Policy Research Working Paper 10970 Weathering Shocks Unraveling the Effects of Short-Term Weather Shocks on Poverty in Paraguay Teresa Janz Franziska Gassmann Lyliana Gayoso de Ervin Poverty and Equity Global Practice November 2024 Policy Research Working Paper 10970 Abstract In Paraguay, poverty reduction has stalled since 2014 reductions: 5 percent in urban areas and up to 8.8 per- due to a deceleration in economic growth, which has cent in rural areas, on average. Heat shocks also increased been argued to be partly due to a series of climate-related poverty by 1.7 and 4.2 percentage points in urban and events. Nevertheless, little is known about the impacts of rural areas, respectively. Floods primarily affected urban climate-related shocks on the poor. This study analyzes the areas, increasing poverty by 1.9 percentage points. The extent to which short-term weather shocks have affected impacts vary substantially across regions and household incomes and poverty in Paraguay. It combines data from characteristics: female-headed households in rural areas are the yearly household survey series, the Permanent Con- particularly vulnerable to heat shocks, while households tinuous Household Survey; the fifth generation European that are active in the primary sector suffer most from both Centre for Medium-Range Weather Forecasts atmospheric heat and drought shocks. These findings evidence the reanalysis of the global climate temperature data; and Cli- disproportional impacts of short-term weather shocks on mate Hazards Group InfraRed Precipitation with Station income and poverty across regions and household charac- data from 2004 to 2019. The research design exploits vari- teristics. The results highlight the need to consider actions ation in weather shocks across districts and time, using to promote adaptation and climate risk strategies tailored to ordinary least squares pooled regression analysis. The results subpopulations that are vulnerable to climate, to enhance show that heat shocks led to significant household income overall resilience in the country. This paper is a product of the Poverty and Equity 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 lgayoso@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Weathering shocks: Unraveling the effects of short-term weather shocks on poverty in Paraguay Teresa Janz∗ 1 , Franziska Gassmann† 1 , and Lyliana Gayoso de Ervin‡ 2 1 UNU-MERIT, United Nations University and Maastricht University 2 The World Bank, Poverty and Equity Global Practice ∗ Corresponding author. E-mail: tjanz@merit.unu.edu † E-mail: franziska.gassmann@maastrichtuniversity.nl ‡ E-mail: lgayoso@worldbank.org The opinions and results expressed in this paper are those of the authors and do not necessarily reflect the views of the World Bank, its Board of Directors, the countries it represents. The authors thank Eliana Rubiano-Matulevich, Ruth Hill, Mariana Conte, and Pablo Valdivia (World Bank), as well as Britta Augsburg (IFS) and Zina Nimeh (UNU-MERIT) for their feedback and helpful comments and suggestions. The authors are also grateful to Luis Miguel García and Silvia Juliana Granados Ibarra (World Bank) for their support in the provision of data. 1 Introduction In Latin America and the Caribbean, the average number of extreme weather events1 has in- creased by more than 60% in the past two decades compared to the 1980s and 1990s, while in Paraguay it has more than doubled (OECD, 2022). These increases in frequency and intensity of extreme weather events are expected to continue as a consequence of climate change (IPCC, 2021). Due to its consequences on economic activities, ecosystems, and populations, which are significant and, in many cases, irreversible, climate change is considered one of the most systemic threats to our societies. Empirical evidence indicates that low- and middle-income countries (LMCs) are disproportionally affected by the consequences of climate change. In the most recent Global Climate Risk Index –an index that analyzes to what extent countries and regions have been affected by impacts of weather-related events– eight of the ten countries most affected by the quantified impacts of weather shocks in 2019 were LMCs (Eckstein, Künzel, & Schäfer, 2021). In the Global Climate Risk Index, Paraguay is ranked 20th worldwide (Eckstein et al., 2021). Climate change is expected to worsen existing inequalities and poverty levels across the globe (Olsson et al., 2014). A recent cross-country study finds that a one-degree Celsius increase in temperature leads to a 9.1 percent increase in poverty (measured at the USD $1.90 poverty line), as well as a 0.8 percent increase in the Gini index (Dang, Cong Nguyen, & Trinh, 2023). Furthermore, existing socio-economic inequalities can lead to uneven impacts of weather shocks in LMCs, in particular, if vulnerable communities depend on natural resources for income gen- eration, or when the poor are largely concentrated in environmentally risky areas (Reckien et al., 2017). In Paraguay, climate-related shocks may have contributed to stalled poverty levels in the last decade. While Paraguay was able to reduce poverty in half between 2003 and 2022, from 51.4% to 24.7%, the pace of poverty reduction has stalled since 2014. At the same time, Paraguay experienced an increase in the frequency of floods and droughts, as well as in temperatures. In 2019 alone, the country not only recorded the highest average temperature, with an annual temperature of 24◦ C, but also experienced severe droughts, floods, and even heat waves (GRID- Genever European Comission, 2023). Furthermore, historical data denote that years in which the country experienced droughts and floods were associated with sharp reductions in Gross Domestic Product (GDP) per capita and with increases in poverty levels (Gayoso de Ervin & Rubiano, 2023), evidencing the possible link between climate change and socioeconomic indica- 1 Defined as natural disasters (landslides, storms, droughts and floods) resulting in 100,000 or more people affected, or 1,000 or more deaths, or at least 2% of GDP in estimated economic damages (OECD, 2022). 1 tors. Despite the growing concern about the potential impacts of climate change, however, not much is known about the impacts of extreme weather events on incomes and poverty. Exist- ing studies have either focused on the potential impacts of climate change on specific economic sectors at a macro-level of analysis (CEPAL, 2014) or on more micro analysis estimating the potential impacts of climate change on food security (Ervin & Gayoso de Ervin, 2019). In this paper, we analyze to what extent weather shocks have had an effect on (labor) in- comes earned and poverty in Paraguay. Towards this goal, we combine yearly household survey data from the Permanent Continuous Household Survey (Encuesta Permanente de Hogares Con- tinua (EPHC)) with ERA5 temperature data and CHIRPS precipitation data between 2004 and 2019. We merge these datasets at the lowest administrative level possible (districts) to account for large variations in weather across the country and construct short-term weather shock vari- ables characterized as anomalies from long-term means. Then, our empirical strategy consist of exploiting variations in weather shocks across districts and time, following closely Letta, Mon- talbano, and Tol (2018), Sedova and Kalkuhl (2020), and Aggarwal (2021) through pooled OLS cross-sectional regression models. Our main findings indicate heat shocks lead to average reductions of households’ incomes of 5% in urban areas, and up to 8.8% in rural areas, over the period of study. Drought shocks show similar negative impacts on rural areas. Our results also show that short-term weather shocks are associated with increases in poverty, with heat shocks increasing poverty levels by 4.2 percentage points in rural areas, and 1.7 percentage points in urban areas, on average. Flood shocks primarily affect urban areas, increasing poverty by 1.9 percentage points, on average. The results presented in this paper also evidence the regional heterogeneity of the impacts of short-term weather shocks: while heat shocks and flooding are most detrimental in urban areas, heat shocks and droughts have the largest negative effects on rural socio-economic well-being. These findings suggest that Paraguay’s considerable poverty reduction over the past decades may be threatened in the face of climate change. Therefore, protecting the incomes of the most poor and vulnerable will become increasingly important to avoid set-backs in poverty reduction efforts and enhance resilience to future shocks. The reminder of the paper is structured as follows. Section 2 reviews the existing literature, and Section 3 provides in-depth background information on the context of poverty and weather shocks in Paraguay. Data and methods are presented in Section 4, while Section 5 provides the results. In Section 6, we discuss the findings, and Section 7 concludes and presents some policy options. 2 2 Literature Review A growing strand of literature on the economics of climate change focuses primarily on the effect of changes in long-term climate patterns, such as an increase in average annual temper- atures, on economic output. Evidence shows that climate change affects the level and growth of economic output, among others, through reduced productivity growth (Dell, Jones, & Olken, 2012; Deryugina & Hsiang, 2014; Somanathan, Somanathan, Sudarshan, & Tewari, 2021). A recent study by Bilal and Känzig (2024) suggests that macroeconomic damages from climate change are even larger than previously thought, and a 1°C rise in global temperature causes global GDP to decline by 12%, mainly due to a surge in extreme weather events. Apart from impacts from long-term climate changes, short-term, extreme weather events such as hurricanes, floods, and heatwaves can cause immediate and severe impacts on affected areas and populations (Felbermayr, Gröschl, Sanders, Schippers, & Steinwachs, 2022; Gao & Mills, 2018). However, a priori, it is unclear whether the impacts of more long-term climate change or short-term weather shocks translate into adverse impacts on well-being and income distributions, as there may be large heterogeneity both across and within countries, for instance across economic sectors or income classes (Hallegatte, Fay, & Barbier, 2018). As one of the human activities most depen- dent on climate, agriculture is perceived to be the dominant channel through which climate change affects livelihoods (Ahmed, Diffenbaugh, & Hertel, 2009; Hallegatte & Rozenberg, 2017; Hertel, Burke, & Lobell, 2010). However, while globally, overall impacts of weather shocks like heat-waves, droughts and floods on agricultural production are projected to be negative, there is significant variation in the magnitude and direction of change in crop yields at national or regional levels. In fact, this impact of will depend on the openness of the economy, and the integration of markets within the economy (Gouel & Laborde, 2021; Wang et al., 2021). On a more micro scale, weather shocks can significantly impact the level of well-being of agri- cultural communities by reducing agricultural productivity and production, which directly affects their income and food security (Alpízar et al., 2020; Aragón, Oteiza, & Rud, 2021; Ngoma, Finn, & Kabisa, 2024). Due to the relatively inelastic demand for food, a reduction in agricultural production likely translates to significant price increases for agricultural commodities, exacer- bating food insecurity. In line with that, poverty may also rise for non-agriculturally specialized households due to the decline in relative prices of non-agricultural commodities compared to agri- cultural goods or food price spikes as a result of a supply shock, while net agricultural producers may benefit when selling their surplus (Ahmed et al., 2009; Hallegatte, Vogt-Schilb, Rozenberg, Bangalore, & Beaudet, 2020; Hertel et al., 2010). Adverse weather conditions also reduce la- 3 bor productivity, limiting farmers’ ability to work and earn income. Moreover, severe weather events can destroy physical assets like livestock and equipment, further impeding future income generation and recovery efforts. Although richer households may lose more in absolute terms after a weather shock (such as a flood) due to higher levels of income and asset endowments, poorer households are found to experience higher losses relative to their income (Hallegatte et al., 2020). This may be due to their generally lower quality of assets, for instance housing, and less diversified economic portfolios such as high dependency on agricultural income and natural ecosystems (Béné, 2009; Carter, Little, Mogues, & Negatu, 2007; Krishna, 2006; Sen, 2003). On top of that, the poor often settle in areas at risk of weather shocks, as areas close to risky neighborhoods may offer attractive economic opportunities and higher incomes. Along with a rising urban share of poverty (Ravallion, 2007), poor and vulnerable urban communities often inhabit densely populated, and environmentally risky areas (Hallegatte et al., 2020; Reckien et al., 2017; Revi et al., 2014). Additionally, households are pushed into more risky areas due to unfavorable land and housing markets, especially in urban areas, and within them in slums (Erman et al., 2019; Hallegatte et al., 2020; Hardoy & Pandiella, 2009). While both poor and non-poor households may live in areas that are affected by a large-scale, one-off weather shock, areas with recurring hazards such as yearly floods during the monsoon season are almost exclusively inhabited by the poor (CEPAL, 2014; Hallegatte et al., 2020). Eventually, urban neighborhoods become more vulnerable due to the lack of adequate infrastructure and services, such as water and sanitation facilities (Hardoy & Pandiella, 2009). Lastly, households and individuals may apply different risk-management strategies in the face of a weather shock, among others saving, borrowing, informal or formal insurance, or a change in income-generating activities, such as migrating to more productive agricultural areas or shifting to different economic sectors (Bandyopadhyay & Skoufias, 2015; Dallmann & Millock, 2017; Del Ninno, Dorosh, & Smith, 2003; Dercon, 2004; Zeleke, Beyene, Deressa, Yousuf, & Kebede, 2021). However, poor households often lack access to these strategies, and as they are typically averse to fluctuation in consumption, they may experience a transitory welfare loss. This, in turn, may lead to chronic effects on income or income growth through (ex-ante) avoidance of profitable but risky investment opportunities in favor of low-risk low-return strategies, as well as (ex-post) loss of physical or social capital (Dercon, 2004). Moreover, although the poor often rely on risk-sharing strategies such as borrowing and receiving assistance from family and neighbors (Ansah, Gardebroek, & Ihle, 2021; Nguyen, Nguyen, & Grote, 2020), they are typically less able to rely on social networks or government support in the form of social safety nets in post- shock situations and may not have sufficient and immediate resources to recover (Hallegatte & 4 Rozenberg, 2017; Hallegatte et al., 2020). The negative impacts of weather shocks on poverty eventually depend on the extent to which the poor are exposed to weather shocks and more susceptible to their adverse consequences, which may push them into or prevent them from moving out of poverty. A growing body of re- search provides evidence that in contexts where the poor are highly exposed to extreme weather events and economically dependent on environmental systems, weather shocks may increase poverty levels, while simultaneously making those who are already poor poorer (Hallegatte & Rozenberg, 2017). However, these impacts may be uneven within and across countries. While evidence on (cross-country) macro-level studies on the impact of climate change on economic output (Cavallo, Hoffmann, & Noy, 2023; Ishizawa & Miranda, 2016; Lachaud, Bravo-Ureta, & Ludena, 2022; Nagy et al., 2018; Reyer et al., 2017), or conceptual papers on the vulnera- bility to climate change in Latin America (Hardoy & Lankao, 2011; Hardoy & Pandiella, 2009; Zuñiga, Lima, & Villoria, 2021) exist, only few studies build on those findings and analyze micro- level relationships between weather shocks and well-being, such as Zapata (2023) in Ecuador or Rozenberg et al. (2021) in Argentina. A few other studies investigate the link between weather extremes and well-being in very localized contexts, for example Metcalfe et al. (2020) for agri- cultural communities in Yucatan, Mexico, Aragón et al. (2021) for subsistence farmers in Peru, or Harvey et al. (2018) for smallholder farmers in different Central American countries. In Paraguay, a number of studies have analyzed the potential impacts of climate change on socioe- conomic indicators in Paraguay (CEPAL, 2014; Ervin & Gayoso de Ervin, 2019), infrastructure (Elkadi et al., 2019; Silvero, Lops, Montelpare, & Rodrigues, 2019), and health (Conte Grand, Schulz-Antipa, García-Witulski, & Rabasssa, 2024; Gómez Gómez et al., 2022). However, no studies have looked at the relationship between short-term weather shocks and poverty. We are, to the best of our knowledge, the first to implement a quantitative study on the relationship between short-term weather shocks and income and poverty in Paraguay. 