Policy Research Working Paper 11215 Uneven Waters Examining Poverty and Urban and Rural Households’ Exposure to Flood Risk in Paraguay Paul Ervin Lyliana Gayoso Eliana Rubiano Matulevich Poverty Global Department September 2025 Policy Research Working Paper 11215 Abstract Floods are becoming more frequent and severe due to cli- facing depths of flooding nearly four times higher than non- mate change, population growth, and land cover changes. poor households, in smaller, more common flood events. In Paraguay, floods are the most common weather-related The approach provides valuable insights for targeting flood hazard and disproportionately impact poor and vulnera- risk reduction efforts and highlights the importance of ble populations. This study contributes to understanding considering socioeconomic vulnerability in disaster risk household-level exposure to flood risk in Paraguay by com- management. These findings underscore the multidimen- bining geolocated household survey data with novel flood sional nature of vulnerability to flood risk, particularly in hazard maps. The study estimates that more than 23 per- rapidly urbanizing areas, and the need for integrated urban cent of households are exposed to flood risk, with exposure planning and poverty reduction strategies to address flood varying by geography and household characteristics. Urban risk disparities effectively, particularly in rapidly urbanizing households living in poverty are among the most exposed, areas. This paper is a product of the Poverty Global Department. 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 paervin@gmail.com, lgayoso@worldbank.org, and erubiano@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 Uneven Waters: Examining Poverty and Urban and Rural Households' Exposure to Flood Risk in Paraguay ∗ Paul Ervin, 1 Lyliana Gayoso, 2 and Eliana Rubiano Matulevich 3 Authorized for distribution by Carlos Rodriguez-Castelán, Practice Manager, Poverty Global Department, World Bank Group Keywords: poverty, welfare, flood hazard maps, household survey data, Paraguay JEL: I3, Q50, Q54, R2 ∗ 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 Mariana Conte, Alejandro de la Fuente, and Klaas de Groot for their feedback and suggestions on earlier versions of this paper. 1 Economic and Human Development Center (CEDEH), Paraguay and The World Bank. Email: paervin@gmail.com 2 [Corresponding Author] Economic and Human Development Center (CEDEH), Paraguay and The World Bank. Email: lgayosoervin@gmail.com, lgayoso@worldbank.org. 3 The World Bank. Email: erubiano@worldbank.org 1. Introduction Understanding flood exposure among those living in poverty is important for development outcomes and poverty reduction. Globally, the poor are often more exposed to flooding and typically the most vulnerable to flood impacts, as they are unable to afford mitigation measures and any reduction in income or consumption risks pushing them below subsistence levels (Hallegate et al., 2020). While poverty can contribute to flood exposure and vulnerability, flooding can also increase poverty through direct loss of income and assets (Rodriguez-Oreggia et al., 2012), long-term indebtedness, and reduced labor productivity, to mention a few, which may reinforce cycles of poverty (Banerjee & Duflo, 2012). Furthermore exposure to risk may induce behavioral responses where people reduce investment in productive assets and choose lower-risk, lower-return activities (Elbers, Gunning, & Kinsey, 2007; Cole, et al., 2013; Winsemius, et al., 2018), all of which can have significant implications for economic growth and development. Recent research in Paraguay estimates that floods have contributed to an increase in poverty of nearly 2 percentage points on average, primarily in urban areas (Janz, Gassmann, & Gayoso de Ervin, 2024). Flooding occurs more frequently than other natural hazards in Paraguay (EM-DAT, 2024) and has had significant impacts on the country’s population, displacing a large share of the population and impacting health and well-being (Nagy et al., 2016). Not surprisingly, urban flooding due to inadequate infrastructure and planning has emerged as a pressing issue for the country (World Bank, 2023). According to the Paraguay National Emergency Secretariat (Secretaría de Emergencia Nacional), floods in 2019 alone displaced over 60,000 families. Recurrent and severe floods continue to put lives and livelihoods at risk, posing a threat particularly to vulnerable households in floodplains (Ervin et al., 2025). Climate change, coupled with continued population growth and urbanization, is likely to contribute to increased flood risk in Paraguay in the future. The impacts of climate change in Paraguay are already evident. Changes in El Niño and La Niña, together with anomalies of the Intertropical Convergence Zone (ITCZ), have led to an increasing number of intense rainfall and flooding events, along with heat waves (Cai, et al., 2020; Glantz & Ramirez, 2020). In the last century, the frequency of floods and droughts has increased in Paraguay, while temperatures have risen and the occurrence of heat waves has tripled (GRID-Geneva & European Commission , 2023). Despite the growing evidence of climate change impacts in Paraguay, there remains a critical gap in our understanding of how these increasing flood risks intersect with socioeconomic vulnerabilities at the household level. The existing evidence suggests that poverty is both a driver and a consequence of flood exposure (Ervin et al., 2025). This implies that the poor are not necessarily the most exposed to flood risks. Indeed, while some studies provide evidence that people living in poverty are often more exposed to flood hazards (Pelling, 1997; Akter and Mallick, 2013), 4 this is not always the case, as other studies report similar exposure rates between poor and non-poor households (del Ninno et al., 2001; Opondo, 2013). On one hand, poor people may be more exposed primarily for the following three reasons: First, because flood prone areas offer economic opportunities and amenities to the poor, who then choose to live in risky areas. For example, poor households may live along a riverbank to have access to the river for fishing and water for agricultural 4 See Hallegatte et al. (2020) for a more thorough literature review of the multiple factors that explain why poor people are disproportionally affected by natural hazards and disasters. 2 and other uses; or, the poor may choose to reside in other flood prone areas, due to proximity to their workplace (e.g., dockworkers) or schools, among other things (Hallegatte, 2012; Patankar, 2015). Second, land and housing markets, especially in areas with land scarcity, may push poor people to live in riskier, cheaper areas (Daniel V, Florax, & Rietveld, 2009). Third, flood risk in an area may be unknown. For example, in a newly developed area flood risk may be unknown and as the area is affected by flooding and risk becomes more known, richer households may move away, while poorer households may not have the resources to leave. However, on the other hand richer households may be more exposed to flooding than the poor, because bodies of water such as rivers, lakes, and coastal shorelines may provide environmental amenities, such as beach access, river walks, lake access, or enjoyable views, that raise property prices and reduce the affordability of living in such areas that are also at flood risk (Zhenshan & Towe, 2024; Bin, Crawford, Kruse, & Landry, 2008). The empirical evidence on flood exposure and poverty indicates that a key pattern emerges in urban areas, where land scarcity often pushes poor households toward higher risk areas. Studies by Narloch and Bangalore (2018) and Winsemius et al. (2018) found that poor urban households face higher flood exposure than the average urban population, while flood exposure among rural households shows less variation by income level. This rural-urban distinction may reflect differences