individual aspiration index −.07 2.64 −.37 2.46 * Community aspiration index .00 2.96 −.10 3.57 B: Dimensions of individual aspirations Child’s education 13.76 1.44 13.58 1.42 ** Plot size 15.55 13.09 10.37 9.39 ** Number of cows 26.11 27.49 23.91 25.14 Number of goats 37.56 36.89 40.21 37.67 C: Dimensions of community aspirations Housing 62.60 55.61 72.25 71.74 ** School distance .09 .06 .08 .07 *** Security 6.23 3.42 6.28 3.71 Village contributions 130.80 160.02 144.35 174.52 Mutual support 33.77 33.70 30.86 55.89 Meetings 3.18 1.43 2.70 1.38 *** Note:N=990. P-values are calculated using two-tailed t-tests. ***p<.01, **p<.05, * =p<.1. SD =standard deviation. seems to be driven by lower aspirations regarding plot size and cow ownership. We also observed a small drop in plot size from 2018 to 2019, which is not an uncommon response of Zambian farmers to drought (Alfani et al., 2021). Differences regarding to the community aspiration index between 2018 and 2019 are not statistically significant. There are a few differences in specific dimensions. Aspirations for school distance, share receiving support, and the number of meetings decreased from 2018 to 2019, while aspirations regarding the contributions to the village even increased by approximately 12 Kwacha. As mentioned earlier, attrition in the study is relatively high. To better understand reasons for attrition and potential threats to the validity of our analysis, we first compare baseline characteristics of non-attriters and attriters. Results are shown in Table 2. Attriters are on average more often male, are younger, and have received more education. They are also more likely to be single and are less likely to be a savings group member. These systematic differences are intuitive since they can be linked to either a person’s need to leave the study area or their chances to find sources of income outside their rural village. Gender roles in the region often suggest that women should stay at home and men work outside, which makes it easier for men to leave the village and find work elsewhere. Similar arguments can be made for education. Differences with regards to the aspiration indices are not statistically significant. 4.2 Regression results and discussion Table 3summarizes the impacts of drought on individual aspirations. Column (1) shows the results following the modelshowninEquation(5) without adjusting for control variables. Column (2) shows the results for the same regression with control variables included, but without a dummy for the year 2019. As discussed earlier, this is our preferred specification. Results shown in column (3) contain both control variables as well as a dummy for the year 2019. Generally, we 992 PARLASCA et al. TABLE 2 Attrition – Baseline comparison. Non-Attriter (N=495) Attriter (N=148) Difference Female (1/0) .51 .41 ** Age (years) 44.05 40.96 ** Highest grade 7.145 7.621 * Single (1/0) .07 .14 ** Locus of control .96 .94 Herd size (TLU) 4.09 4.10 Plot size control 5.89 5.15 Children’s education 3.81 3.84 Health group member .02 .01 Savings group member .19 .12 ** Individual Aspirations −.07 .25 Community Aspirations −.00 .00 Notes:N=643. We were able to re-interview 495 of 643 people. P-values are calculated using two-tailed t-tests. *** p<.01, **p<.05, *p<.1. Abbreviation: TLU, tropical livestock unit. find a negative and statistically significant effect of drought intensity on individual aspirations. The coefficients for the drought intensity in columns (1) and (2) are very similar, which lends some support to the argumentation that rainfall is exogenous to the time variant control variables The parameter δfor the main specification in column (2) is well above the threshold of one, even for conservative values for 𝑅2 𝑚𝑎𝑥. However, the results in column (3) show that when a dummy variable for the year 2019 is also included, the coefficient for rainfall deviation is not statistically significant anymore. This may be caused by the limited spatial variation of rainfall deviation and, hence, a strong correlation between the two variables. In columns (4) to (7), we explore different drought indicators, specifically the square of rainfall deviation (columns 4 and 5) and the ASI for cropland (columns 6 and 7). The results remain highly consistent across these alternative specifications and are also statistically significant when the dummy for 2019 is included. This finding is noteworthy, given that the correlation between rainfall deviation and the ASI in the drought year 2019 is relatively modest (approximately 12%). This implies that the drought indicators capture different aspects of drought intensity for the affected population, yet all exhibit a consistent negative association with individual aspirations. The square brackets below the coefficient for the drought indicators in columns (2), (4), and (6) contain the parameter δfor the relative degree of selection under proportional selection of observables and unobservables. The values are well above 1, meaning that even under conservative values for 𝑅2 𝑚𝑎𝑥, the results are insensitive to potential missing variable bias. Column (8) further presents the results for the main specification when respondents are FIGURE 2 Specification curve for the effect of drought intensity on individual aspirations. Notes: N=990. Point estimates for every combination of control variables without a 2019 dummy are shown as black dots. Confidence intervals are shown in dark grey (90%) and lighter grey (95%). weighted using inverse probability weights to address potential attrition bias. Importantly, the key findings are similar and robust even under this adjustment. In Figure 2, we show the specification curve to visualize the sensitivity of the regression coefficient for rainfall deviation in column (2) of Table 3to the inclusion of control variables. The coefficient is negative and statistically significant at a 5% level for all possible combinations of control variables. The respective specification curves for the drought intensity coefficient in columns (4) and (6) of Table 3are qualitatively similar and can be found in the supplementary material. In Figure 3we show how the treatment effects in Table 3 change, depending on the assumption of endline aspirations for attriters. We show results for the regressions without control variables and with the treatment variable being either rainfall deviation or ASI. The coefficients remain statistically significant at a 10% level up to a bound of .3 standard deviations. If the endline aspirations of attriters were .4 standard deviations higher than the endline mean of non-attriters, the results would become statistically insignificant. A change of .4 standard deviations is twice as large as the aspiration change of non-attriters and, therefore, quite a large bound. We view this as a further indication that attrition bias is unlikely to be a significant concern for the validity of the key results. Overall, we therefore find strong and robust empirical support for hypothesis 1. Table 4summarizes the impacts of drought on respondents’ aspirations for their community. Again, column (1) shows the results following the model shown in equation (5) without adjusting for control variables. Column (2) shows the results for the same regression with control PARLASCA et al. 999 UNCCD. (2022). Drought in numbers 2022 - Restoration for readiness and resilience.. Vins, H., Bell, J., Saha, S., & Hess, J. J. (2015). The mental health outcomes of drought: A systematic review and causal process diagram. International Journal of Environmental Research and Public Health,12(10), 13251–13275. https://doi.org/10.3390/ ijerph121013251 Zamani, G. H., Gorgievski-Duijvesteijn, M. J., & Zarafshani, K. (2006). Coping with drought: Towards a multilevel understanding based on consERVATION OF RESOURCES THEORy. Human Ecology,34(5), 677–692. https://doi.org/10.1007/s10745-006-90 34-0 SUPPORTING INFORMATION Additional supporting information can be found online in the Supporting Information section at the end of this article. How to cite this article: Parlasca, M. C., Martini, C. A., Köster, M., & Ibañez, M. (2024). Aspirations and weather shocks: Evidence from rural Zambia. Agricultural Economics,55, 985–999. https://doi.org/10.1111/agec.12858