3 Background Over the last two decades, Paraguay has made impressive progress in terms of economic and social development. Political and economic stability have allowed the country some of the highest economic growth rates in Latin America (World Bank, 2021). Poverty at the national poverty line more than halved between 2003 and 2022, from 51.4 % to 24.7% (INE, n.d.-b). However, a major challenge for socio-economic well-being in Paraguay is the increasing frequency and intensity of weather shocks as a result of long-term changes in climate patterns, and their consequences on people’s livelihoods. As Paraguay is hit by periods of drought and flooding 5 events, for instance induced by changes in the El Niño Southern Oscillation weather pattern, some of the gains in poverty reduction are at risk of reversal. The economic recession induced by the 2019 drought coupled with subsequent inflation, and not least the onset of the COVID-19 pandemic in 2020, led to a poverty rate of 26.9% in 2021, more than three percentage points higher than pre-pandemic levels (23.5% in 2019) (World Bank, n.d.-b, 2023). While by 2022, poverty had fallen to 24.7%, the increases in poverty seen after this series of shocks highlight Paraguay’s vulnerability in the face of more severe and frequent weather shocks. Paraguay’s climate is characterized by high humidity and warm temperatures throughout the year, with hot and rainy summers, as well as high temperature variability enabling both heat waves and frost across the country (World Bank, 2021). Between 1991 and 2020, annual average temperatures in Paraguay have increased by 0.29°C per decade, with particularly large increases in the summer months (World Bank, n.d.-a). Extreme weather events, such as flooding during the summer months (November to December), and heat waves or droughts, especially during December to February, are increasingly common. Warm spells and the number of heat days (days with temperatures above 30°C) have also become more frequent over time and are projected to further increase based on climate change projections at different emission scenarios (Magrin et al., 2014; World Bank, 2021). At the same time, inter-annual precipitation variability has been increasing, with greater rainfall in the summer months leading to flooding, and an increase in the number of days with heavy rainfall. A combination of the increasing frequency and intensity of short-term weather shocks and the country’s economic structure makes Paraguay particularly vulnerable to the impacts of climate change (Coronel et al., 2015; Lovino et al., 2021; World Bank, n.d.-a). For instance, agriculture, forestry and fishing, a sector that is extremely dependent on changes in weather patterns and climate hazards (World Bank, 2021), contributed 11.6% to the country’s GDP in 2021, and employed 18.9% of the population in the fourth quarter (INE, n.d.-a; World Bank, n.d.-c). Farming in Paraguay is predominantly rain-fed, and thus, subject to extreme precipitation events and heat waves that pose large risks for crop production and agricultural yields, or can lead to livestock loss (World Bank, 2021). Furthermore, around 90% of producers are small rural farmers, whose income and consumption heavily dependent on agriculture and feel immediate economic and food security consequences of reductions in agricultural output while possessing limited capacity to adapt to negative consequences (Benitez Rodriguez, Wolf, Trotter, & Gurgel, 2023; Ervin & Gayoso de Ervin, 2019). Simultaneously, excessive rain or flooding can destroy crops and affect domestic food markets, leading to an increase in food prices that may eventually translate into a decrease in food consumption. Furthermore, heat shocks can exacerbate existing 6 tensions for water between agricultural and human needs and decrease yields, with particularly negative effects on soy and livestock farmers (World Bank, 2021). Despite Paraguay’s relatively large rural population in comparison to other countries in Latin America, nearly 63% of its population was living in urban areas in 2021 (INE, 2021), which is projected to increase to 74% by 2050 (World Bank, 2021). Urban communities in Paraguay are increasingly exposed to flooding after intense rainfall events, and subject to property damage, loss of assets or negative health impacts due to inadequate infrastructure and draining systems (World Bank, n.d.-a). As urbanization rates grow, increasing pressure on dense urban struc- tures and public infrastructure elevates urban populations’ vulnerability to weather shocks. In conclusion, Paraguay is highly vulnerable to the impacts of climate change due to a combination of the country’s economic structure, lack of resilient infrastructure and socio-economic activities that are particularly sensitive to climate risk. 4 Empirical Strategy 4.1 Data Household survey data. We use household survey data from the Permanent Continuous Household Survey (Encuesta Permanente de Hogares Continua (EPHC)). The EPHC is an annual, cross-sectional survey that provides indicators related to labor markets, income, and other socio-economic characteristics that allow to analyze the evolution of well-being in the Paraguayan population over time (Instituto Nacional de Estadística, 2022). The EPHC also represents the main data source for official poverty estimations in Paraguay, with poverty being measured on an annual basis. For our purposes, we use data from 2004 to 2019, from which we generate a pooled cross-sectional data set of 379,170 individual observations (Table A.1 in Appendix A). Since 2017, the EPHC has been collected on a quarterly basis, while previously, the EPHC was collected only on the fourth quarter of each year. While data for the years 2020- 2022 is available, the observed increases in poverty or decreases in incomes during the pandemic and subsequent economic recession may not be primarily attributed to weather shocks, and including such data may introduce contamination in our estimations. Therefore, our primary analyses in this study focuses on pre-pandemic years. It is also important to note that the poverty estimates used in this study are based on weights derived from national projections, 2015 revision. In 2022, the statistical office conducted a Population and Housing Census, and updated the weights for household surveys, but at the completion of this study, data for 2022 and 2023 were only available. Therefore, the historical series used in this study are not strictly 7 comparable with the poverty figures using the revised weights. Table 1 shows descriptive statistics for selected economic and demographic variables, disag- gregated by urban and rural areas. Our main variables of interest are total per capita income,2 labor income, and poverty status. Total income and labor income are initially measured as individual income, with a reference period of the ’past month’, which is then aggregated to monthly income at the household level and finally divided by household size to obtain per capita income values. We convert monthly per capita income as well as the poverty lines (absolute and extreme) measured in local currency, Guaraní (PGY), into per capita income in 2017 US Dollars (PPP). Further, we spatially deflate incomes by the factor 0.897 for urban households to account for differences in price and income levels between urban and rural areas (Dirreción General de Estadística, Encuestas y Censos, 2012). Following from this, poverty is estimated by comparing monthly per capita incomes to the national poverty lines, separately for urban and rural areas. On average, over the period of 16 years, per capita labor income represented 75% of total income in urban, and 78% in rural areas. Moreover, a median income and labor income of $372.5 and $282.9 in urban and rural areas, respectively, which are considerably lower than the mean, points towards a skewed income distribution in Paraguay. Temperature and precipitation data. The lowest level of geographical aggregation in the EPH is the district-level, representing the smallest political-administrative constituencies in Paraguay (Instituto Nacional de Estadística, n.d.). In total, there are 263 districts in Paraguay, including the capital city Asunción with population sizes from as low as 717 inhabitants in San Carlos, Concepción, to approximately 243,800 inhabitants in Ciudad del Este, Alto Paraná and 530,5000 in Asunción in 20233 (de Catastro, 2022). However, not all 263 districts are sampled for the EPH in every year. Based on all available districts in a given year, we match household survey data from the EPH with weather data. We use data on average and maximum daily temperature from the ERA5 Land Aggregates provided by the European Centre for Medium- Range Weather Forecasts (ECMWF) and Copernicus Climate Change Service (Muñoz Sabater, 2019; Service, 2017). ERA5 data are available from 1950 to three months from real-time, and have a spatial resolution of 0.1° (approximately 9km × 9km grid spacing). For precipitation data, we use daily rainfall data from the Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) global rainfall data set, which are available from 1981 and have a spatial 2 Total income is composed of labor income, rental income, income from interest and dividends, income from divorce benefits or child care, income from family assistance (remittances) from both within Paraguay and abroad, income from pension, income from social assistance programs (e.g., Tekoporã), social pension (the program Adulto Mayor), and other miscellaneous incomes. In 2019, labor incomes made up 85.6% of total incomes, on average (INE, 2020). 3 The average population size per district in 2003 was 22,334 including Asunción, and 20,293 excluding Asun- ción. 8 Table 1: Descriptive statistics Encuesta Permanente de Hogares (EPH) Urban Rural Mean SD Mean SD Income Monthly per capita income, 2017$ PPP 559.7 973.1 335.8 803.8 Monthly per capita labor income, 2017$ PPP 425.9 715.6 260.4 763.4 Poor (%) 25.8 43.8 43.9 49.6 Extremely poor (%) 3.0 17.0 14.5 35.2 Demographics Household size 5.0 2.3 5.3 2.4 Age dependency ratio 0.72 0.71 0.95 0.86 Female household head (%) 33.8 47.3 25.6 43.6 Age household head 47.2 14.2 47.1 14.5 Years of schooling household head 8.5 4.6 5.7 3.6 Head and spouse only speak Guaraní (%) 15.9 36.6 52.5 49.9 Employment Any HH member in agriculture (%) 23.8 42.6 80.1 39.9 Employed (%) 92.9 25.7 96.1 19.3 Share of employment in primary sector (%) 3.8 19.2 54.3 49.8 Share of employment in secondary sector (%) 20.9 40.7 13.4 34.0 Share of employment in tertiary sector (%) 75.3 43.1 32.3 46.8 Assets Household has a car/truck (%) 31.0 46.3 11.9 32.4 Household has heating (%) 9.9 29.8 2.1 14.3 Household has a microwave (%) 23.4 42.4 6.0 0.238 Observations 242,921 187,673 Note: All calculations are based on the pooled data from the Encuesta Permanente de Hogares (EPH) for 2004- 2019. Individual-level survey weights are applied. Income and labor income is expressed in monthly per capita values at $2017 PPP. Poverty and extreme poverty rates are calculated using Paraguay’s national poverty lines for urban and rural areas, poverty rates show the national aggregate poverty estimate. The age dependency ratio is defined as the number of children aged 0 to 14 and the number of elderly aged above 65, divided by the number of (working-age) adults, aged 15-65. The employment rate is based on all individuals aged 10 or above. Shares of employment by economic sector are based on individuals aged 10 or above who are employed and/or who indicated their type of occupation. resolution of 0.05° (approximately 5km × 5km grid spacing) (Funk et al., 2014). We use the Google Earth Engine API to download daily temperature and precipitation data, and aggregate raw daily values by calculating geo-spatial averages at the district-level. We produce moving averages for precipitation and temperature for a time bin of 90 days, which we match with survey data based on the exact day of interview. A 90-day, or three-month window is a useful period for capturing short-term weather conditions, as this period is long enough to smooth out daily fluctuations and provide a representative snapshot of seasonal weather patterns, but short enough to reflect the specific weather conditions that individuals and households experience and report in surveys. Although the reference period for income is the last month before the interview, potentially negative effects from adverse weather conditions may not be immediate, so while we do cover the income reference period, we allow for a lead time period of another 9 Figure 1: Average temperature and total precipitation during October-December 2019 (a) Temperature (b) Precipitation Note: Districts in white are not included in EPH. Source: authors’ calculations based on ERA5 and CHIRPS data. two months in order to allow for income changes to materialize. This is particularly relevant in rural areas, where incomes depend on among others harvest, that in turn depend on weather conditions during growing seasons, and should even less immediately react to a weather shock.4 In addition, a three-month period corresponds to the natural seasons in Paraguay and captures seasonal effects without the noise of longer-term weather trends during the year. Figure 1 shows the average short-term temperature and total precipitation across districts in Paraguay during the survey months October to December of 2019 (see Figures B.1 and B.2 in Appendix B for all years). Generally, we observe higher average temperatures in more northern districts, as well as in districts west of the Paraguay River (the river that divides the country in two distinct regions). Correspondingly, districts east of the Paraguay River and in the south receive more precipitation. In this study, we characterize weather shocks as short-term extreme atmospheric conditions in reference to long-term averages. Due to the high relevance of temperature and precipita- tion (Auffhammer, Hsiang, Schlenker, & Sobel, 2013), and their potential link with poverty 4 A precise analysis covering the entire (or main) growing and harvest seasons in Paraguay is however difficult, as there are different growing seasons for different crops like soy bean, maize, wheat, and so on (FAO, 2023). Disentangling the effects of weather shocks on different households involved in different types of agricultural activities lies beyond the scope of this paper. 