in wealth, land availability and the potential benefits of proximity to water for agricultural production areas (Hallegatte, 2016). In general, the various findings on the relationship between flood exposure and poverty in the existing literature are likely partly driven by the contexts of the various study areas, as well as the differing methodologies employed (Ervin et al., 2025). Many of the existing studies limit their focus to smaller communities heavily impacted by historical flooding events of interest. It is not always known if the chosen areas were relatively poorer in general or if relatively more poor than non-poor households were exposed to flooding. Other studies survey households on their perceptions of flood exposure over various recall periods, as well as future climate change expectations. Furthermore, most studies are not representative at the national or subnational level, making it difficult to generalize findings to larger populations. This study contributes to the literature by providing a granular analysis of how flood risks intersect with socioeconomic vulnerabilities at the household level for a country where this topic remains unexplored. Despite the growing problem of flooding in Paraguay and its contribution to poverty, little is known about the relationship between poverty and flood exposure in the country, making it difficult to design disaster risk management policies that support poor and vulnerable communities. The existing literature on flooding in Paraguay has focused on impacts on road infrastructure (Elkadi, et al., 2019), as well as the risks that flooded roads pose to vehicles and people as floods on urban streets reach high velocities (Díaz-Cardoso et al., 2018). In this study, we analyze the relationship between flood exposure, income, and poverty nationally and across rural and urban areas and regions in Paraguay. Our approach uses a continuous flood exposure index, capturing nuances in the severity of flood risk beyond simple binary classifications, following Ervin et al. (2025). We perform a flood exposure analysis at the household level by combining geolocated households from a national and subnational representative survey with novel, country-wide flood hazard maps. Our methodology follows the extended flood hazard risk analysis framework, where we use the information from multiple return period flood hazard maps to derive exceedance probability curves and calculate a flood exposure index based on the average annualized depth of flooding at each household’s 3 location. We then explore both the relationship between flood exposure and the severity of exposure to household income, poverty status, and other household-level vulnerability factors available in the household survey. This includes social vulnerability factors, such as the education and sex of the household head and the presence of children or elderly in the household, as well as vulnerability characteristics of the dwelling structure. Our results reveal that 23.3% of Paraguayan households are exposed to flood risk. This exposure shows distinct patterns across urban and rural areas. In urban settings, poor households face significantly higher exposure rates (27.1%) compared to non-poor households (21.0%), with this disparity most pronounced in Asunción, the capital city of the country, where 62.6% of poor households face flood risk versus 37.2% of non-poor households. The results show that in urban areas, a 10% increase in household income is associated with a 2.8% reduction in flood exposure and a 3.2% reduction in exposure severity. This urban pattern contrasts with rural areas, where non-poor households show slightly higher exposure rates (26.4%) than poor households (22.9%), though this difference is not statistically significant. Beyond income, we find that urban households with lower education levels, larger household sizes, presence of children, and lower- quality building materials face significantly higher flood risk, while in rural areas, female-headed households and overcrowded dwellings show higher exposure rates. Our analysis advances the literature on poverty and hazard risk analysis in several ways. We employ geolocated household surveys, representative at the national level, allowing for more accurate exposure analysis than studies using approximated locations or administrative-level data. By incorporating data from ten return periods of fluvial and pluvial flood maps, we construct detailed household flood exposure profiles that capture both the likelihood and severity flooding. This goes beyond binary exposure classifications to create a flood exposure index based on expected annualized flood depths at each location. Our methodology enables identification of not only which households face flood risk, but also their relative severity of exposure, providing policy makers with crucial information for targeting interventions. The analysis is representative at both national and subnational levels, allowing us to examine how flood exposure patterns vary across regions and urban-rural divides while accounting for household characteristics and vulnerability factors. We differ from existing studies in that we use precise georeferenced household locations collected at the time of the household survey. Accurate georeferenced households are important to understanding flood exposure across households as flood exposure and depths can be very different across households just 100 meters away from each other. We assure anonymity of interviewed households by only presenting aggregate results at representative areas, although the analysis is at the household level. Additionally, our sample is representative at the national and subnational levels, e.g., rural/urban areas and across several departments of Paraguay, which allows us to explore how exposure to flood risk varies across household income levels and poverty status within representative subnational administrative units, as well as by other characteristics. Finally, in our analysis we derive exposure to flood risk from relatively high-resolution, multiple return period flood maps that estimate the probability of flooding exceeding a flood depth at the household’s location, rather than using a flood map derived from one historical event, which likely only impacted a part of the country. This allows us to compare exposure to flood risk across households nationally and regionally and estimate both exposure to flood risk and the severity of exposure. 4 Our findings highlight how existing geolocated household survey data can be combined with hazard data to explore hazard exposure and household information. We believe this is an important first step in understanding the scale and scope of flood risk on households and the relationship between flood exposure and poverty. However, existing household surveys in most countries will likely need to be adapted to collect additional information relevant for policy makers to tailor mitigation, emergency response, and recovery programs to better understand the needs of those most at risk. In this context, vulnerability factors for specific hazards, such as characteristics and height of the housing structure or raised yards, as well as coping strategies would be useful. This study provides a methodological framework that can be adapted globally, potentially contributing to more informed disaster risk management practices. The rest of this paper is organized as follows. In the next section, we present the data and methodology used in the analysis. This is followed by the results and a conclusion summarizing key findings and policy implications. 