10 and inequality, we use both types of weather anomalies simultaneously. Following Sedova and Kalkuhl (2020), our measures of weather shocks are, therefore, variables that capture positive and negative temperature and precipitation anomalies during a time period of 90 days before a household’s EPH survey date. First, we construct a Standard Temperature Index (STI) and Standard Precipitation Index (SPI) that measure temperature and rainfall variability as the difference between short-term average temperature or precipitation during the time bin t and the historical average (or long-term climate pattern) during this time of the year between 1981 and 2003, divided by the historical standard deviation (Patel, Chopra, & Dadhwal, 2007), Pdt − Pdh Tdt − Tdh SP Idt = and ST Idt = σPdh σTdh where Pdt is average precipitation during 90 days prior to the interview, Pdh is the historical average precipitation during this time bin, and σPdh is the historical standard deviation of precipitation during the same time bin, which equally applies to the STI. Historical values as reference period of long-term climate conditions cover the 23 years between 1981-2003, corre- sponding to the start of the CHIRPS and leading up to the start of the EPH data collection. In line with previous literature, we use the STI and SPI to capture weather shocks, as levels of temperature or precipitation matter less in an absolute sense but rather in proportion to a district’s historical variation (Heim, 2002; Hirvonen, 2016; Letta et al., 2018; Randell & Gray, 2016). Based on these indices, we construct four measures of positive and negative weather anomalies as: T (+)dt = max{0, ST Idt }, T (−)dt = max{0, −ST Idt }, P (+)dt = max{0, SP Idt } and P (−)dt = max{0, −SP Idt }, where a positive (negative) anomaly is defined as the absolute z-score of temperature or precipi- tation, if the index is larger (smaller) than zero, and zero otherwise. Therefore, in this study, the four weather shock measures are T (+)dt , a positive temperature anomaly (also referred to as heat shock), T (−)dt , a negative temperature anomaly (cold shock), P (+)dt , a positive precipitation anomaly (flood shock), and P (−)dt , a negative precipitation anomaly (drought shock). Figure 2 shows the averages of these weather shocks between 2004 and 2019. Eventually, our measure of weather shocks captures short-term temperature and precipitation extremes in relation to long-term climatic conditions in the respective districts. In doing so, we capture meteorologi- cal conditions as expressed by the SPI and STI, instead of more complex indices such as the Standardized Precipitation Evapotranspiration Index (SPEI) or other drought indices for several 11 Figure 2: Average temperature and precipitation anomalies (2004-2019) (a) Positive temperature anomaly (heat shock) (b) Negative temperature anomaly (cold shock) 2 2 1.5 1.5 1 1 .5 .5 0 0 2005 2007 2009 2011 2013 2015 2017 2019 2005 2007 2009 2011 2013 2015 2017 2019 Year Year (c) Positive precipitation anomaly (flood shock) (d) Negative precipitation anomaly (drought shock) 2 2 1.5 1.5 1 1 .5 .5 0 0 2005 2007 2009 2011 2013 2015 2017 2019 2005 2007 2009 2011 2013 2015 2017 2019 Year Year Note: Average temperature and precipitation anomalies in survey sample, using individual weights. The time period refers to a time bin of 90 days before a household’s date of interview in the EPH survey. reasons. First, raw data from the CHIRPS and ERA5 data bases are more granular than the SPEI that is downloadable for instance from Google Earth engine. As the district is the lowest level of geographical aggregation in the EPH data, we want to ensure maximum variation of our weather shock variables at the district level, and as districts in (southern) Paraguay are rather small, we opt for the more granular weather data. Furthermore, several studies have shown high and significant correlation between the SPI and SPEI, so our estimates should be effective in capturing drought conditions (Danandeh Mehr, Sorman, Kahya, & Hesami Afshar, 2020; Salimi, Asadi, & Darbandi, 2021; Torelló-Sentelles & Franzke, 2022). Table 2 displays descriptive statistics for relevant temperature and precipitation variables. Temperature and precipitation were, on average, positive during the reference period 2004-2019, meaning that temperatures were on average higher. Similarly, more rainfall was observed than historically (1981-2003). The number of heat days (>30°C) and extreme heat days (>35°C) was also higher than the historical average. Regarding the range of weather shocks (anomaly), they range from 0.31 to 0.91 standard deviations, on average. Furthermore, positive anomalies such as heat and flooding shocks seem to be larger in magnitude than negative anomalies, capturing cold and drought shocks. 12 Table 2: Descriptive statistics ERA5 and CHIRPS weather data Mean SD Min Max Temperature Average temperature 22.59 2.29 15.93 28.52 Historical average temperature 22.00 1.96 16.45 27.37 Z-score average temperature 0.68 1.25 -2.24 4.42 Average maximum temperature 28.58 2.13 18.96 33.76 Historical average maximum temperature 27.90 1.89 21.92 32.50 Z-score average maximum temperature 0.64 1.04 -2.00 3.87 Heat days Number of days above 30°C 38.76 14.43 0.00 81.00 Historical number of days above 30°C 34.54 12.10 0.00 74.50 Number of days above 35°C 8.80 6.68 0.00 39.00 Historical number of days above 35°C 5.63 3.37 0.00 23.52 Precipitation Average precipitation 4.36 1.81 0.63 10.80 Historical average precipitation 4.15 1.13 1.27 6.01 Z-score average precipitation 0.14 0.95 -2.46 4.11 Total precipitation 392.40 163.27 56.73 972.43 Historical total precipitation 373.46 101.33 114.26 540.54 Z-score total precipitation 0.14 0.95 -2.46 4.11 Anomalies Positive temperature anomaly (heat shock) 0.91 0.97 0.00 4.42 Negative temperature anomaly (cold shock) 0.23 0.44 0.00 2.24 Positive precipitation anomaly (flood shock) 0.45 0.65 0.00 4.11 Negative precipitation anomaly (drought shock) 0.31 0.45 0.00 2.46 Observations 379,170 Note: Authors’ calculations based on ERA5 and CHIRPS data. Averages within survey sample, estimated using individual weights. The time frame refers to a time bin of 90 days before a household’s EPH interview. Historical values are calculated based on the years 1981-2003, all other values are averaged over the years 2004-2019. The number days above 30°C/35°C are calculated based on the daily maximum temperature. 4.2 Methodology Our empirical framework builds on previous work that aims to estimate micro-level effects on poverty and inequality resulting from the observed increase in frequency and severity of extreme weather events due to climate change, such as Letta et al. (2018) and Aggarwal (2021). By exploiting exogeneity in weather shocks across districts and time we are able to estimate rela- tionships between weather shocks and (labor) income or poverty in Paraguay during the years 2004-2019. Following Sedova and Kalkuhl (2020), we construct the regression model of the form: − − ′ ′ + Yidt = β0 + β1 Tdt + β2 Tdt + + β3 Pdt + β4 Pdt + Wd γ + Xidt δ + ηdt + θm + ϵit (1) where Yidt is our main variable of interest, which is measured in two different ways: as lnYidt , where Y is either monthly total or labor income of household i in district d and time period t, log transformed using an inverse hyperbolic sine transformation (Burbidge, Magee, & Robb, 13 1988), or a binary variable identifying individuals living in a household whose per capita income lies below the national poverty or extreme poverty line. We estimate these models using OLS − − and probit models with robust standard errors,5 respectively. Tdt + , Tdt , Pdt + and Pdt represent measures of temperature and precipitation shocks in district d at time period t, measured as ′ is a vector of weather and climate positive and negative anomalies as explained in 4.1, Wd variables, namely the historical average temperature and precipitation in district d, and the ′ is a vector of average temperature and precipitation in the district during time period t,6 Xidt household demographic variables, ηdt are department-year dummies,7 θm are interview-month dummies, and ϵidt is the error term. In all models, we use individual-level survey weights to obtain population representativeness. After estimating the models, we calculate marginal effects which allow us to obtain estimates of the changes in the expected value of the dependent variable associated with a one-unit change in the independent variable of interest, in this case temperature and precipitation anomalies, while holding other variables constant. To account for differences in incomes between urban and rural areas, and because we expect the impact of weather shocks to be structurally different given the different economic structure of production in rural vs. urban areas, we split the models by area of residence. Since districts are the lowest level of geographical aggregation in the EPH data, our results will capture the effects of an average weather shocks on income or poverty at the district level. In other words, there may be underlying heterogeneity this analysis will not be able to highlight, such as more nuanced differences in the intensity of weather shocks within districts, and may perhaps under-estimate the severity of the impact for highly-affected households. However, most districts in Paraguay are reasonably small, allowing for sufficient variation across units of analysis. 5 Ideally, we would cluster standard errors at the primary sampling unit (PSU) level. However, PSU codes vary across years in the EPH, so that it can currently not be used for clustering. 6 We include those variables to control for differential weather conditions across areas of Paraguay, similar to previous studies like Azzarri and Signorelli (2020), Gao and Mills (2018) or Asfaw and Maggio (2018). 7 We use the variable for ’representative departments’ in the EPH survey, which groups households into the following areas: Asunción, San Pedro, Caaguazú, Itapúa, Alto Paraná, Central, as well as other urban and other rural areas. ’Other urban and ’other rural’ contains all households who do not live in any of the aforementioned departments, divided by rural or urban area. Only the departments Boquerón and Alto Paraguay are not covered in the EPH due to the small population size living in those departments (1.4% according to the latest census). As a result, our study is representative of the Paraguayan population except for Boquerón and Alto Paraguay. As our weather shock variables are fixed at the district level, we cannot use district-fixed effects but use (representative) department-fixed effects instead. 14 5 Results 5.1 Main Results 5.1.1 Differential impacts of weather shocks on urban and rural areas In this section, we present the main results of our analysis, which show that weather shocks have significantly different impacts on urban and rural areas. While heat shocks are particularly detrimental to both, floods in urban areas and droughts in rural areas have large negative impacts. Figure 3 reports the results of Equation 1, which captures the effects of short-term temperature and precipitation shocks, as defined in Section 4.1, on total as well as on labor income. Our shock measures are linear (Section 4.1), but here we report coefficients of a one- unit change in the shock variable, corresponding to a weather shock with the intensity of one standard deviation above the mean, a common threshold used when defining shocks. Hence, all coefficients reported in the following sections correspond to temperature or precipitation anomalies of one standard deviation above the mean. Therefore, the effects could be even larger for more severe weather shocks. Notably, the results indicate that short-term weather shocks have a differential impact on urban and rural areas. In particular, in urban areas, heat shocks are associated with reductions in total income of approximately 5%, and labor income reductions of 4.6%, on average. In contrast, periods of cold temperatures and low rainfall are positively related to total income in urban areas. In rural areas, both heat and drought shocks have a negative effect on income, with income declines of up to 8.8%, and labor income declines of up to 12.1%. Cold and flood shocks, however, are positively associated with (labor) income in rural areas, which may increase by up to 7.7% (see Tables A.2 and A.3 in Appendix A). The impact of short-term temperature and precipitation shocks on poverty in Paraguay are largely in line with the results presented above. In urban areas, heat and flood shocks are important drivers of poverty increases. In particular, the results indicate that heat shocks lead to a rise of 1.7 percentage points in poverty levels, and a 0.4 percentage points rise in extreme poverty, on average (Figure 4, and Tables A.2 and A.3 in Appendix A for regression results). Furthermore, flooding may increase urban poverty by 1.9 percentage points, on average, indicating the relevance of flooding events in urban areas. Periods of cold temperatures and lack of rainfall (drought shocks), however, are negatively associated with poverty in urban areas. In rural areas, the results indicate large (extreme) poverty increases during times of heat shocks (4.2 and 1.2 percentage points, on average, respectively), albeit with rather large confidence intervals. 15 Figure 3: Effects of temperature and precipitation shocks on total income and labor income Income Labor income .1 .05 Income change 0 -.05 -.1 -.15 ck k k ck ck k k ck oc oc oc oc ho ho ho ho sh sh sh sh ts s ts s ld d ht ld d ht ea ea oo oo Co ug Co ug H H Fl Fl ro ro D D Urban Rural Note: Time periods refers to the 90-day time bin before the interview. All coefficients shown in the figure represent marginal effects. The corresponding regression tables with estimated coefficients as well as marginal effects can be found in Tables A.2 and A.3 in Appendix A. Drought shocks are also relevant for the rural population, as our results indicate that they are an important driver of extreme poverty in rural areas (up to 1.7 percentage points). Similarly to urban areas, cold and rainy periods seem to be beneficial for rural poverty reduction. Those results are in line with previous literature in the sense that we find evidence that short-term weather shocks in Paraguay increase, and in some cases deepen poverty, which is particularly true for the most common and devastating weather shocks in the context of Paraguay, namely extreme heat, drought and flooding. In addition to the main results presented above, we also apply a series of robustness tests (Tables A.4, A.6 A.5 and A.7 in Appendix A) with regards to the timing of weather shocks and excluding outliers. We find that the results are largely robust to varying the time period considered for the construction of weather shocks, which shows a similar negative impact of longer-term heat and drought shocks on income, while increasing poverty. In urban areas, the results seem to be more sensitive to the timing of the weather shocks. Yet, our results are robust to excluding outliers for both rural and urban models. 16 Figure 4: Effects of temperature and precipitation shocks on absolute and extreme poverty Poverty Extreme poverty .05 .025 Poverty change 0 -.025 -.05 ck k k ck ck k k ck oc oc oc oc ho ho ho ho sh sh sh sh ts s ts s ld d ht ld d ht ea ea oo oo Co ug Co ug H H Fl Fl ro ro D D Urban Rural Note: Time periods refers to the 90-day time bin before the interview. All coefficients shown in the figure represent marginal effects. The corresponding regression tables with estimated coefficients as well as marginal effects can be found in Tables A.2 and A.3 in Appendix A. 5.1.2 Regional variations in weather shock impacts Furthermore, we also seek to understand the impacts of weather shocks at the regional level, given the large geographical disparities observed in poverty rates in the country. The analysis reveals that weather shocks have varying regional impacts on poverty in Paraguay, with heat shocks significantly increasing poverty in urban centers like Itapúa and Central, while droughts primarily affect rural areas and flood shocks lead to substantial increases in poverty in Alto Paraná and Central. We disaggregate the analysis geographically and run separate estimations by the representa- tive departments included in the EPH, namely Asunción (the capital city), San Pedro, Caaguazú, Itapúa, Alto Paraná, Central, as well as two groups that capture all other urban and rural areas in the country (Figure 5). While different weather shocks have varying impacts on poverty in different departments, sometimes with rather large confidence intervals, the following results stand out. Heat shocks are particularly detrimental for the populations living in Itapúa and Central—two of the departments with the largest urban centers in the country—with average increases in poverty of 7.3 and 3.4 percentage points, respectively, during the time period ana- lyzed. Other departments, however, do not show, on average, an increase or decrease in poverty. 17 Figure 5: Effects of temperature and precipitation shocks on poverty by representative departments .15 .1 .05 Poverty change -.05 0 -.1 -.15 Heat shock Cold shock Flood shock Drought shock Asunción San Pedro Caaguazú Itapúa Alto Paraná Central Other Urban Other Rural Note: Time periods refers to the 90-day time bin before the interview. All coefficients shown in the figure represent marginal effects. The corresponding regression tables with estimated coefficients as well as marginal effects can be found in Tables A.8 and A.9 in Appendix A. Cold shocks have similarly limited, or even beneficial effects on poverty across those regions. Flood shocks largely affect Alto Paraná and Central, with 8.3 percentage points increases in poverty (though the former is marginally insignificant at the 5% confidence level), while other rural areas are mostly affected by droughts (6.4% increase in poverty, on average). 5.2 Heterogeneous Effects Existing literature suggests that the effects of short-term weather shocks on poverty may be heterogeneous, for instance with regards to a households’ economic activities and sex of the household head (Angelsen & Dokken, 2018; Hallegatte, 2016; Hertel & Rosch, 2010). Therefore, we investigate the effects across various demographic and economic dimensions. The analysis shows that female-headed households in rural areas are most vulnerable to heat shocks, while in urban areas, floods have a stronger impact on poverty for these households. Male-headed households are most affected by heat, and by droughts in rural areas. Households active in the primary sector, particularly agriculture, are the most affected by heat and drought shocks, while in urban areas, the secondary sector experiences significant poverty increases due to heat and floods. 18 5.2.1 Differential impacts of weather shocks on female- and male-headed house- holds First, we split the sample into households whose head is male or female, and re-estimate Equation 1. The results of this analysis are shown in Figure 6. Among the most notable find- ings are that poverty rate increases due to heat shocks in rural areas are significantly greater for individuals living in female-headed households than in male-headed households. On average, poverty increases by 7.1 percentage points after a heat shock for those living in female-headed households, compared to a 3.9 percentage point increase in male-headed households. However, for individuals living in male-headed households in rural areas, drought shocks are largely asso- ciated with an increase in poverty by up to 4.8 percentage points on average. In urban areas, floods are largely driving increases in poverty for those living in female- headed households, by up to 3.4 percentage points, with no significant effects among male-headed households. The latter suffer disproportionately from heat shocks, with poverty increases of up to 2.6 percentage points on average. Figure 6: Effects of temperature and precipitation shocks on poverty by gender of household head Male household head Female household head .1 .05 Poverty change 0 -.05 -.1 ck k k ck ck k k ck oc oc oc oc ho ho ho ho sh sh sh sh ts ts ts ts ld d ld d ea h ea h oo oo Co ug Co ug H H Fl Fl ro ro D D Urban Rural Note: Time periods refers to the 90-day time bin before the interview. All coefficients shown in the figure represent marginal effects. The corresponding regression tables with estimated coefficients and marginal effects can be found in Tables A.10 and A.11 in Appendix A. 19 Figure 7: Effects of temperature and precipitation shocks on poverty by economic sector of household head HH head primary sector HH head secondary sector HH head tertiary sector .15 .1 .05 Poverty change 0 -.05 -.1 -.15 ck k k ck ck k k ck ck k k ck oc oc oc oc oc oc ho ho ho ho ho ho sh sh sh sh sh sh ts s ts s ts s ld d ht ld d ht ld d ht ea ea ea oo oo oo Co ug Co ug Co ug H H H Fl Fl Fl ro ro ro D D D Urban Rural Note: Time periods refers to the 90-day time bin before the interview. All coefficients shown in the figure represent marginal effects. The corresponding regression tables with estimated coefficients and marginal effects can be found in Table A.12 and A.13 in Appendix A. Results for primary-sector rural areas exclude households living in the Central department due to insufficient sample size. 5.2.2 Variations in weather shock impacts by economic activity Secondly, we analyze the relationship between short-term weather shocks and poverty based on whether the head of household is economically active in the primary, secondary or tertiary sector (Figure 7). The results indicate that heat shocks in rural areas disproportionally affect households whose head is working in the primary sector, including agriculture, livestock, hunting or fishing. The findings suggest that poverty is estimated to increase by up to 5.3 percentage points, on average, for this population. At the same time, heat and drought shocks decrease incomes by 10% and 12.5%, respectively, although the latter does not seem to translate into effects on absolute poverty (Table A.13 in Appendix A). This further supports the evidence presented in Section 2 that agriculture is the dominant channel through which weather shocks affect well-being, at least in rural areas, and heat and drought have the largest potential to destroy yields and entail further negative economic consequences for both smallholder and large- scale farmers. In urban areas, effects of weather shocks on poverty or income among households whose head is active in the primary-sector are not statistically significant at conventional significance levels. 20 However, heat and floods shocks are associated with increases in poverty for the urban popula- tion living in households whose head is working in the secondary sector, with an average poverty increase of 3.8 and 10 percentage points, respectively. Given that the secondary sector includes the manufacturing industries and construction, among others, high temperatures and floods pre- sumably impede workers to perform regular working hours, reducing their incomes, and pushing them and their dependents into poverty. Again, our results confirm that flooding is one of the most devastating weather shocks in urban Paraguay, with particularly distribution-sensitive effects for secondary-sector workers, as poverty increases, while income remains unaffected. Finally, the population living in households whose head is engaged in the tertiary sector is practically unaffected by negative consequences of weather shocks in terms of poverty. Never- theless, heat shocks lead to income declines of 4.6% in urban, and as much as 8.3% in rural areas, respectively, and drought shocks to a reduction in income by up to 6.8% on average in rural areas (Figure B.3 in Appendix B and Table A.15 in Appendix A). 5.3 Extensions: Heat days To extend our study of short-term weather shocks and poverty in Paraguay, we add an additional analysis in which we change our measure for heat shocks to the number of heat days during the 90-day time bin. The analysis shows that an increase in the number of heat days is strongly associated with rising poverty for rural households engaged in agriculture, while non-agricultural and urban households are less affected, except on days with extreme temperatures. Measuring temperature shocks as days with extreme temperatures has been used widely literature and in various forms, and is considered an effective measure for labor productivity, health, and crop yields, amongst others (Aragón et al., 2021; D’Agostino & Schlenker, 2016; He & Chen, 2022; Schlenker, Hanemann, & Fisher, 2007). We use temperature thresholds of 25°C, 30°C and 35°C, respectively, which are considered as reference thresholds for ’summer days’, ’hot days’, and ’very hot days’ in the context of Paraguay (World Bank, n.d.-a), and thus capture absolute and short-term severe heat shocks. Figure 8 shows the impacts of the number of heat days on poverty disaggregated by whether the household is engaged in agriculture. In rural areas, there seems to be a strong association between the number of heat days and poverty for households in agriculture, with increases in poverty of nearly 0.3 percentage points per additional heat day, on average. The mechanisms here could be either a destruction of crop and reduction of yields, or reduction in physical labor hours due to the heat, as identified in Section 2. Non-agricultural households are less affected, although every additional very hot day (days with temperatures above 35°C) leads to large increases in poverty, by up to 0.4 percentage 21 Figure 8: Effects of heat days on poverty Household not in agriculture Household in agriculture .006 .004 Poverty change .002 0 -.002 Days >25°C Days >30°C Days >35°C Days >25°C Days >30°C Days >35°C Urban Rural Note: Time periods refers to the 90-day time bin before the interview. All coefficients shown in the figure represent marginal effects. The corresponding regression tables with estimated coefficients and marginal effects can be found in Tables A.16 and A.17 in Appendix A. points. Similarly, days with extreme temperatures have no (or rather a positive) effect on urban, non-agricultural households, but those engaged in agriculture experience an increase in poverty of around 0.2 percentage points per additional day above 30°C. 6 Discussion In this study, we provide an in-depth analysis of the effects of short-term weather shocks on income and poverty in Paraguay. Our results show that weather shocks may disproportionally affect the poor in Paraguay, and are thus consistent with previous literature that has established association between weather shocks and poverty (Hallegatte & Rozenberg, 2017). However, their impacts are not even, but vary amongst others by area of residence, economic activities, or the sex of the household head. Our most important findings are summarized and put into context below. Differential effects across urban and rural areas First, the impact of weather shocks is heterogeneous across urban and rural areas. Heat and flood shocks are detrimental for income in urban areas, while households in rural areas 22 suffer from heat and drought shocks, in particular. On the other hand, periods of extremely low temperatures and periods with little rainfall can even increase incomes and decrease poverty in certain contexts. Moreover, the magnitudes of effects are mostly larger in rural than in urban areas (see Figures 3 and 4). In addition, while weather shocks can have immediate negative effects on income, thus increasing the poverty headcount, they also can deepen poverty by increasing extreme poverty, mostly in rural areas. We find that heat shocks deepen poverty by increasing extreme poverty, both in urban and rural areas, with especially large magnitudes for rural populations. Drought shocks also can lead to increases in extreme poverty, by up to 1.7 percentage points on average. As mentioned in Section 5.1, these results are a lower bound estimate of the impact of weather shocks on poverty, as they capture the effect of a temperature or precipitation anomaly of the intensity of one standard deviation above the mean. However, as Figure 2 shows, temperature, and in some cases precipitation anomalies in Paraguay, can exceed one standard deviation above the mean, meaning that the impacts of extreme heat shocks or floods can be much more severe than the estimates reported above. The evidence points out that urbanization in rural areas might be beneficial in terms of resiliency, given the larger impacts of weather shocks on poverty on rural areas relative to urban area. However, the smaller impacts of weather shocks on urban areas could be explained by the lower dependency of income generation activities on natural ecosystems. Indeed, the findings also reveal that flooding and extreme heat are a real threat to poverty in urban areas. Given that in the process of urbanization, vulnerable communities often settle in areas that are susceptible to weather shocks and lack adequate infrastructure and sanitation, urbanization could even lead to a deterioration in resilience to weather shocks in Paraguay. Therefore, interventions that seek to increase the urbanization of rural settlements need to take into account adaptation ac- tions to prepare for the consequences that can derive from weather shocks for urban populations. Heterogeneous effects across population sub-groups Moreover, in some of our models, confidence intervals are rather large, meaning that differ- ent weather shocks can have varying effects for various population subgroups. Heterogeneous effects exist with regards to demographic categories and economic sectors. For instance, rural female-headed households seem to experience larger increases in poverty after a heat shock or a flood shock in urban areas. Presumably, this could be due to the economic structure of these households and gendered labor market decisions, as women are more likely than men to work in occupations that give them more flexibility, but at the same time provide less security, and thus respond more quickly to an economic or weather shock (Otto et al., 2017). As women are 23 typically in charge of household tasks, like cleaning, and care work, female-headed households could be more likely to incur an income shock due competing time demands (Stokes et al., 2015). Also, as women tend to be more risk-averse, they might show more conservative behavior once a weather shock occurs (Molua & Ayuk, 2021; Perez et al., 2015). Furthermore, the impacts of drought shocks on poverty in rural areas are mainly seen in male-headed households. This could be due to the fact that these households are more likely to be engaged