2. Analytical framework Floods and poverty are interconnected through various reinforcing mechanisms, particularly in developing countries where vulnerability is higher due to weaker infrastructure, inadequate financial safety nets, and limited institutional capacity. Different frameworks have been proposed in the literature to illustrate the different pathways through which floods exacerbate poverty and reduce households’ income. In this study, we follow the extended risk assessment framework proposed by Hallegatte et al. (2020), which is grounded on the common flood risk assessment framework (IPCC, 2012; Koks, Jongman, Husby, & Botzen, 2015; Ikram, Jamalzi, Hamidi, Ullah, & Shahab, 2024), where flood risk is the result of three interconnected factors: = × × . Under the extended framework (Figure 1), the hazard is defined as the probability of a flood event occurring, while exposure is defined as the people, households, and assets located in flood-prone areas. Vulnerability is defined as the susceptibility to be adversely affected by a flood event, such as loss of assets or incomes, damage to property, and other disruptions, and Socioeconomic Resilience is the ability of the affected population to cope with and recover from disaster losses (Figure 1). Notably, this framework is consistent with recent theoretical advancements in disaster economics, which increasingly highlight the disproportionate vulnerability of economically vulnerable populations to natural hazards (Hallegatte, Fay, & Barbier, 2018). 5 Figure 1: Extended risk assessment framework Source: Hallegatte et al. (2020) In addition to this, we ground our work on previous studies that address, conceptually or empirically, how floods affect both the vulnerability and resilience of households, influencing their ability to cope with and recover from disasters. It is important to acknowledge that under this framework flood exposure is not exclusively a physical phenomenon, but rather a socioeconomic process determined by household decision-making, resource constraints, and prevailing institutional arrangements (Ervin et al., 2025). In developing countries, the dynamics of poverty and floods are influenced by several factors. For instance, exposure to flooding is determined by limited land availability, lower land prices, and limited housing options (Samarasinghe & Sharp, 2010; Rentschler, Salhab, & Jafino, Flood exposure and poverty in 188 countries, 2022; Kemwita, Kombe, & Nguluma, 2023). Overall, studies consistently demonstrate that limited land, high prices in safe areas, and poverty drive a disproportionate number of low-income households to flood-prone areas. Adding to this complex picture, historical settlement patterns and distinctive topographic features can further determine the specific geographical areas where both hazards are intensified and vulnerabilities are clustered. Furthermore, we explore the relationship between exposure to flood risk and vulnerability indicators to better understand the characteristics of households that are exposed to flood risk. In Paraguay, many factors can determine the vulnerability of households, such as household income, language spoken, educational level, gender, sector of employment, and occupation of the household head, or housing characteristics such as dwelling materials or access to basic services (electricity, water, etc.), which often shaped households’ decisions on where to live and define the opportunities they may have. We also acknowledge that vulnerability factors may be correlated among them, e.g., educational level is a correlate of income, and thus, of poverty. Because the examination of vulnerability factors that are correlated among them may lead to endogeneity, the empirical strategy exploits different model specifications. 3. Data and Methods 3.1. Data 3.1.1. Flood Hazard Data 6 To estimate household exposure to flood risk, we consider two types of flooding common in Paraguay: 5 fluvial (river) flooding, which occurs when a river or other bodies of water overflow, and pluvial (surface water) flooding, which occurs when rainfall overwhelms the terrain's ability to absorb it, creating localized floods that are independent of overflowing rivers or lakes. The country-wide fluvial and pluvial flood hazard data are from the 2019 global Fathom flood hazard datasets. 6 The Fathom flood-hazard model uses a methodology that combines a two-dimensional hydrodynamic model based on land morphology provided by MERIT-DEM, an enhanced digital elevation model of the NASA Shuttle Radar Topography Mission (SRTM) DEM, that corrects for multiple height errors from existing spaceborne DEMs (Yamazaki, et al., 2017) and river channel hydrography by MERIT-HYDRO. The Fathom flood model is based on observational records of meteorological and hydrological events, and uses gage data, where gages exist. Fathom's flood frequency analysis uses data from gaged regions to predict the behavior of extreme events in regions where no data exists. 7 The Fathom flood hazard datasets provide gridded information on the flood extent and depth of flooding above ground in meters at a 3 arcsecond (equivalent to 90 meters at the equator) resolution. While higher resolution data would be preferable for household or structure-level flood analyses, the 3 arcsecond maps are the only known flood maps available country-wide for multiple return periods in Paraguay that consider all flood types common in the country. Flood data is available for flood hazard scenarios simulated for 10 return periods including 5, 10, 20, 50, 75, 100, 200, 250, 500, and 1,000 years. 8 For each return period, Fathom flood maps provide the expected flood depth in meters for fluvial defended, fluvial undefended, and pluvial flooding. Fluvial defended maps include flood defenses depending on the availability of information. In this analysis, because a household can be exposed to flooding from multiple sources, we estimate the depth of exposure at a household’s location by considering the maximum depth of flooding over all flood types, including pluvial, fluvial defended, and fluvial undefended for each return period. This might result in overestimation of exposure to flood risk in areas where hydraulic structures provide flood protection. However, we do not expect this to significantly impact our results, as Paraguay has limited flood protection systems in place. 3.1.2. Paraguay Household Survey Data The household survey data comes from the 2021 Paraguay Permanent Household Survey (Encuesta Permanente de Hogares Continua (EPHC)), the latest available with information on households’ GPS coordinates. The EPHC is an annual household survey implemented since 2002 by the National Statistical Office of Paraguay (INE) and serves as the reference for welfare measurement and poverty in Paraguay. 5 Paraguay is a landlocked country and, therefore, is not at risk of coastal flooding. 6 https://fathom.global 7 For more information on the Fathom flood model and methods, see https://www.fathom.global/academic-papers. 8 Return periods provide an estimated average time between the occurrence of flood events of a certain magnitude. The inverse of the return period is the average frequency of occurrence, which indicates the probability of a hazard occurring in any one year, e.g., a 10-year flood has a 10% (1/10) chance of being exceeded in any given year. Less frequently occurring flood events, such as the 100-year flood, generally have larger flood extents and are more severe. 7 The 2021 EPH surveyed 4,646 households and is representative at the national level, for urban and rural areas, the capital city of Asunción, and the following departments (or states): San Pedro, Caaguazú, Itapúa, Alto Paraná, Central, and a group called “Other departments” that combines the rest of the departments in the country excluding Boquerón and Alto Paraguay. The survey contains modules that collect information on demographic characteristics of all household members, education, health, labor markets, incomes, as well as households’ assets, and dwelling characteristics, among others. Individual households may floodproof their structure, for example by elevating their terrain or structure. Information on the elevation of the household’s structure and other floodproofing activities is not available in the household survey data. We expect such measures to be more commonly employed among the more well-off, so that our estimates may overestimate exposure to flood risk among the wealthy. We discuss this further in the results section. 