in agriculture. We also find larger negative effects on labor income in contrast to to- tal income in rural areas, which suggests that heat and drought shocks are indeed detrimental to economic activities among those households and during the time period studied. Disaggregating the results by economic activity of the household head reflect that negative effects of weather shocks are most severe for rural households whose head works in agriculture. This is also high- lighted by our results from Section 5.3, which shows that extremely hot days have large and persistent detrimental effects on welfare for households who live in rural areas and are engaged in agriculture. Yet, income-decreasing effects of heat (and drought) shocks in urban and rural areas persist in the tertiary sector, for instance for the commerce, hospitality or social services sectors. Next to the direct impacts on the primary sector, the latter can be more indirectly impacted by weather shocks through disrupted trade and transportation, or indirect impacts through increased prices of primary goods. However, those declines in income do not seem to translate into poverty increases, which could be due to the fact that those relatively better off lose more in absolute terms, as highlighted in Section 2. Moreover, the negative effects of weather shocks on poverty in urban areas seem to be driven by households whose head is working in the secondary sector. For instance, in occupations such as those related to construction, weather shocks can force the temporary suspension of work. As the secondary sector constitutes around 20% of urban employment, detrimental impacts can affect a large share of households. Heterogeneous effects within population sub-groups These heterogeneity analyses may still hide variation in the relationship between weather shocks and welfare, possibly because a disaggregation by gender or economic activity of the household head might not paint a sufficiently comprehensive picture of economic activities and income generation within the entire household. In other words, even when the household head’s sector is affected by a weather shock, the financial contribution of a different household mem- ber to household income could compensate potential income shocks. Therefore, recognizing inter-sectional effects and potential heterogeneity both within economic activities or different demographic groups, but also across the income distribution becomes vital. Weather shocks 24 seem to not only affect the poor or near-poor. Some weather shocks induce income declines (or increases) on average, also in traditionally more stable sectors, like the tertiary, and to some extent the secondary sector. This may be driven by the fact that richer households have a larger asset base and may as such incur in larger financial losses, for instance business or farm owners involved in soy production (Wesz Junior, 2022). Secondly, the poorest may find themselves at welfare levels that are so low, that (weather) shocks create limited financial havoc on a living standard that is already at subsistence level. However, on top of that, there also seems to be a vulnerable middle class that remains exposed to the negative effects of precipitation shocks in urban areas. Cumulative and overlapping effects of weather shocks We conclude that the effects of different temperature and precipitation shocks on income and poverty in Paraguay are not to be interpreted in isolation. Within a year, or across years, there may be multiple and overlapping weather shocks with which households must cope. More importantly, the interaction between such shocks and the rate at which they re-occur may have cumulative effects on poverty. A preliminary analysis using the data from this paper suggests that in urban areas, the combined effect of heat and flood shocks on poverty is significant and would increase poverty levels, though a deeper exploration of these joint effects remains for future research. Moreover, as the EPH data does not cover the departments Boquerón and Alto Paraguay, which are in the north of the country and may exhibit the most extreme weather conditions and anomalies (Figure 1), the effects of weather shocks on poverty could be even higher than our estimates suggest. In light of this and the increasing frequency and intensity of weather shocks in the future, we believe that our results are a lower-bound estimation of the impacts of weather shocks on poverty, and more severe effects are to be expected without long-term adaptation and climate risk mitigation strategies. The long-term effects of weather shocks can be profound and complex, not only having an immediate impact on income and poverty, but also exacerbating structural vulnerability over time. Historically, Paraguay has experienced high rates of deforestation. However, over the past decade, this trend has shown signs of improvement, with deforestation rates gradually declining.8 Despite this positive development, the country still faces significant challenges. Climate change poses a particular threat, as it is expected to exacerbate deforestation through increased droughts and wildfires if preventive measures are not implemented (World Bank, 2024, Forthcoming ). In addition, repeated weather shocks may further degrade the country’s natural resources and 8 Between 2000 and 2020 alone, 30 percent of the country’s forest area was lost (OECD, 2022; World Bank, 2021). 25 ecosystems, which are critical to the livelihoods of many rural communities, as detailed in section 3. Soil erosion, deforestation, and water scarcity can reduce agricultural productivity and biodiversity, undermining the sustainability of local economies and food systems. In addition, repeated weather shocks can lead to chronic poverty. Households that are fre- quently affected by employment and income losses may need to deplete their savings or assets, making it increasingly difficult to recover between shocks. As a consequence, households may be trapped in a state of persistent poverty. Finally, as weather shocks become more frequent and severe, households may be forced to migrate in search of better living conditions and economic opportunities. This can lead to further urbanization in Paraguay, and put additional strain on its urban infrastructure and services. Internal displacement can also lead to social tensions and deteriorate the vulnerability of both displaced and host communities. Estimation challenges and limitations Our results are consistent with previous literature that has established association between weather shocks and poverty (Hallegatte & Rozenberg, 2017). Still, some noise in the estimates remain, which we attribute mostly to the timing of the weather shocks, the timing of data collection, and the cyclical nature of economic activities. First, EPH data is collected over a time period of three (and in some years slightly more) months, between October and December (or January), corresponding to the spring and summer months in Paraguay. As described in Section 3, extreme weather events are most common during the summer months, so there may be a variation in the effects we are able to observe depending on whether the household is interviewed in, say, the beginning of October, versus the end of December. As such, the time periods we use to construct our weather variables, which we match with survey data based on the date of the interview, cover different parts of the year, and in the case of households involved in agriculture, the agricultural calendar. We control for seasonal variation by including dummies based on the month of data collection, but there may be unobserved variation we fail to capture. Against this backdrop, we expect the true impacts of extreme weather events to be larger in magnitude than what we can estimate with our data, given that the impacts will be most harshly felt only after the summer, or most severely if weather extremes occur during crop growing cycles. Furthermore, temporal fluctuation in our main variable of interest may persist, especially in rural settings where growing and harvest seasons add to seasonal variation. There are a few limitations to our study. First, since we only have cross-sectional data, we are unable to track households over time and estimate relationships between weather shocks and poverty over time. Furthermore, these results may underestimate the true impact of weather 26 shocks on household income and poverty due to the lack of granularity in the weather data. Since we are not able to match households with weather data corresponding to their exact (or slightly offset) place of residence, but only at the district level, there may be underlying variation we fail to capture, as there is little to no variation across households living in the same district. Districts in Paraguay are reasonably small and differences in weather shocks may only vary slightly from district to district, so we expect this issue to be small. Also, as Asuncion and Central are the smallest districts, but simultaneously the most populated ones, we expect the results to be more robust in those locations, covering a large share of the Paraguayan population. Still, more geographically-coded survey data, potentially enhanced with self-reported shocks may help overcome this limitation, and remains a topic for future data collection and research. Third, there may be other weather shocks with devastating welfare effects, such as river flooding, which we do not or only imperfectly cover in our analysis. 7 Conclusion and Policy Recommendations In this study, we find that weather shocks have significant and heterogeneous impacts on poverty and income in Paraguay. Heat shocks emerge as the most pervasive threat, reducing household incomes by 5% in urban areas and 8.8% in rural areas, while increasing poverty rates by 1.7 and 4.2 percentage points, respectively. Urban areas face distinct challenges from flooding, which increases poverty by 1.9 percentage points, while rural areas are particularly vulnerable to both heat and drought shocks. These impacts are especially pronounced for specific groups: female- headed households in rural areas experience poverty increases of up to 7.1 percentage points during heat shocks, while households engaged in primary sector activities see income declines of up to 10% during heat shocks and 12.5% during droughts. The evidence provided in this study suggests that Paraguay’s considerable poverty reduction over the past decades may be threatened by the increasing frequency and intensity of weather shocks. In spite of that, as the country recovers from the negative economic impacts induced by the COVID-19 pandemic and the protracted drought that followed, poverty levels have fallen to 22.7% in 2023 (Instituto Nacional de Estadística, 2024).9 To build on this progress in reducing poverty in the face of climate change, it is increasingly important to improve resilience to weather shocks and protect the incomes of the most vulnerable to avoid setbacks. In light of these findings, policy makers and practitioners in Paraguay should consider sev- eral strategies to mitigate the negative impacts of weather shocks on vulnerable populations. 9 Note that this poverty level is not strictly comparable to previous years in our data, as poverty levels are being revised to reflect the population level observed in the 2022 Census. 27 First, expanding and improving social safety nets can provide immediate relief to households affected by weather shocks. Conditional cash transfers, such as Paraguay’s Tekoporã program, or expanded social security coverage can help cushion unexpected income losses and prevent households from falling even deeper into poverty. Moreover, improving access to financial ser- vices, including savings accounts, insurance, and credit, can help households manage financial risks and recover more quickly from weather shocks. Second, promoting and facilitating the diversification of income sources, especially in rural areas, can reduce the vulnerability of house- holds that are dependent on a single economic activity, such as agriculture. This could include the promotion of agroforestry, or non-agricultural income-generating activities. As an exam- ple, the Government of Paraguay approved the PROEZA program in 2019, which combines the goals of poverty reduction, reforestation and increasing the consumption of renewable en- ergy. The main component of the program, Sembrando Futuro, is an environmental conditional cash transfer targeting rural households affected by poverty and vulnerability. An expansion of this program may ensure greater coverage of cash assistance while simultaneously promoting sustainable agricultural practices and the diversification of agricultural activities. Moreover, investing in effective extension services to agricultural producers is crucial to enhance their awareness and adoption of climate-smart technologies and practices. By providing farmers with timely and relevant information, training, and support, these services can help mitigate the adverse impacts of weather shocks. Effective extension services could contribute to stabilizing agricultural output and productivity, thereby increasing the resilience of agricultural communities against weather shocks and long-term climate change. Additionally, adopting a differentiated approach that acknowledges the unique challenges faced by small-scale farmers is essential. Despite comprising 90% of all farmers, small-scale producers collectively occupy only about 11% of total agricultural land. They often face significant budgetary and commercial constraints, which hinder their capacity to enhance their resilience to adverse weather shocks. Tailored (financial) support and policies are needed to address these specific limitations and help small-scale farmers effectively adapt to climate change. Furthermore, greater prioritization of and increased funding for agricultural research and development is necessary to drive innovation in seed technologies and livestock management practices, helping to mitigate the impacts of short-term weather shocks and enhance the long-term resilience of the agriculture sector. Finally, developing infrastructure that can withstand extreme weather events is crucial. This includes improving drainage systems in urban areas to prevent flooding and the subsequent de- struction of homes, property and income-generating infrastructure. In rural areas, the focus should be on improving water storage and irrigation systems to mitigate the impact of drought 28 shocks, that proved to be poverty-enhancing for agricultural households. Investing in and im- proving early warning systems for extreme weather events, such as heatwaves and floods, is crucial. Alerting the population and ensuring institutions are prepared to respond can reduce vulnerability and economic impact, even in the face of extreme temperatures and precipitation, enabling timely protection. Such systems could even be connected to social assistance programs and allow for the scaling up of support in times of weather shocks, based on the example of the PSNP program in Ethiopia (Sabates-Wheeler, Hirvonen, Lind, & Hoddinott, 2022). 29 8 References Aggarwal, R. (2021). Impacts of climate shocks on household consumption and inequality in India. Environment and Development Economics , 26 (5-6), 488–511. Ahmed, S. A., Diffenbaugh, N. S., & Hertel, T. W. (2009). Climate volatility deepens poverty vulnerability in developing countries. Environmental Research Letters , 4 (3), 034004. Alpízar, F., Saborío-Rodríguez, M., Martínez-Rodríguez, M. R., Viguera, B., Vignola, R., Capitán, T., & Harvey, C. A. (2020). Determinants of food insecurity among small- holder farmer households in Central America: Recurrent versus extreme weather-driven events. Regional Environmental Change , 20 , 1–16. Angelsen, A., & Dokken, T. (2018). Climate exposure, vulnerability and environmental re- liance: A cross-section analysis of structural and stochastic poverty. Environment and Development Economics , 23 (3), 257–278. Ansah, I. G. K., Gardebroek, C., & Ihle, R. (2021). Shock interactions, coping strategy choices and household food security. Climate and Development , 13 (5), 414–426. Aragón, F. M., Oteiza, F., & Rud, J. P. (2021). Climate change and agriculture: Subsistence farmers’ response to extreme heat. American Economic Journal: Economic Policy , 13 (1), 1–35. Asfaw, S., & Maggio, G. (2018). Gender, weather shocks and welfare: Evidence from Malawi. The Journal of Development Studies , 54 (2), 271–291. Auffhammer, M., Hsiang, S. M., Schlenker, W., & Sobel, A. (2013). Using weather data and climate model output in economic analyses of climate change. Review of Environmental Economics and Policy . Azzarri, C., & Signorelli, S. (2020). Climate and poverty in Africa South of the Sahara. World development , 125 , 104691. Bandyopadhyay, S., & Skoufias, E. (2015). Rainfall variability, occupational choice, and welfare in rural Bangladesh. Review of Economics of the Household , 13 , 589–634. Béné, C. (2009). Are fishers poor or vulnerable? Assessing economic vulnerability in small-scale fishing communities. The Journal of Development Studies , 45 (6), 911–933. Benitez Rodriguez, A. M., Wolf, R., Trotter, I. M., & Gurgel, A. C. (2023). Assessing the economic implications of climate change impacts on the Paraguayan agricultural sector. Climate and Development , 1–8. Bilal, A., & Känzig, D. R. (2024). The Macroeconomic Impact of Climate Change: Global vs. Local Temperature (Tech. Rep.). National Bureau of Economic Research, Inc. 30 Burbidge, J. B., Magee, L., & Robb, A. L. (1988). Alternative transformations to handle extreme values of the dependent variable. Journal of the American statistical Association , 83 (401), 123–127. Carter, M. R., Little, P. D., Mogues, T., & Negatu, W. (2007). Poverty traps and natural disasters in Ethiopia and Honduras. World Development , 35 (5), 835–856. Cavallo, E., Hoffmann, B., & Noy, I. (2023). Disasters and climate change in Latin America and the Caribbean: An introduction to the special issue. Economics of Disasters and Climate Change , 7 (2), 135–145. CEPAL. (2014). La economía del cambio climático en el Paraguay. ((LC/W.617), Santiago de Chile) Conte Grand, M., Schulz-Antipa, P., García-Witulski, C., & Rabasssa, M. (2024). From rising temperature to rising health concerns: A study of climate change effects in Paraguay. Coronel, G., Pastén, M., Báez, J., Monte Domecq, R., Bidegain, M., & Nagy, G. J. (2015). Improving capacities and communication on climate threats for water resources adaptation in Paraguay. Handbook of Climate Change Adaptation; Leal Filho, W., Ed.; Springer: Berlin, Germany , 1091–1108. D’Agostino, A. L., & Schlenker, W. (2016). Recent weather fluctuations and agricultural yields: Implications for climate change. Agricultural Economics , 47 (S1), 159–171. Dallmann, I., & Millock, K. (2017). Climate variability and inter-state migration in India. CESifo Economic Studies , 63 (4), 560–594. Danandeh Mehr, A., Sorman, A. U., Kahya, E., & Hesami Afshar, M. (2020). Climate change impacts on meteorological drought using SPI and SPEI: case study of Ankara, Turkey. Hydrological Sciences Journal , 65 (2), 254–268. Dang, H.-A. H., Cong Nguyen, M., & Trinh, T.-A. (2023). Does hotter temperature increase poverty and inequality? Global evidence from subnational data analysis. International Inequalities Institute, London School of Economics. de Catastro, S. N. (2022). Estudio de los límites distritales de la República del Paraguay. Retrieved from https://www.catastro.gov.py/public/967e7c_ESTUDIO% 20DE%20LÃŊMITES%20DISTRITALES%20DE%20LA%20REPÃŽBLICA%20DEL%20PARAGUAY.pdf. Dell, M., Jones, B. F., & Olken, B. A. (2012). Temperature shocks and economic growth: Evidence from the last half century. American Economic Journal: Macroeconomics , 4 (3), 66–95. Del Ninno, C., Dorosh, P. A., & Smith, L. C. (2003). Public policy, markets and household coping strategies in Bangladesh: Avoiding a food security crisis following the 1998 floods. 31 World Development , 31 (7), 1221–1238. Dercon, S. (2004). Growth and shocks: evidence from rural Ethiopia. Journal of Development Economics , 74 (2), 309–329. Deryugina, T., & Hsiang, S. M. (2014). Does the environment still matter? Daily temperature and income in the United States (Tech. Rep.). National Bureau of Economic Research. Dirreción General de Estadística, Encuestas y Censos. (2012). Metodología para la estimación de las líneas de pobreza. Gobierno Nacional. Eckstein, D., Künzel, V., & Schäfer, L. (2021). Global climate risk index 2021, who suffers most from extreme weather events? Weather-related loss events in 2019 and 2000–2019. Germanwatch e.V., Germany. Elkadi, A., Woning, M., Bles, T., Abraham, G., Casares, A., Sethi, K., & Flor, L. (2019). Climate-resilient roads in Paraguay; Mapping the risks and advising adaptive mitigation measures. Proceedings of the XVII ECSMGE-2019 . Erman, A. E., Tariverdi, M., Obolensky, M. A. B., Chen, X., Vincent, R. C., Malgioglio, S., . . . Yoshida, N. (2019). Wading out the storm: The role of poverty in exposure, vulner- ability and resilience to floods in Dar Es Salaam. World Bank Policy Research Working Paper (8976). Ervin, P. A., & Gayoso de Ervin, L. (2019). Household vulnerability to food insecurity in the face of climate change in Paraguay. FAO Agricultural Development Economics Working Paper 19-04. Rome, FAO. FAO. (2023). Sistema mundial de información y alerta: Resúmenes informativos por países Paraguay. Retrieved from https://www.fao.org/giews/countrybrief/country.jsp ?lang=es&code=PRY. Organización de las Naciones Uidas para la Alimentación y la Agri- cultura. Felbermayr, G., Gröschl, J., Sanders, M., Schippers, V., & Steinwachs, T. (2022). The economic impact of weather anomalies. World Development , 151 , 105745. Funk, C. C., Peterson, P. J., Landsfeld, M. F., Pedreros, D. H., Verdin, J. P., Rowland, J. D., . . . others (2014). A quasi-global precipitation time series for drought monitoring. US Geological Survey Data Series , 832 (4), 1–12. Gao, J., & Mills, B. F. (2018). Weather shocks, coping strategies, and consumption dynamics in rural Ethiopia. World Development , 101 , 268–283. Gayoso de Ervin, L., & Rubiano, E. (2023). Climate risk and poverty: Prevention is an invest- ment. Retrieved from https://blogs.worldbank.org/en/latinamerica/climate-risk -and-poverty-prevention-investment. 32 Gouel, C., & Laborde, D. (2021). The crucial role of domestic and international market-mediated adaptation to climate change. Journal of Environmental Economics and Management , 106 , 102408. GRID-Genever European Comission. (2023). Paraguay country profile. Retrieved from https:// dicf.unepgrid.ch/paraguay. Gómez Gómez, R. E., Kim, J., Hong, K., Jang, J. Y., Kisiju, T., Kim, S., & Chun, B. C. (2022). Association between climate factors and dengue fever in Asunción, Paraguay: A generalized additive model. International Journal of Environmental Research and Public Health , 19 (19), 12192. Hallegatte, S. (2016). Shock waves: Managing the impacts of climate change on poverty. World Bank Publications. Hallegatte, S., Fay, M., & Barbier, E. B. (2018). Poverty and climate change: Introduction. Environment and Development Economics , 23 (3), 217–233. Hallegatte, S., & Rozenberg, J. (2017). Climate change through a poverty lens. Nature Climate Change , 7 (4), 250–256. Hallegatte, S., Vogt-Schilb, A., Rozenberg, J., Bangalore, M., & Beaudet, C. (2020). From poverty to disaster and back: A review of the literature. Economics of Disasters and Climate Change , 4 , 223–247. Hardoy, J., & Lankao, P. R. (2011). Latin American cities and climate change: Challenges and options to mitigation and adaptation responses. Current Opinion in Environmental Sustainability , 3 (3), 158–163. Hardoy, J., & Pandiella, G. (2009). Urban poverty and vulnerability to climate change in Latin America. Environment and Urbanization , 21 (1), 203–224. Harvey, C. A., Saborio-Rodríguez, M., Martinez-Rodríguez, M. R., Viguera, B., Chain- Guadarrama, A., Vignola, R., & Alpizar, F. (2018). Climate change impacts and adap- tation among smallholder farmers in Central America. Agriculture & Food Security , 7 (1), 1–20. He, X., & Chen, Z. (2022). Weather, cropland expansion, and deforestation in Ethiopia. Journal of Environmental Economics and Management , 111 , 102586. Heim, R. R. (2002). A review of twentieth-century drought indices used in the United States. Bulletin of the American Meteorological Society , 83 (8), 1149–1166. Hertel, T. W., Burke, M. B., & Lobell, D. B. (2010). The poverty implications of climate-induced crop yield changes by 2030. Global Environmental Change , 20 (4), 577–585. Hertel, T. W., & Rosch, S. D. (2010). Climate change, agriculture, and poverty. Applied 33 Economic Perspectives and Policy , 32 (3), 355–385. Hirvonen, K. (2016). Temperature changes, household consumption, and internal migration: Evidence from Tanzania. American Journal of Agricultural Economics , 98 (4), 1230–1249. INE. (n.d.-a). Población ocupada por año y trimestre, según área de residencia y sector económico de la ocupación principal. Instituto Nacional de Estadística. Retrieved from https:// www.ine.gov.py/publicacion/3/empleo. INE. (n.d.-b). Principales indicadores de pobreza de la población por año de la encuesta, según área de residencia. Periodo 1997/98 - 2022. Instituto Nacional de Estadística. Retrieved from https://www.ine.gov.py/default.php?publicacion=4. INE. (2020). Principales resultados de pobreza monetaria y distribución de ingresos EPHC 2019. Instituto Nacional de Estadística. INE. (2021). Proyecciones de población nacional, áreas urbana y rural, por sexo y edad. Retrieved from https://www.ine.gov.py/Publicaciones/Biblioteca/documento/7eb5 _Paraguay_2021.pdf. (Instituto Nacional de Estadística) Instituto Nacional de Estadística. (n.d.). Códigos geográficos: Censo nacional de población y viviendas 2012 - Paraguay. Retrieved from https://www.datos.gov.py/ dataset/cÃşdigos-geogrÃąficos-censo-nacional-de-poblaciÃşn-y-viviendas-2012 -paraguay. Instituto Nacional de Estadística. (2022). Boletín trimestral de empleo – 2022. Gobierno Nacional. Instituto Nacional de Estadística. (2024). Principales resultados de pobreza monetaria y dis- tribución de ingresos – 2023. Retrieved from https://www.ine.gov.py/Publicaciones/ Biblioteca/documento/246/Pobreza%20Monetaria_%20EPHC%202023_INE..pdf. IPCC. (2021). Climate change 2021: The physical science basis. Contribution of working group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (V. Masson-Delmotte et al., Eds.). Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press. Retrieved from https://www.ipcc.ch/report/ar6/wg1/ doi: 10.1017/9781009157896 Ishizawa, O., & Miranda, J. J. (2016). Weathering storms: Understanding the impact of natural disasters on the poor in Central America. World Bank Policy Research Working Paper (7692). Krishna, A. (2006). Pathways out of and into poverty in 36 villages of Andhra Pradesh, India. World Development , 34 (2), 271–288. Lachaud, M. A., Bravo-Ureta, B. E., & Ludena, C. E. (2022). Economic effects of climate change 34 on agricultural production and productivity in Latin America and the Caribbean (LAC). Agricultural Economics , 53 (2), 321–332. Letta, M., Montalbano, P., & Tol, R. S. (2018). Temperature shocks, short-term growth and poverty thresholds: Evidence from rural Tanzania. World Development , 112 , 13–32. Lovino, M. A., Pierrestegui, M. J., Müller, O. V., Berbery, E. H., Müller, G. V., & Pasten, M. (2021). Evaluation of historical CMIP6 model simulations and future projections of temperature and precipitation in Paraguay. Climatic Change , 164 , 1–24. Magrin, G., Marengo, J., Boulanger, J.-P., Buckeridge, M. S., Castellano, E., Poveda, G., . . . others (2014). Central and South America. In: Barros VR, Field CB, Dokken DJ et al. (eds). Climate change 2014: Impacts, adaptation, and vulnerability. Part B: Regional aspects. Contribution of Working Group II to the Fifth Assessment Report of the In- tergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, pp. 1499–1566. Metcalfe, S. E., Schmook, B., Boyd, D. S., De la Barreda-Bautista, B., Endfield, G. E., Mardero, S., . . . others (2020). Community perception, adaptation and resilience to extreme weather in the Yucatan Peninsula, Mexico. Regional Environmental Change , 20 , 1–15. Molua, E. L., & Ayuk, J. E. (2021). Male–female sensitivity in climate-induced income insecurity: Some empirical evidence from farming households in Northern Cameroon. Development in Practice , 31 (8), 1014–1039. Muñoz Sabater, J. (2019). ERA5-land monthly averaged data from 1981 to present. (Copernicus Climate Change Service (C3S) Climate Data Store (CDS).) Nagy, G. J., Filho, W. L., Azeiteiro, U. M., Heimfarth, J., Verocai, J. E., & Li, C. (2018). An assessment of the relationships between extreme weather events, vulnerability, and the impacts on human wellbeing in Latin America. International journal of environmental research and public health , 15 (9), 1802. Ngoma, H., Finn, A., & Kabisa, M. (2024). Climate shocks, vulnerability, resilience and liveli- hoods in rural Zambia. Climate and Development , 16 (6), 490–501. Nguyen, T.