3.2. Methodology Our methodological approach uses a flood hazard risk analysis framework similar to Ervin, Gayoso de Ervin, Rubiano-Matulevich and Canavire-Bacarreza (2025) 9 and involves intersecting household locations with flood risk information using household coordinates and the flood hazard maps for each return period. This allows us to obtain flood exposure profiles for each household, expressed as exceeded probability curves that show the relationship between flood depths and their probability of occurrence for each return period. A map of Paraguay showing household points and a flood hazard map is shown in Figure 2: Exceedance Probability Curves for Different Table A 1 in the Appendix. Our approach Households goes beyond simple binary classifications of flood exposure by constructing a continuous flood exposure index. This index allows us to capture nuances in the severity of flood risk, providing a more detailed picture of household vulnerability. Figure 2 shows the flood exposure profiles for three different households. For example, a 100-year flood map shows the flood depths corresponding to a 1% (1/100) probability of exceedance in a given year. 10 The shape of the exceedance probability curve is influenced by the source of flooding and the topology of the terrain, which vary by household locations. Using each household’s flood exposure profile, we estimate the expected level of flood exposure by computing the area under each household’s exceedance curves using numerical integration. We calculate each household’s average annualized flood depth (AAD), using the following formula 9 See also Bellelli (2022, May 22). F.S.Bellelli: A brief introduction to natural hazard risk analysis. Retrieved from https://fbellelli.com/posts/2022-05-22-a-brief-introduction-to-natural-hazard-risk-analysis/ for a further illustrative example of flood risk analysis similar to the method employed in this paper. 10 These exceedance probabilities assume a stationary distribution and are independent of each other. Therefore, the probability of a 100-year event occurring (or being exceeded) in a 30-year period is 1-0.74=0.26 or 26%. 8 1 | | 1 1 = ∑=1 � + � �,+1 − , � ……(1), 2 +1 where = {5, 10, 20, 50, 75, 100, 200, 250, 500, 1,000} is the set of flood return periods and represents the jth element in the set of years corresponding to the 10 return periods, so that 1/1 equals 1/5, capturing the probability of occurrence of the flood event, and so on. , is the depth of flooding above ground for the return period at household i’s location. While AAD may be understood as the probability-weighted depth of flooding at a household’s location in any given year, flood risk is cumulative with many years of no flooding followed by an often-unexpected large flood event. Therefore, we rescale the AAD metric using min-max normalization for each household to create a Flood Exposure Index (FEI) as, −min () = …… (2). max ()−min () The normalization is done, so that the households with the highest level of exposure to flood risk score a one on the FEI and households with no risk score zero. Like AAD, the FEI is a continuous variable that ranks households by their likelihood and severity of flood exposure. However, while it preserves the ranking of households by exposure to flood risk, its units are meaningless. To capture whether a household is estimated to be exposed to any flood risk, we create the following indicator variable: 1 > 0 = � …… (3), 0 = 0 where denotes whether a household is estimated to face any flood exposure. Therefore, a household is exposed to flood risk if there is any chance of flooding at any depth greater than 0 meters at the household’s location. While this estimate may be conservative on the high end (e.g., overestimate exposure), we also explore flood exposure across return periods by poverty status. We further perform statistical analyses of the relationship between household exposure to flood risk and household poverty status and income. We control for and explore other variables that could be related to flood exposure and vulnerability common in the existing literature (Lottering, Mafongoya, & Lottering, 2021; Bucherie, et al., 2022; Kawasaki, Kawamura, & Zin, 2020; Bangalore, Smith, & Veldkamp, 2019). It is hypothesized that characteristics of the household head play a role in the decision of where to live and, thus, flood exposure. The construction material of the house also has been shown to influence the severity of damage in the event of a flood, but can also be determined by exposure to flood risk (e.g., building a cheaper house expecting it to flood). Household composition and demographics and other information may also be related to exposure and vulnerability. To explore the correlates of household flood exposure we consider the following regression: = ∑ 0, + 1 + ∑ + , …… (4). 9 where the dependent variable, , is either the binary variable, , identifying whether household i is exposed to any flood risk or the continuous flood exposure index, , estimating the likelihood and severity of flood exposure. To analyze , we estimate Equation (4) using a logistic regression. To analyze , we estimate Equation (4) using ordinary least squares (OLS). is a vector of department fixed effects with d as the department indicator. is a household welfare variable, which is either poverty status or income. denotes a vector of household head, demographics, and other household vulnerability characteristics. is an error term for household idiosyncrasies not captured in other terms. Several caveats are worth noting. First, the household survey lacks information on flood protection infrastructure and mitigation measures (e.g., elevated terrain, raised structures), which are more prevalent among well-off households, and on quality of the infrastructure, which may influence the impact of floods. This data limitation may lead on one hand, to systematic overestimation of flood exposure among wealthier households. On the other hand, exposure to flood risks, both pluvial and fluvial, may also be underestimated. 11 In the context of Paraguay, the underestimation may be partly due to the continued occupation of poor households on flood-prone areas, like rivers and streams, and the reduced capacity for water flow caused by extensive landfills, for instance, those who are seen along the Paraguay River. For pluvial flooding, poor solid waste management also plays a significant role in the country. Clogged drainage systems in the country frequently cause or exacerbate floods, in particular in urban areas. However, these peculiarities of the country are not accounted for in current flood hazard mapping models. Second, our flood hazard estimates rely on modeled data that may not capture all topographical features or protection measures. While this introduces some uncertainty in exposure estimates, particularly at fine spatial scales, the methodology provides valuable insights for targeting disaster risk reduction interventions when combined with local expertise. Finally, while our analysis identifies correlations between flood exposure and poverty, the potential endogeneity between these variables – where flood risk areas may attract poor households while flooding events can induce poverty through direct asset losses and supply chain disruptions - is beyond the scope of this paper. These dynamics could be explored through catastrophe risk modeling frameworks to understand the broader economic and poverty impacts of flooding (Grossi, Kunreuther, & Patel, 2005; Ranger, et al., 2011). 3.3. Study Sample, Poverty, and Household Characteristics The 2021 Paraguay household survey (EPH) contains information on 4,646 households. Our study sample includes all 4,646 households with complete information relevant for our study, including valid GPS coordinates required to link household locations to flood hazard maps. Variable definitions are presented in Table A 2 in the Appendix and descriptive statistics for the study sample are presented in Table 1. For this study, the unit of analysis is the household. The average monthly household income per capita is PYG 1,771,626 (which is approximately 260 USD in 2021 prices) 12 and about 15% of households receive some form of social assistance. The 2021 national household poverty rate 13 was 11 Pelling (1997) found raised living spaces, i.e., houses raised to be higher than anticipated flood levels, to be common both among lower income and higher income households in Guyana, with 60% of higher income households living in a raised dwelling compared to 40% of lower income households living in raised dwellings. 12 Using the average exchange rate 6,774.16 $/PYG. for 2021 from https://wdi.worldbank.org/table/4.16. 13 The 2021 national poverty line was PYG 666,847 or approximately 98 USD. 10 20.48%, which was slightly below the national population poverty rate of 26.9%, because poor households tend to be larger than non-poor households. About 63% of households reside in urban areas and 37% are in rural areas. Table 1: Descriptive Statistics for Study Sample Variable Obs. Mean St. Dev. Min. Max. Income per capita (monthly, PYG 2021) 4,646 1,771,626 2,031,701 0.00 85,756,613 Poverty status 4,646 0.2048 0.4036 0.00 1.00 Receives social assistance 4,646 0.1450 0.3522 0.00 1.00 Head age 4,646 47.7860 15.5516 17.00 98.00 Head years schooling 4,645 8.8995 4.7272 0.00 18.00 Head female 4,646 0.3906 0.4879 0.00 1.00 Head language Guarani 4,646 0.3868 0.4871 0.00 1.00 Head born rural 4,646 0.5762 0.4942 0.00 1.00 Head sector agriculture 4,646 0.1769 0.3816 0.00 1.00 Head inactive 4,646 0.1914 0.3934 0.00 1.00 Household size 4,646 3.7483 1.9409 1.00 19.00 Child under age of 15 present 4,646 0.5779 0.4939 0.00 1.00 Adult over age of 65 present 4,646 0.2042 0.4032 0.00 1.00 Overcrowded 4,646 0.2295 0.4206 0.00 1.00 Dwelling floor made of tile, granite, 4,646 0.6157 0.4865 0.00 1.00 porcelain, or parquet Dwelling wall made of brick, cement, 4,646 0.8333 0.3727 0.00 1.00 stucco, or adobe Urban 4,646 0.6343 0.4817 0.00 1.00 Rural 4,646 0.3657 0.4817 0.00 1.00 Asunción 4,646 0.0698 0.2548 0.00 1.00 San Pedro 4,646 0.0584 0.2346 0.00 1.00 Caaguazú 4,646 0.0823 0.2748 0.00 1.00 Caazapá 4,646 0.0280 0.1651 0.00 1.00 Itapúa 4,646 0.0934 0.2911 0.00 1.00 Alto Parana 4,646 0.1197 0.3247 0.00 1.00 Central 4,646 0.2923 0.4549 0.00 1.00 Other departments 4,646 0.2561 0.4365 0.00 1.00 Note: Statistics are weighted by household weights and nationally representative of households. To explore the profile of households exposed to flood risk, we include additional characteristics of the household in our study. This includes information on the demographics of the household head: years of schooling, language spoken, and whether the household head works in the agriculture sector and may benefit from floods for agricultural production. Additionally, we include information on household demographics, such as the household size and the presence of young children and older adults. Finally, we 11 include information on the dwelling, such as whether more than 2 people inhabit a room of the house—a measure of overcrowding—and the construction material of the floor and wall, which may be more resilient to short-term flooding. 4. Results 4.1. Household Exposure to Flood Risk and Poverty We estimate that nationally, 23% (95% CI: 21.8%-24.7%) of households in Paraguay, or approximately 445,145 households of the total 1,935,412 households, are located in areas with exposure to flood risk. This indicates that nearly one in four households is estimated to face some level of flood risk. Figure 3 presents estimates of household exposure to flood risk by location and poverty status. Exposure to flood risk varies significantly across rural and urban areas and by department. Rural households are estimated to be slightly more exposed (25%) than urban households (22%). The capital city of Asunción shows the highest exposure rate, with 40% of households at risk, followed by the departments of San Pedro (31%), the grouping of other departments (28%), and Central (23%). Figure 3: Household Flood Exposure by Poverty Status in Paraguay by Location 70% 60% 50% Households 40% 30% 20% 10% 0% National Urban* Caazapa Itapua* Rural Alto Parana Caaguazu Central* Other depts. San Pedro Asuncion* Area Department Total households Non-poor Poor Note: * indicates difference between poor and non-poor households is statistically significant at the 5% significance level. Table 2 shows estimates of household exposure to flood risk by location and poverty status and 95% confidence intervals. Our analysis reveals significant disparities in flood exposure between poor and non- poor households, particularly in urban areas. In urban settings, 27.1% of poor households are exposed to flood risk, compared to 21% of non-poor households. This difference is statistically significant and highlights the vulnerability of urban poor populations to flood hazards. The gap between poor and non-poor households is most pronounced in the capital city of Asunción, where 62.6% of poor households and 37.2% of non-poor households are estimated to be exposed to flood risk. 12 This is followed by the department of Central, where 32.2% of poor households and 21% of non-poor households are estimated to be exposed to flood risk. These results are both statistically significant. This stark difference underscores the spatial inequality in flood risk within urban areas and suggests that poverty and flood exposure are closely interlinked in urban settings. In rural areas, we find that non-poor households are slightly more exposed to flood exposure risk (26.4%), compared to poor households (22.9%). However, these results are not statistically significant, indicating that the relationship between poverty and flood exposure is less clear in rural contexts. Only in the department of Itapúa is the gap between non-poor and poor households statistically significant, with 15.8% of non-poor households estimated to be exposed to flood risk versus 7.6% of poor households. Table 2: Household Flood Exposure and Poverty by Location with Confidence Intervals Non-poor Households Poor households Households Lower Upper Poor Lower Upper Non-poor Lower Upper Location Households 95% CI 95% CI households 95% CI 95% CI Households 95% CI 95% CI National 23.3% 21.8% 24.7% 25.1% 21.8% 28.3% 22.8% 21.2% 24.4% Urban* 22.0% 20.1% 24.0% 27.1% 21.9% 32.3% 21.0% 18.9% 23.1% Rural 25.4% 23.4% 27.5% 22.9% 19.1% 26.6% 26.4% 23.9% 28.8% Asunción* 39.7% 33.1% 46.3% 62.6% 40.4% 84.8% 37.2% 30.4% 44.1% San Pedro 30.8% 26.3% 35.2% 28.4% 20.3% 36.5% 31.9% 26.5% 37.3% Caaguazú 20.7% 16.2% 25.2% 23.1% 15.1% 31.0% 19.5% 14.0% 24.9% Caazapá 22.0% 17.3% 26.8% 26.5% 17.4% 35.7% 20.0% 14.5% 25.4% Itapúa* 13.9% 10.4% 17.3% 7.6% 1.6% 13.6% 15.8% 11.8% 19.9% Alto Paraná 11.7% 9.0% 14.4% 15.5% 7.9% 23.1% 10.7% 7.9% 13.5% Central* 22.6% 19.2% 25.9% 32.2% 22.0% 42.4% 21.0% 17.6% 24.5% Other depts. 27.7% 25.0% 30.4% 26.1% 20.3% 31.9% 28.2% 25.2% 31.2% Note: * indicates difference between poor and non-poor households is statistically significant at the 5% confidence level. Rural-urban differences in household exposure to flood risk between poor and non-poor households stand out particularly across exposure and severity by flood return periods. Figure 4 presents estimates of household exposure to flood risk and the severity of flood exposure between rural and urban areas and by household poverty status. Panels in Figure 4 show the likelihood of exposure, by return period, and the average depth of flooding households are likely to be exposed to. Panels (a) and (c) show that poor urban households face significant and high flood risk relative to non-poor households across return periods. Poor households in urban areas are exposed to depths of flooding nearly four times higher, on average, than non-poor households in smaller, more common flooding and in larger, less common flood events, depths of flooding at poor, urban households’ locations are twice as high as non-poor households. Rural household exposure to flood risk is more similar across household poverty status, with non-poor households estimated to be exposed to slightly more flood risk than poor households (panels (b) and (d)). However, these results are not statistically significant between non-poor and poor households across return periods. Overall, these findings suggest that in urban areas the poor are both more likely to be exposed to flood risk, as well as to face much higher levels of flood depth. 