-T., Nguyen, T. T., & Grote, U. (2020). Multiple shocks and households’ choice of coping strategies in rural Cambodia. Ecological Economics , 167 , 106442. OECD. (2022). Latin American Economic Outlook 2022: Towards a Green and Just Transition. OECD Publishing, Paris. https://doi.org/10.1787/3d5554fc-en. Olsson, L., Opondo, M., Tschakert, P., Agrawal, A., Eriksen, S., Ma, S., . . . Zakieldeen, S. (2014). Livelihoods and poverty. In C. Field et al. (Eds.), Climate Change 2014: Impacts, Adaptation, and Vulnerability. part A: Global and Sectoral Aspects. Contribution of Work- 35 ing Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 793-832. Otto, I. M., Reckien, D., Reyer, C. P., Marcus, R., Le Masson, V., Jones, L., . . . Serdeczny, O. (2017). Social vulnerability to climate change: A review of concepts and evidence. Regional Environmental Change , 17 , 1651–1662. Patel, N., Chopra, P., & Dadhwal, V. (2007). Analyzing spatial patterns of meteorological drought using Standardized Precipitation Index. Meteorological Applications: A journal of forecasting, practical applications, training techniques and modelling , 14 (4), 329–336. Perez, C., Jones, E., Kristjanson, P., Cramer, L., Thornton, P. K., Förch, W., & Barahona, C. a. (2015). How resilient are farming households and communities to a changing climate in Africa? A gender-based perspective. Global Environmental Change , 34 , 95–107. Randell, H., & Gray, C. (2016). Climate variability and educational attainment: Evidence from rural Ethiopia. Global environmental change , 41 , 111–123. Ravallion, M. (2007). Urban poverty. Finance and Development , 44 (3), 15–17. Reckien, D., Creutzig, F., Fernandez, B., Lwasa, S., Tovar-Restrepo, M., Mcevoy, D., & Sat- terthwaite, D. (2017). Climate change, equity and the sustainable development goals: An urban perspective. Environment and urbanization , 29 (1), 159–182. Revi, A., Satterthwaite, D., Aragón-Durand, F., Corfee-Morlot, J., Kiunsi, R., Pelling, M., . . . Solecki, W. (2014). Urban areas. In C. Field et al. (Eds.), Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 535-612. Reyer, C. P., Adams, S., Albrecht, T., Baarsch, F., Boit, A., Canales Trujillo, N., . . . others (2017). Climate change impacts in Latin America and the Caribbean and their implications for development. Regional Environmental Change , 17 , 1601–1621. Rozenberg, J., Dborkin, D. V., Giuliano, F. M., Jooste, C., Mikou, M., Rodriguez Chamussy, L., . . . Walsh, B. J. (2021). Argentina - poverty and macro economic impacts of climate shocks (Tech. Rep.). Washington, D.C.: World Bank Group. Sabates-Wheeler, R., Hirvonen, K., Lind, J., & Hoddinott, J. (2022). Expanding social pro- tection coverage with humanitarian aid: Lessons on targeting and transfer values from Ethiopia. The Journal of Development Studies , 58 (10), 1981–2000. Salimi, H., Asadi, E., & Darbandi, S. (2021). Meteorological and hydrological drought monitor- 36 ing using several drought indices. Applied Water Science , 11 , 1–10. Schlenker, W., Hanemann, W. M., & Fisher, A. C. (2007). Water availability, degree days, and the potential impact of climate change on irrigated agriculture in California. Climatic Change , 81 (1), 19–38. Sedova, B., & Kalkuhl, M. (2020). Who are the climate migrants and where do they go? Evidence from rural India. World Development , 129 , 104848. Sen, B. (2003). Drivers of escape and descent: changing household fortunes in rural Bangladesh. World Development , 31 (3), 513–534. Service, C. C. C. (2017). ERA5: Fifth generation of ECMWF atmospheric reanalyses of the global climate. (Copernicus Climate Change Service (C3S) Climate Data Store (CDS)) Silvero, F., Lops, C., Montelpare, S., & Rodrigues, F. (2019). Impact assessment of climate change on buildings in Paraguay—overheating risk under different future climate scenarios. In Building Simulation (Vol. 12, pp. 943–960). Somanathan, E., Somanathan, R., Sudarshan, A., & Tewari, M. (2021). The impact of temper- ature on productivity and labor supply: Evidence from Indian manufacturing. Journal of Political Economy , 129 (6), 1797–1827. Stokes, E., Lauff, C., Eldridge, E., Ortbal, K., Nassar, A., & Mehta, K. (2015). Income generating activities of rural Kenyan women. Journal of Sustainable Development , 8 (8), 42. Torelló-Sentelles, H., & Franzke, C. L. (2022). Drought impact links to meteorological drought indicators and predictability in Spain. Hydrology and Earth System Sciences , 26 (7), 1821– 1844. Wang, D., Jenkins, K., Forstenhäusler, N., Lei, T., Price, J., Warren, R., . . . Guan, D. (2021). Economic impacts of climate-induced crop yield changes: Evidence from agri-food indus- tries in six countries. Climatic Change , 166 (3), 30. Wesz Junior, V. J. (2022). Soybean production in Paraguay: Agribusiness, economic change and agrarian transformations. Journal of Agrarian Change , 22 (2), 317–340. World Bank. (n.d.-a). Paraguay: Climate Change Knowledge Portal. Retrieved from https:// climateknowledgeportal.worldbank.org/country/paraguay/vulnerability. World Bank. (n.d.-b). The World Bank in Paraguay. Retrieved from https://www.worldbank .org/en/country/paraguay/overview. World Bank. (n.d.-c). World Development Indicators. Retrieved from https://databank .worldbank.org/source/world-development-indicators. World Bank. (2021). Climate Risk Profile: Paraguay. The World Bank Group, Washington, 37 DC. World Bank. (2023). Poverty & Equity Brief Paraguay, April 2023. Retrieved from https://databankfiles.worldbank.org/public/ddpext_download/poverty/ 987B9C90-CB9F-4D93-AE8C-750588BF00QA/current/Global_POVEQ_PRY.pdf. World Bank. (2024, Forthcoming ). Paraguay Country Climate Development Report. Executive Summary. World Bank. Zapata, O. (2023). Weather Disasters, Material Losses and Income Inequality: Evidence from a Tropical, Middle-Income Country. Economics of Disasters and Climate Change , 7 (2), 231–251. Zeleke, T., Beyene, F., Deressa, T., Yousuf, J., & Kebede, T. (2021). Vulnerability of smallholder farmers to climate change-induced shocks in East Hararghe Zone, Ethiopia. Sustainability , 13 (4), 2162. Zuñiga, R. A. A., Lima, G. N., & Villoria, A. M. G. (2021). Impact of slow-onset events related to Climate Change on food security in Latin America and the Caribbean. Current Opinion in Environmental Sustainability , 50 , 215–224. 38 A Appendix Table A.1: Sample: Encuesta Permanente de Hogares Year Nr. Districts Urban HH Rural HH Urban Ind. Rural Ind 2004 210 4,125 3,698 17,455 17,181 2005 158 2,568 1,896 10,674 8,905 2006 179 2,857 2,435 11,813 10,920 2007 163 2,771 2,041 11,466 9,587 2008 154 2,615 1,986 10,694 8,772 2009 156 2,478 1,961 9,791 8,628 2010 162 3,048 1,955 12,211 8,264 2011 147 3,137 1,757 12,389 7,351 2012 160 3,434 1,854 13,503 7,648 2013 154 3,507 1,917 13,646 7,561 2014 157 3,337 1,828 13,100 7,172 2015 199 5,352 2,877 19,932 10,966 2016 225 5,584 4,635 20,669 17,145 2017 220 5,175 4,395 18,908 16,206 2018 165 2,715 2,284 9,939 8,557 2019 165 2,764 2,335 9,812 8,355 Total 55,467 39,854 216,002 163,168 Note: HH = Household, Ind = Individuals. 39 Table A.2: Effects of weather shocks on income and poverty Total income Labor income Poverty Extreme poverty (1) (2) (3) (4) (5) (6) (7) (8) Urban Rural Urban Rural Urban Rural Urban Rural Positive temperature anomaly -0.050∗∗∗ -0.070∗∗∗ -0.046∗∗ -0.121∗∗∗ 0.061∗∗ 0.107∗∗∗ 0.166∗∗∗ 0.073∗∗ (Heat shock) (0.009) (0.011) (0.020) (0.022) (0.026) (0.025) (0.047) (0.030) Negative temperature anomaly 0.056∗∗∗ 0.050∗∗∗ 0.043∗ 0.077∗∗∗ -0.094∗∗∗ -0.070∗∗ -0.108∗∗ -0.114∗∗∗ (Cold shock) (0.010) (0.013) (0.022) (0.025) (0.029) (0.028) (0.054) (0.032) Positive precipitation anomaly -0.005 0.071∗∗∗ 0.043∗∗ 0.069∗∗∗ 0.070∗∗ -0.052∗ -0.074 -0.090∗∗∗ (Flood shock) (0.009) (0.012) (0.020) (0.025) (0.028) (0.028) (0.049) (0.035) Negative precipitation anomaly 0.028∗∗∗ -0.088∗∗∗ -0.018 -0.118∗∗∗ -0.086∗∗∗ 0.051 0.073 0.103∗∗∗ (Drought shock) (0.010) (0.013) (0.022) (0.026) (0.030) (0.031) (0.055) (0.038) Average temperature 0.067∗∗∗ 0.074∗∗∗ 0.068∗∗∗ 0.126∗∗∗ -0.090∗∗∗ -0.123∗∗∗ -0.253∗∗∗ -0.055 (0.010) (0.012) (0.021) (0.024) (0.029) (0.027) (0.054) (0.033) -0.049∗∗∗ -0.063∗∗∗ -0.047∗∗ 0.073∗∗∗ 40 Average precipitation 0.003 -0.020 0.024 0.026 (0.007) (0.009) (0.015) (0.017) (0.021) (0.020) (0.036) (0.024) Historical average temperature -0.037∗∗∗ -0.042∗∗∗ -0.022 -0.081∗∗∗ 0.061∗∗ 0.070∗∗ 0.222∗∗∗ 0.033 (0.010) (0.013) (0.022) (0.025) (0.030) (0.028) (0.058) (0.035) Historical average precipitation -0.033∗∗∗ -0.018∗ -0.011 0.016 0.082∗∗∗ 0.085∗∗∗ 0.145∗∗∗ 0.040 (0.008) (0.010) (0.018) (0.020) (0.024) (0.024) (0.039) (0.029) Household controls Yes Yes Yes Yes Yes Yes Yes Yes Year-department dummies Yes Yes Yes Yes Yes Yes Yes Yes Month dummies Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.54 0.44 0.36 0.32 Pseudo R-squared 0.34 0.27 0.28 0.22 N 216,002 163,168 216,002 163,168 216,002 163,168 216,002 163,168 Standard errors in parentheses ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01 Note: The time period covers 90-day bin before each household’s interview. Household controls include: household size, dependency ratio, age of household head, age of household head squared, years of schooling of household head, and dummy variables if the household head is female, if the head and spouse only speak Guaraní, the economic sector of the household head (primary, secondary, tertiary, unemployed), whether the household is engaged in agriculture, and whether the household has a car, heating or microwave. Table A.3: Marginal effects of weather shocks on income and poverty Total income Labor income Poverty Extreme poverty (1) (2) (3) (4) (5) (6) (7) (8) Urban Rural Urban Rural Urban Rural Urban Rural Positive temperature anomaly (Heat shock) -0.050*** -0.070*** -0.046** -0.121*** 0.017** 0.042*** 0.004*** 0.012** (0.009) (0.011) (0.020) (0.022) (0.007) (0.010) (0.001) (0.005) Negative temperature anomaly (Cold shock) 0.056*** 0.050*** 0.043** 0.077*** -0.026*** -0.028** -0.002** -0.018*** (0.010) (0.013) (0.022) (0.025) (0.008) (0.011) (0.001) (0.005) 41 Positive precipitation anomaly (Flood shock) -0.005 0.072*** 0.043** 0.070*** 0.019** -0.021* -0.002 -0.014** (0.009) (0.012) (0.020) (0.025) (0.008) (0.011) (0.001) (0.006) Negative precipitation anomaly (Drought shock) 0.028*** -0.088*** -0.018 -0.118*** -0.023*** 0.020 0.002 0.017*** (0.010) (0.013) (0.022) (0.026) (0.008) (0.012) (0.001) (0.006) N 216,002 163,168 216,002 163,168 216,002 163,168 216,002 163,163 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: The time period covers 90-day bin before each household’s interview. Marginal effects estimated using Stata’s margins, dydx(*) command. Table A.4: Robustness checks: Effects of weather shocks on total income 30 days before interview 180 days before interview Without outliers (1) (2) (3) (4) (5) (6) Urban Rural Urban Rural Urban Rural Positive temp. anomaly -0.006 0.014 -0.067∗∗∗ -0.080∗∗∗ -0.047∗∗∗ -0.056∗∗∗ (Heat shock) (0.008) (0.012) (0.019) (0.023) (0.008) (0.011) Negative temp. anomaly 0.013 -0.033∗∗ 0.051∗∗ 0.044 0.043∗∗∗ 0.035∗∗∗ (Cold shock) (0.010) (0.014) (0.025) (0.029) (0.010) (0.012) Positive prec. anomaly 0.036∗∗∗ 0.034∗∗∗ -0.010 0.089∗∗∗ -0.001 0.079∗∗∗ (Flood shock) (0.006) (0.010) (0.009) (0.012) (0.009) (0.011) Negative prec. anomaly 0.011 -0.030∗∗∗ 0.024∗∗ -0.050∗∗∗ 0.022∗∗ -0.081∗∗∗ (Drought shock) (0.007) (0.011) (0.010) (0.015) (0.009) (0.012) Average temperature 0.013∗ -0.013 0.093∗∗∗ 0.085∗∗∗ 0.065∗∗∗ 0.064∗∗∗ (0.007) (0.011) (0.026) (0.031) (0.009) (0.012) Average precipitation -0.014∗∗∗ -0.013∗∗∗ 0.016 -0.085∗∗∗ -0.002 -0.052∗∗∗ (0.003) (0.004) (0.011) (0.014) (0.007) (0.008) Historical average temp. 0.004 0.048∗∗∗ -0.057∗∗ -0.064∗∗ -0.039∗∗∗ -0.036∗∗∗ (0.007) (0.011) (0.026) (0.031) (0.010) (0.012) Historical average prec. 0.001 -0.024∗∗∗ -0.042∗∗∗ -0.008 -0.023∗∗∗ -0.010 (0.004) (0.006) (0.014) (0.017) (0.008) (0.010) Household controls Yes Yes Yes Yes Yes Yes Year-department dummies Yes Yes Yes Yes Yes Yes Month dummies Yes Yes Yes Yes Yes Yes R-squared 0.54 0.44 0.54 0.44 0.54 0.44 N 216,002 163,168 216,002 163,168 211,858 159,899 Standard errors in parentheses ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01 Note: The time periods refer to the number of days before each household’s interview. In columns (5) and (6), we exclude outliers below the 1s t and above the 99t h percentile of the income distribution. Household controls include: household size, dependency ratio, age of household head, age of household head squared, years of schooling of household head, and dummy variables if the household head is female, if the head and spouse only speak Guaraní, the economic sector of the household head (primary, secondary, tertiary, unemployed), whether the household is engaged in agriculture, and whether the household has a car, heating or microwave. 42 Table A.5: Marginal effects of weather shocks on total income, robustness checks 30 days before interview 180 days before interview Without outliers (1) (2) (3) (4) (5) (6) Urban Rural Urban Rural Urban Rural Positive temp. anomaly -0.006 0.014 -0.067*** -0.080*** -0.047*** -0.056*** (Heat shock) (0.008) (0.012) (0.019) (0.023) (0.008) (0.011) Negative temp. anomaly 0.013 -0.033** 0.051** 0.044 0.043*** 0.035*** (Cold shock) (0.010) (0.014) (0.025) (0.029) (0.010) (0.012) Positive prec. anomaly 0.036*** 0.034*** -0.010 0.089*** -0.001 0.079*** (Flood shock) (0.006) (0.010) (0.009) (0.012) (0.009) (0.011) Negative prec. anomaly 0.011 -0.030*** 0.024** -0.050*** 0.022** -0.081*** (Drought shock) (0.007) (0.011) (0.010) (0.015) (0.009) (0.012) N 216,002 163,168 216,002 163,168 211,858 159,899 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: The time period in columns (5) and (6) covers 90-day bin before each household’s interview. Marginal effects estimated using Stata’s margins, dydx(*) command. 43 Table A.6: Robustness checks: Effects of weather shocks on poverty 30 days before interview 180 days before interview Without outliers (1) (2) (3) (4) (5) (6) Urban Rural Urban Rural Urban Rural Positive temp. anomaly -0.006 -0.066∗∗ 0.073 0.139∗∗∗ 0.061∗∗ 0.104∗∗∗ (Heat shock) (0.025) (0.029) (0.055) (0.052) (0.026) (0.025) Negative temp. anomaly 0.009 0.095∗∗∗ 0.078 -0.039 -0.087∗∗∗ -0.068∗∗ (Cold shock) (0.028) (0.031) (0.066) (0.064) (0.029) (0.028) Positive prec. anomaly -0.058∗∗∗ -0.058∗∗∗ -0.021 -0.141∗∗∗ 0.065∗∗ -0.057∗∗ (Flood shock) (0.021) (0.022) (0.028) (0.028) (0.028) (0.028) Negative prec. anomaly 0.012 0.064∗∗ -0.043 0.078∗∗ -0.073∗∗ 0.049 (Drought shock) (0.024) (0.025) (0.032) (0.034) (0.030) (0.031) Average temperature 0.007 0.054∗∗ -0.104 -0.082 -0.090∗∗∗ -0.121∗∗∗ (0.021) (0.025) (0.075) (0.071) (0.029) (0.027) Average precipitation 0.031∗∗∗ 0.024∗∗∗ -0.008 0.133∗∗∗ -0.041∗ 0.027 (0.009) (0.009) (0.033) (0.031) (0.021) (0.020) Historical average temp. -0.015 -0.120∗∗∗ 0.078 0.043 0.063∗∗ 0.069∗∗ (0.023) (0.026) (0.076) (0.072) (0.031) (0.029) Historical average prec. -0.021∗ 0.022∗ 0.051 0.036 0.072∗∗∗ 0.081∗∗∗ (0.011) (0.013) (0.039) (0.037) (0.024) (0.024) Household controls Yes Yes Yes Yes Yes Yes Year-department dummies Yes Yes Yes Yes Yes Yes Month dummies Yes Yes Yes Yes Yes Yes Pseudo R-squared 0.34 0.27 0.34 0.27 0.33 0.26 N 216,002 163,168 216,002 163,168 211,858 159,899 Standard errors in parentheses ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01 Note: The time periods refer to the number of days before each household’s interview. In columns (5) and (6), we exclude outliers below the 1s t and above the 99t h percentile of the income distribution. Household controls include: household size, dependency ratio, age of household head, age of household head squared, years of schooling of household head, and dummy variables if the household head is female, if the head and spouse only speak Guaraní, the economic sector of the household head (primary, secondary, tertiary, unemployed), whether the household is engaged in agriculture, and whether the household has a car, heating or microwave. 44 Table A.7: Marginal effects of weather shocks on poverty, robustness checks 30 days before interview 180 days before interview Without outliers (1) (2) (3) (4) (5) (6) Urban Rural Urban Rural Urban Rural Positive temp. anomaly -0.002 -0.026** 0.020 0.055*** 0.017** 0.041*** (Heat shock) (0.007) (0.011) (0.015) (0.020) (0.007) (0.010) Negative temp. anomaly 0.002 0.037*** 0.021 -0.016 -0.024*** -0.027** (Cold shock) (0.008) (0.012) (0.018) (0.025) (0.008) (0.011) Positive prec. anomaly -0.016*** -0.023*** -0.006 -0.056*** 0.018** -0.023** (Flood shock) (0.006) (0.009) (0.008) (0.011) (0.008) (0.011) Negative prec. anomaly 0.003 0.025** -0.012 0.031** -0.020** 0.019 (Drought shock) (0.007) (0.010) (0.009) (0.014) (0.008) (0.012) N 216,002 163,168 216,002 163,168 211,858 159,899 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: The time period in columns (5) and (6) covers 90-day bin before each household’s interview. Marginal effects estimated using Stata’s margins, dydx(*) command. 