13 Figure 4: Household Exposure to Flood Risk and Estimated Flood Depth by Return Period and Poverty Status (a) Urban household flood exposure by return (b) Rural household flood exposure by return period and poverty status period and poverty status 30% 30% 25% 25% Households 20% Households 20% 15% 15% 10% 10% 5% 5% 0% 0% 1,000 5 10 20 50 75 100 200 250 500 1,000 5 10 20 50 75 100 200 250 500 Flood Return Period (years) Flood Return Period (years) Poor Non-poor Poor Non-poor Note: Differences are statistically significant between poor and Note: Differences are not statistically significant between poor non-poor households. and non-poor households. (c) Flood depth at urban household's location by (d) Flood depth at rural household's location by return period and poverty status return period and poverty status 1 0.35 0.3 Flood Depth (m) 0.8 Flood Depth (m) 0.25 0.6 0.2 0.15 0.4 0.1 0.2 0.05 0 0 Flood Return Period (years) Flood Return Period (years) Poor Non-poor Poor Non-poor Note: Differences are statistically significant between poor and Note: Differences are not statistically significant between poor non-poor households. Results are conditional on flood and non-poor households. Results are conditional on flood exposure. exposure. Our study is similar to recent works by Winsemius et al. (2018), Kawasaki, Kawamura, and Zin (2020), and Rentschler, Salhab, and Jafino (2022), but differs in important ways. Winsemius et al. (2018) explored exposure to river flooding across 52 countries using spatial wealth data sets from USAID’s DHS surveys. These data sets provide households’ approximate location georeferenced to the survey cluster centroid with random error to maintain anonymity of the interviewed household. Furthermore, poverty is estimated using the bottom quintile of country-specific wealth indices. Kawasaki, Kawamura, and Zin (2020) use a household survey representative of Bago city, Myanmar, to analyze flood exposure to the large 2011 flood at the household level and explore the relationship between flood depth in the 2011 flood, household income, and other factors, such as education, house type, and household demographics. Rentschler, Salhab, and Jafino (2022) use similar relatively high-resolution, country-wide flood maps as we use in this 14 study, but they use population density maps and poverty levels at subnational administrative levels. In a given area within the subnational administrative area Rentschler, Salhab, and Jafino (2022) assume that flood exposure is uniform across income groups. The findings presented in this section are aligned with the results obtained by recent studies conducted by Pelling (1997), Akter and Mallick (2013), and Ervin et al. (2025). These studies report the disproportionate flood exposure of poor households in Guyana, Bangladesh, and Grenada. However, our study, similar to that of Ervin et al. (2025), adds an additional dimension to the complex relationship that exists between exposure to flood risk and poverty, by extending the analysis to account for flood depths. 4.2. Flood Exposure and Other Household Characteristics The preceding analysis highlights how household-level exposure to flood risk varies by geographic area and poverty status. In this sub-section, we investigate the relationship of household exposure to flood risk and household income levels and other household characteristics. Figure 5 presents estimates of household flood exposure and the severity of flood exposure, measured by the FEI, across household income quintiles. In urban areas, household exposure to flood risk decreases as income increases (statistically significant at the 10% level). In rural areas, household exposure to flood risk shows more variation across income quintiles, slightly increasing over the lower income quintiles and highest among the top income quintile. However, these results are not statistically significant. Figure 5: Household Exposure to Flood Risk and Severity of Flood Exposure by Income Quintile (a) Household Flood Exposure across Income (b) Household Severity to Flood Exposure across Quintiles by Area Income Quintiles by Area 31% 0.55 29% 0.5 Flood Exposure Index 27% 25% Households 0.45 23% Rural Rural 0.4 21% Urban Urban 19% 0.35 17% 0.3 15% 1 2 3 4 5 1 2 3 4 5 Income Quintile Income Quintile Note: Urban differences with the top quintile are statistically Note: Results conditional on flood exposure. Urban difference significant at the 10% level. Rural differences are not between top and bottom quintiles is statistically significant statistically significant. below the 5% level. Rural difference between the top and second quintile is statistically significant below the 5% level. Income elasticities from regression analyses (See Appendix, Table A 5 and Table A 6) suggest that in urban areas a 10% increase in household income per capita is associated with a 2.8% reduction in household 15 exposure to flood risk and a 3.2% reduction in the level of exposure (FEI), and these results are statistically significant at conventional levels. In rural areas, while household flood exposure is positively related with household income, the results are not statistically significant. While no information is available in the survey on household flood protection, such as raising the elevation of the terrain or structure to reduce risk of flooding, we expect this to be more prevalent among well-off households. If this is the case, flood exposure and severity would be overestimated for the richest quintiles and the relationship between income and household flood exposure would be even stronger. Household profiles for urban (Appendix, Table A 3) and rural (Appendix, Table A 4) households by flood exposure status are presented in the Appendix. In urban areas, households exposed to flood risk are more likely to be in poverty and have lower income levels and these results are statistically significant. Beyond income and poverty status, we find that several other household characteristics are associated with higher flood exposure, particularly in urban areas. Urban households with higher flood risk are statistically more likely to have household heads with lower education levels, have more household members, have more children present, to be overcrowded, and have a dwelling constructed out of lower quality materials, such as dirt, or low-quality wood, brick, or cement. In rural areas, households at flood exposure risk are statistically more likely to be headed by a woman and be overcrowded. These results highlight the multidimensional nature of vulnerability to flood risk, suggesting that socioeconomic factors beyond income play a crucial role in determining household exposure to flood hazards. For instance, lower quality housing is more likely to be substantially damaged during a flood, and larger households with more children means more people will be impacted and need an adult to care for children while housing is repaired or the family is relocated. Regression analyses (Appendix, Table A 5 and Table A 6) further provide support on the relationship between household exposure to flood risk and education levels, household size, and overcrowding. However, we did not find evidence that household exposure to flood risk is related to households already receiving some form of social assistance, a potential avenue for targeting at-risk households; the area the household head was born, a measure for internal migration and urbanization; nor the head working in the agricultural sector, which we hypothesized may encourage households to locate near water sources for agricultural inputs. Household income and human capital appear to be among the most important factors related to household exposure to flood risk, but more information is needed to understand the mechanism driving low-income households to risky areas in Paraguay. 5. Conclusion and Policy Implications Our analysis demonstrates that flood exposure affects a substantial portion of Paraguay’s population, with 23.3% of households exposed to flood risk. This exposure varies across geographic and socioeconomic dimensions, revealing important patterns for policy consideration, while acknowledging that our findings identify correlations rather than establish causal relationships. The relationship between poverty and flood exposure differs between urban and rural areas. In urban settings, poor households face higher exposure rates (27.1%) compared to non-poor households (21.0%), with this disparity most pronounced in Asunción where 62.6% of poor households face flood risk versus 16 37.2% of non-poor households. Rural areas show a different pattern, with non-poor households having slightly higher exposure rates (26.4%) compared to poor households (22.9%), though this difference is not statistically significant. Beyond exposure rates, poor urban households face higher flood depths across both frequent and rare flooding events. Our analysis reveals that a 10% increase in urban household income corresponds to a 2.8% reduction in flood exposure probability and a 3.2% reduction in exposure severity. Several household characteristics beyond income correlate with flood exposure. In urban areas, households with lower education levels, larger family sizes, presence of children, and lower-quality building materials face significantly higher flood risk. In rural areas, female-headed households and overcrowded dwellings show higher exposure rates. Urban flood exposure among the poor is driven primarily by increased fluvial (river) flood risk, while rural patterns show different dynamics influenced by pluvial (surface water) flooding patterns. Based on our empirical findings and acknowledging the limitations in establishing causal mechanisms, we propose evidence-informed policy directions that address the disparities documented in our analysis. Given that 62.6% of poor households in Asunción face flood risk, urban-focused strategies are essential. These could include developing risk-informed urban planning that incorporates flood hazard information into zoning and development decisions, voluntary relocation programs for households in high-risk areas with adequate support and livelihood opportunities, and targeted infrastructure investments in high-risk, low- income urban areas prioritizing drainage systems and flood protection measures, consistent with global evidence on effective flood risk reduction. Our findings on housing quality correlations suggest exploring home improvement grants or low-interest loans for flood-proofing measures in at-risk areas, developing and enforcing building codes mandating flood-resistant design in flood-prone zones, and creating incentives for affordable housing development in lower-risk areas to reduce concentration of vulnerable populations in flood-prone regions. In addition, the different patterns observed between urban and rural areas indicate the need for differentiated approaches. Urban interventions need to address land scarcity and housing market dynamics that may push poor households into high-risk areas, while rural interventions could consider gender-specific approaches given higher exposure among female-headed households. Both contexts require addressing overcrowding as a common vulnerability factor. A starting point for enhanced data collection and risk communication strategies involves expanding household surveys to include flood protection measures, flood experiences, and adaptation strategies. Risk communication strategies need to consider education levels and language barriers, particularly among Guaraní speakers, while establishing systematic data collection that combines socioeconomic data with flood risk information for better targeting interventions. Infrastructure improvements need to include enhanced early warning systems ensuring coverage of all communities, investments in both traditional flood protection infrastructure and green solutions and addressing underlying causes such as solid waste management that can exacerbate pluvial flooding. This study provides valuable insights while acknowledging limitations. Our analysis may overestimate flood exposure among wealthier households due to lack of information on household-level flood protection measures. Additionally, current flood hazard models may underestimate exposure due to factors like 17 informal settlements along waterways and infrastructure degradation not captured in the modeling. While we identify strong correlations between flood exposure and socioeconomic factors, establishing causal relationships requires further research. The potential endogeneity between flood risk and poverty—where flood-prone areas may attract poor households while flooding events can induce poverty—merits deeper investigation through approaches such as catastrophe risk modeling frameworks to understand broader economic impacts. Future research could explore city-level analysis when larger datasets become available, extend the methodology to other natural hazards beyond flooding, and investigate the mechanisms driving the relationship between poverty and flood exposure. The methodological framework we provide can be adapted to other countries, but effective policy implementation requires local adaptation, community engagement, and recognition that flood exposure patterns reflect complex interactions between physical hazards, socioeconomic vulnerabilities, and institutional capacities. Our findings highlight the multidimensional nature of vulnerability to flood risk, particularly in rapidly urbanizing areas. The disparities we document, especially the nearly four times higher flood depths faced by poor urban households during common flooding events, underscore the urgent need for integrated approaches that combine urban planning, poverty reduction, and disaster risk management strategies. The evidence base we provide could inform locally-informed policy design that considers the specific contexts and needs of affected communities, while contributing to the broader understanding of how flood risks intersect with socioeconomic vulnerabilities in developing countries facing increasing climate-related challenges. 