45 Table A.8: Effects of weather shocks on poverty by department (1) (2) (3) (4) (5) (6) (7) (8) Asunción San Pedro Caaguazú Itapúa Alto Parana Central Other Urban Other Rural Positive temperature anomaly -0.175∗ 0.039 0.077∗ 0.195∗∗∗ 0.039 0.134∗∗∗ -0.038 0.062 (Heat shock) (0.106) (0.053) (0.044) (0.052) (0.045) (0.046) (0.051) (0.040) Negative temperature anomaly 0.070 -0.061 -0.111∗∗ -0.143∗∗∗ 0.008 -0.223∗∗∗ -0.011 -0.043 (Cold shock) (0.109) (0.054) (0.049) (0.053) (0.046) (0.057) (0.058) (0.048) Positive precipitation anomaly -0.451∗∗∗ -0.303∗∗∗ 0.139 -0.141∗∗ 0.263∗ 0.324∗∗∗ 0.010 0.030 (Flood shock) (0.159) (0.060) (0.107) (0.066) (0.136) (0.063) (0.046) (0.041) Negative precipitation anomaly 0.268∗ 0.118∗∗ -0.078 -0.130∗∗ -0.305∗∗ -0.230∗∗∗ -0.006 0.160∗∗∗ (Drought shock) (0.140) (0.058) (0.097) (0.063) (0.141) (0.051) (0.051) (0.050) Average temperature 0.094 -0.039 -0.035 -0.162∗∗ -0.029 -0.197∗∗∗ -0.012 -0.129∗∗∗ (0.105) (0.051) (0.045) (0.064) (0.053) (0.051) (0.055) (0.045) Average precipitation 0.390∗∗∗ 0.235∗∗∗ -0.069 -0.007 -0.145∗ -0.259∗∗∗ 0.000 0.004 (0.144) (0.041) (0.065) (0.042) (0.081) (0.053) (0.034) (0.031) 46 Historical average temperature 0.091 -0.081 0.015 0.434∗∗∗ 0.071 0.164∗∗∗ -0.037 0.096∗∗ (0.133) (0.054) (0.051) (0.068) (0.058) (0.054) (0.058) (0.047) Historical average precipitation -0.822∗∗∗ -0.174∗∗∗ -0.060 -0.485∗∗∗ -0.034 0.307∗∗∗ -0.009 0.189∗∗∗ (0.258) (0.049) (0.075) (0.057) (0.098) (0.059) (0.041) (0.037) HH controls Yes Yes Yes Yes Yes Yes Yes Yes Year-dep. dum. Yes Yes Yes Yes Yes Yes Yes Yes Month dummies Yes Yes Yes Yes Yes Yes Yes Yes Pseudo R-sq. 0.37 0.26 0.27 0.30 0.31 0.29 0.32 0.24 N 36,868 33,613 39,115 34,605 54,460 57,613 54,886 68,010 Standard errors in parentheses ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01 Note: The time period covers 90-day bin before each household’s interview. Household controls include: household size, dependency ratio, age of household head, age of household head squared, years of schooling of household head, and dummy variables if the household head is female, if the head and spouse only speak Guaraní, the economic sector of the household head (primary, secondary, tertiary, unemployed), whether the household is engaged in agriculture, and whether the household has a car, heating or microwave. Table A.9: Marginal effects of weather shocks on poverty by department (1) (2) (3) (4) (5) (6) (7) (8) Asunción San Pedro Caaguazú Itapúa Alto Parana Central Other Urban Other Rural Positive temperature anomaly (Heat shock) -0.025* 0.015 0.031* 0.073*** 0.012 0.034*** -0.013 0.025 (0.015) (0.021) (0.018) (0.019) (0.014) (0.012) (0.017) (0.016) Negative temperature anomaly (Cold shock) 0.010 -0.024 -0.044** -0.054*** 0.003 -0.057*** -0.003 -0.017 (0.016) (0.022) (0.019) (0.020) (0.014) (0.015) (0.019) (0.019) Positive precipitation anomaly (Flood shock) -0.065*** -0.122*** 0.055 -0.053** 0.083* 0.083*** 0.004 0.012 47 (0.023) (0.024) (0.043) (0.025) (0.043) (0.016) (0.015) (0.016) Negative precipitation anomaly (Drought shock) 0.039* 0.048** -0.031 -0.049** -0.097** -0.059*** -0.002 0.064*** (0.020) (0.023) (0.039) (0.024) (0.045) (0.013) (0.017) (0.020) N 36,868 33,613 39,115 34,605 54,460 57,613 54,886 68,010 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: The time period covers 90-day bin before each household’s interview. Marginal effects estimated using Stata’s margins, dydx(*) command. Table A.10: Effects of weather shocks poverty, by gender of the household head Male household head Female household head (1) (2) (3) (4) Urban Rural Urban Rural Positive temperature anomaly 0.098∗∗∗ 0.098∗∗∗ -0.009 0.179∗∗∗ (Heat shock) (0.033) (0.029) (0.044) (0.048) Negative temperature anomaly -0.103∗∗∗ -0.055∗ -0.082∗ -0.158∗∗∗ (Cold shock) (0.036) (0.032) (0.050) (0.056) Positive precipitation anomaly 0.052 -0.055 0.113∗∗ -0.012 (Flood shock) (0.033) (0.034) (0.049) (0.051) Negative precipitation anomaly -0.031 0.121∗∗∗ -0.218∗∗∗ -0.102∗ (Drought shock) (0.037) (0.037) (0.054) (0.058) Average temperature -0.117∗∗∗ -0.110∗∗∗ -0.042 -0.200∗∗∗ (0.036) (0.032) (0.048) (0.053) Average precipitation -0.026 0.039 -0.093∗∗ -0.019 (0.025) (0.024) (0.038) (0.036) Historical average temperature 0.078∗∗ 0.059∗ 0.026 0.143∗∗∗ (0.038) (0.033) (0.051) (0.055) Historical average precipitation 0.052∗ 0.075∗∗∗ 0.148∗∗∗ 0.094∗∗ (0.029) (0.028) (0.043) (0.043) Household controls Yes Yes Yes Yes Year-department dummies Yes Yes Yes Yes Month dummies Yes Yes Yes Yes Pseudo R-squared 0.33 0.29 0.35 0.25 N 146,171 125,607 69,831 37,561 Standard errors in parentheses ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01 Note: The time period covers 90-day bin before each household’s interview. Household controls include: household size, dependency ratio, age of household head, age of household head squared, years of schooling of household head, and dummy variables if the household head is female, if the head and spouse only speak Guaraní, the economic sector of the household head (primary, secondary, tertiary, unemployed), whether the household is engaged in agriculture, and whether the household has a car, heating or microwave. 48 Table A.11: Marginal effects of weather shocks poverty, by gender of the household head Male household head Female household head (1) (2) (3) (4) Urban Rural Urban Rural Positive temperature anomaly (Heat shock) 0.026*** 0.039*** -0.003 0.071*** (0.009) (0.011) (0.013) (0.019) Negative temperature anomaly (Cold shock) -0.027*** -0.022* -0.025* -0.063*** (0.009) (0.012) (0.015) (0.022) Positive precipitation anomaly (Flood shock) 0.014 -0.022* 0.034** -0.005 (0.009) (0.013) (0.015) (0.020) Negative precipitation anomaly (Drought shock) -0.008 0.048*** -0.066*** -0.041* (0.010) (0.015) (0.016) (0.023) N 146,171 125,607 69,831 37,561 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: The time period covers 90-day bin before each household’s interview. Marginal effects estimated using Stata’s margins, dydx(*) command. 49 Table A.12: Effects of weather shocks on poverty, by economic sector of household head HH head primary sector HH head secondary sector HH head tertiary sector (1) (2) (3) (4) (5) (6) Urban Rural Urban Rural Urban Rural Positive temperature anomaly -0.150∗ 0.141∗∗∗ 0.132∗∗ 0.083 0.056 0.121∗ (Heat shock) (0.089) (0.032) (0.063) (0.084) (0.037) (0.068) Negative temperature anomaly 0.158∗ -0.117∗∗∗ -0.175∗∗∗ -0.077 -0.087∗∗ -0.209∗∗∗ (Cold shock) (0.095) (0.036) (0.068) (0.091) (0.041) (0.075) Positive precipitation anomaly 0.138 0.051 0.346∗∗∗ -0.325∗∗∗ -0.045 -0.057 (Flood shock) (0.098) (0.038) (0.062) (0.088) (0.039) (0.068) Negative precipitation anomaly -0.106 0.067 -0.337∗∗∗ -0.033 -0.043 0.123 (Drought shock) (0.107) (0.041) (0.069) (0.097) (0.042) (0.083) Average temperature 0.133 -0.162∗∗∗ -0.224∗∗∗ -0.102 -0.053 -0.209∗∗∗ (0.095) (0.035) (0.068) (0.091) (0.041) (0.076) -0.231∗∗∗ 0.084∗ 50 Average precipitation -0.062 -0.020 0.050 0.028 (0.069) (0.027) (0.047) (0.064) (0.030) (0.051) Historical average temperature -0.155 0.091∗∗ 0.069 0.109 0.068 0.161∗∗ (0.101) (0.037) (0.071) (0.094) (0.043) (0.078) Historical average precipitation 0.165∗∗ 0.171∗∗∗ 0.290∗∗∗ 0.023 -0.062∗ 0.008 (0.081) (0.032) (0.053) (0.074) (0.034) (0.061) Household controls Yes Yes Yes Yes Yes Yes Year-department dummies Yes Yes Yes Yes Yes Yes Month dummies Yes Yes Yes Yes Yes Yes Pseudo R-squared 0.33 0.21 0.32 0.25 0.33 0.29 N 10,984 89,791 39,722 16,726 11,6596 30,122 Standard errors in parentheses ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01 Note: The time period covers 90-day bin before each household’s interview. Household controls include: household size, dependency ratio, age of household head, age of household head squared, years of schooling of household head, and dummy variables if the household head is female, if the head and spouse only speak Guaraní, and whether the household has a car, heating or microwave. The model in column (2) excludes observations from due to insufficient sample size. Table A.13: Marginal effects of weather shocks on poverty, by economic sector of household head HH head primary sector HH head secondary sector HH head tertiary sector (1) (2) (3) (4) (5) (6) Urban Rural Urban Rural Urban Rural Positive temperature anomaly (Heat shock) -0.060* 0.053*** 0.038** 0.025 0.012 0.029* (0.035) (0.012) (0.018) (0.025) (0.008) (0.016) Negative temperature anomaly (Cold shock) 0.063* -0.044*** -0.051*** -0.023 -0.019** -0.049*** (0.038) (0.014) (0.020) (0.027) (0.009) (0.018) 51 Positive precipitation anomaly (Flood shock) 0.055 0.019 0.100*** -0.098*** -0.010 -0.013 (0.039) (0.015) (0.018) (0.027) (0.008) (0.016) Negative precipitation anomaly (Drought shock) -0.042 0.025 -0.097*** -0.010 -0.009 0.029 (0.043) (0.016) (0.020) (0.029) (0.009) (0.020) N 10,984 89,791 39,722 16,726 116,596 30,122 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: The time period covers 90-day bin before each household’s interview. Marginal effects estimated using Stata’s margins, dydx(*) command. The model in column (2) excludes observations from Itapuá due to insufficient sample size. Table A.14: Effects of weather shocks on total income, by economic sector of household head HH head primary sector HH head secondary sector HH head tertiary sector (1) (2) (3) (4) (5) (6) Urban Rural Urban Rural Urban Rural Positive temperature anomaly -0.061 -0.100∗∗∗ -0.077∗∗∗ -0.040 -0.046∗∗∗ -0.083∗∗∗ (Heat shock) (0.047) (0.016) (0.019) (0.031) (0.012) (0.024) Negative temperature anomaly -0.046 0.094∗∗∗ 0.041∗ -0.012 0.054∗∗∗ 0.099∗∗∗ (Cold shock) (0.050) (0.019) (0.022) (0.033) (0.014) (0.028) Positive precipitation anomaly -0.078 0.052∗∗∗ -0.034∗ 0.017 0.011 0.083∗∗∗ (Flood shock) (0.051) (0.019) (0.019) (0.032) (0.012) (0.023) Negative precipitation anomaly 0.058 -0.125∗∗∗ 0.059∗∗∗ 0.000 0.034∗∗ -0.067∗∗ (Drought shock) (0.051) (0.020) (0.023) (0.036) (0.013) (0.027) Average temperature 0.088∗ 0.108∗∗∗ 0.104∗∗∗ 0.057∗ 0.059∗∗∗ 0.104∗∗∗ (0.052) (0.018) (0.020) (0.034) (0.013) (0.027) 52 Average precipitation 0.054 -0.043∗∗∗ 0.036∗∗ 0.029 -0.008 -0.053∗∗∗ (0.037) (0.013) (0.015) (0.023) (0.009) (0.017) Historical average temperature -0.065 -0.068∗∗∗ -0.046∗∗ -0.045 -0.041∗∗∗ -0.081∗∗∗ (0.055) (0.019) (0.022) (0.034) (0.014) (0.028) Historical average precipitation -0.089∗∗ -0.030∗ -0.071∗∗∗ -0.095∗∗∗ -0.002 -0.003 (0.043) (0.016) (0.017) (0.028) (0.011) (0.021) Household controls Yes Yes Yes Yes Yes Yes Year-department dummies Yes Yes Yes Yes Yes Yes Month dummies Yes Yes Yes Yes Yes Yes R-squared 0.56 0.39 0.52 0.45 0.55 0.47 N 11,000 89,795 39,722 16,726 116,596 30,122 Standard errors in parentheses ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01 Note: The time period covers 90-day bin before each household’s interview. Household controls include: household size, dependency ratio, age of household head, age of household head squared, years of schooling of household head, and dummy variables if the household head is female, if the head and spouse only speak Guaraní, and whether the household has a car, heating or microwave. The data for this model excludes the year 2018 which does not provide sufficient information about household head’s economic sector. Table A.15: Marginal effects of weather shocks on total income, by economic sector of household head HH head primary sector HH head secondary sector HH head tertiary sector (1) (2) (3) (4) (5) (6) Urban Rural Urban Rural Urban Rural Positive temperature anomaly (Heat shock) -0.061 -0.100*** -0.076*** -0.041 -0.046*** -0.083*** (0.047) (0.016) (0.019) (0.031) (0.012) (0.024) Negative temperature anomaly (Cold shock) -0.046 0.094*** 0.041* -0.011 0.055*** 0.098*** (0.050) (0.019) (0.022) (0.033) (0.014) (0.028) 53 Positive precipitation anomaly (Flood shock) -0.078 0.053*** -0.034* 0.018 0.011 0.082*** (0.051) (0.019) (0.019) (0.032) (0.012) (0.023) Negative precipitation anomaly (Drought shock) 0.058 -0.125*** 0.058*** 0.000 0.034** -0.068** (0.051) (0.020) (0.023) (0.036) (0.013) (0.027) N 11,000 89,795 39,722 16,726 116,596 30,122 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: The time period covers 90-day bin before each household’s interview. Marginal effects estimated using Stata’s margins, dydx(*) command. The model in column (2) excludes observations from Itapuá due to insufficient sample size. Table A.16: Effects of heat days on poverty Days >25°C Days >30°C Days >35°C (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) U (No ag.) U (Ag.) R (No ag.) R (Ag.) U (No ag.) U (Ag.) R (No ag.) R (Ag.) U (No ag.) U (Ag.) R (No ag.) R (Ag.) Nr. of heat days 0.000 0.000 0.003 0.007∗∗∗ -0.004∗∗∗ 0.005∗∗ 0.000 0.007∗∗∗ -0.000 0.003 0.015∗∗∗ 0.005∗∗∗ (0.001) (0.002) (0.006) (0.002) (0.001) (0.002) (0.004) (0.001) (0.002) (0.003) (0.005) (0.002) Avg. temp. -0.015 -0.026 -0.013 -0.053∗∗∗ 0.009 -0.053∗∗ -0.001 -0.058∗∗∗ -0.014 -0.033∗ -0.043 -0.033∗∗∗ (0.014) (0.019) (0.042) (0.014) (0.015) (0.021) (0.039) (0.013) (0.014) (0.019) (0.031) (0.011) Avg. prec. 0.015∗ 0.004 0.040∗∗ -0.020∗∗∗ 0.012 0.008 0.040∗∗ -0.016∗∗ 0.015∗ 0.007 0.054∗∗∗ -0.017∗∗∗ (0.009) (0.011) (0.017) (0.006) (0.009) (0.011) (0.018) (0.007) (0.008) (0.011) (0.018) (0.007) Hist. avg. temp. 0.006 -0.049∗∗ -0.119∗∗∗ -0.024∗∗ 0.008 -0.051∗∗ -0.120∗∗∗ -0.032∗∗∗ 0.006 -0.049∗∗ -0.126∗∗∗ -0.028∗∗ (0.016) (0.021) (0.035) (0.012) (0.016) (0.021) (0.035) (0.012) (0.016) (0.021) (0.035) (0.012) 54 Hist. avg. prec. 0.033∗ -0.007 -0.030 0.149∗∗∗ 0.031∗ -0.003 -0.027 0.163∗∗∗ 0.032∗ -0.003 0.008 0.169∗∗∗ (0.017) (0.021) (0.032) (0.013) (0.017) (0.021) (0.031) (0.013) (0.018) (0.021) (0.033) (0.013) HH controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Year-dep. dum. Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Month dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Pseudo R-sq. 0.34 0.28 0.36 0.23 0.34 0.28 0.36 0.23 0.34 0.28 0.36 0.23 N 157,405 58,597 25,467 137,701 157,405 58,597 25,467 137,701 157,405 58,597 25,467 137,701 Standard errors in parentheses ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01 Note: The time period covers 90-day bin before each household’s interview. Household controls include: household size, dependency ratio, age of household head, age of household head squared, years of schooling of household head, and dummy variables if the household head is female, if the head and spouse only speak Guaraní, the economic sector of the household head (primary, secondary, tertiary, unemployed), whether the household is engaged in agriculture, and whether the household has a car, heating or microwave. Table A.17: Marginal effects of heat days on poverty (1) (2) (3) (4) Urban (no agriculture) Urban (agriculture) Rural (no agriculture) Rural (agriculture) Number of days with temperatures >25°C 0.000 0.000 0.001 0.003*** (0.000) (0.001) (0.001) (0.001) N 157,405 58,597 25,467 137,701 Number of days with temperatures >30°C -0.001*** 0.002** 0.000 0.003*** 55 (0.000) (0.001) (0.001) (0.001) N 157,405 58,597 25,467 137,701 Number of days with temperatures >35°C -0.000 0.001 0.004*** 0.002*** (0.001) (0.001) (0.001) (0.001) N 157,405 58,597 25,467 137,701 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: The time period covers 90-day bin before each household’s interview. Marginal effects estimated using Stata’s margins, dydx(*) command. B Appendix 56 Figure B.1: Average Temperature during October-December, 2004-2018 (a) 2004 (b) 2005 (c) 2006 (d) 2007 (e) 2008 (f) 2009 (g) 2010 (h) 2011 (i) 2012 57 (k) 2013 (l) 2014 (m) 2015 (n) 2016 (o) 2017 (p) 2018 Note: Districts in white are not included in EPH. 58 Figure B.2: Total Precipitation during October-December, 2004-2018 (a) 2004 (b) 2005 (c) 2006 (d) 2007 (e) 2008 (f) 2009 (g) 2010 (h) 2011 (i) 2012 59 (k) 2013 (l) 2014 (m) 2015 (n) 2016 (o) 2017 (p) 2018 Note: Districts in white are not included in EPH. 60 Figure B.3: Effects of temperature and precipitation shocks on income, by economic sector of household head HH head primary sector HH head secondary sector HH head tertiary sector .15 .1 .05 Income change 0 -.05 -.1 -.15 ck k k k ck k k k ck k k k oc oc oc oc oc oc oc oc oc ho ho ho sh sh sh sh sh sh sh sh sh ts ts ts ld d ht ld d ht ld d ht ea ea ea oo oo oo Co ug Co ug Co ug H H H Fl Fl Fl ro ro ro D D D Urban Rural Note: Time periods refers to the 90-day time bin before the interview. All coefficients shown in the figure represent marginal effects. The corresponding regression tables with estimated coefficients and marginal effects can be found in Tables A.14 and A.15 in Appendix A. 61