18 References Akter, S., & Mallick, B. (2013). The poverty–vulnerability–resilience nexus: Evidence from Bangladesh. Ecological Economics, 96, 114-124. Banerjee, A., & Duflo, E. (2012). Poor Economics: A Radical Rethinking of the Way to Fight Global Poverty. New york: Public Affairs. Bangalore, M., Smith, A., & Veldkamp, T. (2019). Exposure to floods, climate change, and poverty in Vietnam. Economics of Disasters and Climate Change, 3, 79-99. Bin, O., Crawford, T., Kruse, J., & Landry, C. (2008). Viewscapes and Flood Hazard: Coastal Housing Market Response to Amenities and Risk. Land Economics, 84(3), 434-448. doi:https://doi.org/10.3368/le.84.3.434 Bubeck, P., Otto, A., & Weichselgartner, J. (2017). Societal impacts of flood hazards. Oxford Research Encyclopedia of Natural Hazard Science. doi:https://doi.org/10.1093/acrefore/9780199389407.013.281 Bucherie, A., Hultquist, C., Adamo, S., Neely, C., Ayala, F., Bazo, J., & Kruczkiewicz, A. (2022). A comparison of social vulnerability indices specific to flooding in Ecuador: Principal component analysis (PCA) and expert knowledge. International journal of disaster risk reduction, 73. Cai, W., McPhaden, M. J., Grimm, A. M., Rodrigues, R. R., Taschetto, A. S., Garreaud, R. D., . . . Vera, C. (2020). Climate impacts of the El Niño–Southern Oscillation on South America. Nat Rev Earth Environ, 1, 215–231. doi:https://doi.org/10.1038/s43017-020-0040-3 Cole, S., Giné, X., Tobacman, J., Topalova, P., Townsend, R., & Vickery, J. (2013). Barriers to Household Risk Management: Evidence from India. American Economic Journal: Applied Economics, 5(1), 104–35. Daniel V, E., Florax, R., & Rietveld, P. (2009). Flooding risk and housing values: an economic assessment of environmental hazard. Ecol Econ, 69(2), 355–365. del Ninno, C., Dorosh, P., Smith, L., & Roy, D. (2001). The 1998 floods in Bangladesh: disaster impacts, household coping strategies, and response (Vol. 122). International Food Policy Research Institute. Diaz-Cardoso, A., López-Fernández, A., & Camaño-Schettini, E. (2018). Danger of torrents for users of public roads and streets in cities of Paraguay. Revista de la Sociedad Científica del Paraguay, 23(2). doi:https://doi.org/10.32480/rscp.2018-23-2.241-250 Elbers, C., Gunning, J., & Kinsey, B. (2007). Growth and Risk: Methodology and Micro Evidence, The World Bank Economic Review. 21(1), 1-20. doi:https://doi.org/10.1093/wber/lhl008 Elkadi, A. S., 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. EM-DAT. (2024). The International Disaster Database. Centre for Research on the Epidemiology of Disasters, part of the University of Louvain (UCLouvain). Retrieved from https://www.emdat.be/ 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, 44. Retrieved from http://www.fao.org/3/ca3391en/CA3391EN.pdf Ervin, P., Gayoso de Ervin, L., Rubiano-Matulevich, E., & Canavire-Bacarreza, G. (2025). When the Waters Rise: Examining Household-Level Exposure to Inland Flood Risk and Poverty in Grenada. Economics of Disasters and Climate Change. 19 Ervin, P., Gayoso de Ervin, L., Rubiano-Matulevich, E., & Canavire-Bacarreza, G. (2025). When the Waters Rise: Examining Household-Level Exposure to Inland Flood Risk and Poverty in Grenada. Economics of Disasters and Climate Change: https://doi.org/10.1007/s41885-025-00172-w. Glantz, M. H., & Ramirez, I. J. (2020). Reviewing the Oceanic Niño Index (ONI) to enhance societal readiness for El Niño’s impacts. International Journal of Disaster Risk Science, 11, 394-403. GRID-Geneva & European Commission . (2023, November ). Paraguay Country Profile . Retrieved from Interactive Country Fiches : https://dicf.unepgrid.ch/paraguay Grossi, P., Kunreuther, H., & Patel, C. C. (2005). Catastrophe modeling: a new approach to managing risk (Vol. 25). Springer Science & Business Media. Hallegatte, S. (2012). An exploration of the link between development, economic growth, and natural risk. World Bank Policy Research Working Paper 6216. Hallegatte, S. (2016). Shock waves: managing the impacts of climate change on poverty. World Bank Publications. Hallegatte, S., Bangalore, M., Bonzanigo, L., Fay, M., Kane, T., Narloch, U., . . . Vogt-Schilb, A. (2016). From Climate Change to Poverty and Back: A Framework. In S. Hallegatte, M. Bangalore, L. Bonzanigo, M. Fay, T. Kane, U. Narloch, . . . A. Vogt-Schilb, Shock Waves. Managing the Impacts of Climate Change on Poverty. Washignton, D.C.: The World Bank. Hallegatte, S., Fay, M., & Barbier, E. B. (2018). Poverty and climate change: introduction. Environment and Development Economics, 23(3), 217-233. Hallegatte, S., Vogt-Schilb, A., Bangalore, M., & Rozenberg, J. (2016). Unbreakable: building the resilience of the poor in the face of natural disasters. World Bank Publications. 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. Ikram, Q. D., Jamalzi, A. R., Hamidi, A. R., Ullah, I., & Shahab, M. (2024). Flood risk assessment of the population in Afghanistan: A spatial analysis of hazard, exposure, and vulnerability. Natural Hazards Research, 4(1), 46-55. IPCC. (2012). Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation. Cambridge University Press. IPCC. (2023). Contribution of Working Groups I, II and III to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. In H. Lee, & J. Romero, Climate Change 2023: Synthesis Report (pp. 35 - 115). Geneva, Switzerland: IPCC. doi:10.59327/IPCC/AR6- 9789291691647 Janz, T., Gassmann, F., & Gayoso de Ervin, L. (2024). Weathering Shocks: Unraveling the Effects of Short- Term Weather Shocks on Poverty in Paraguay. Policy Research Working Paper; 10970. World Bank. http://hdl.handle.net/10986/42431. Jevrejeva, S., Jackson, L. P., Grinsted, A., Lincke, D., & Marzeion, B. (2018). Flood damage costs under the sea level rise with warming of 1.5 C and 2 C. Environmental Research Letters, 13(7), 074014. Jonkman, S. (2005). Global perspectives on loss of human life caused by floods. Natural hazards, 34(2), 151- 175. Kawasaki, A., Kawamura, G., & Zin, W. W. (2020). A local level relationship between floods and poverty: A case in Myanmar. International Journal of Disaster Risk Reduction, 42. doi:https://doi.org/10.1016/j.ijdrr.2019.101348 20 Kemwita, E. F., Kombe, W. J., & Nguluma, H. M. (2023). Acquisition of land in flood risk informal setlements in Dar es Salaam: Choices and Compromises. African Journal of Land Policy and Geospatial Sciences, 6(1), 188-208. doi:10.22004/ag.econ.334630 Koks, E. E., Jongman, B., Husby, T. G., & Botzen, W. J. (2015). Combining hazard, exposure and social vulnerability to provide lessons for flood risk management. Environmental science & policy, 47, 42- 52. Kvočka, D., Falconer, R. A., & Bray, M. (2016). Flood hazard assessment for extreme flood events. Natural hazards, 84, 1569-1599. Lenderink, G., & Van Meijgaard, E. (2008). Increase in hourly precipitation extremes beyond expectations from temperature changes. Nature Geoscience, 1(8), 511-514. literature, P. F. (2020). From poverty to disaster and back: A review of the literature. Economics of Disasters and Climate Change, 4, 223-247. Lottering, S. J., Mafongoya, P., & Lottering, R. T. (2021). Assessing the social vulnerability of small-scale farmer's to drought in uMsinga, KwaZulu-Natal. International Journal of Disaster Risk Reduction, 65. Min, S., Zhang, X., Zwiers, F., & and Hegerl, G. (2011). Human contribution to more-intense precipitation extremes. Nature, 470(7334), 378-381. Nagy, G., Coronel, G., Pastén, M. B., Monte-Domecq, R., Galeano-Rojas, A., Flores, L., . . . Arana, I. (2016). Impacts on well-being and health by excessive rainfall and floods in Paraguay, Uruguay and Bolivia. Climate change and health: improving resilience and reducing risks, 475-514. Narloch, U., & Bangalore, M. (2018). The multifaceted relationship between environmental risks and poverty: new insights from Vietnam. Environment and Development Economics, 23(3), 298-327. NASA. (2023, November). The Effects of Climate Change. Retrieved from https://climate.nasa.gov/effects/ Olsen, A., Zhou, Q., Linde, J., & Arnbjerg-Nielsen, K. (2015). Comparing Methods of Calculating Expected Annual Damage in Urban Pluvial Flood Risk Assessments. Water, 7, 255-270. Retrieved from https://doi.org/10.3390/w7010255 Opondo, D. (2013). Erosive coping after the 2011 floods in Kenya. International Journal of Global Warming, 5(4), 452-466. Patankar, A. (2015). The exposure, vulnerability and adaptive capacity of households to floods in Mumbai. World Bank Policy Research Working Paper 7481. Pelling, M. (1997). What determines vulnerability to floods; a case study in Georgetown, Guyana. Environment and Urbanization, 9(1), 203-226. Ranger, N., Hallegatte, S., Bhattacharya, S., Bachu, M., Priya, S., Dhore, K., . . . Corfee-Morlot, J. (2011). An assessment of the potential impact of climate change on flood risk in Mumbai. Climatic Change, 104 , 139-167. Reckien, D., Creutzig, F., Fernandez, B., Lwasa, S., Tovar-Restrepo, M., Mcevoy, D., & Satterthwaite, D. (2017). Climate change, equity and the Sustainable Development Goals: an urban perspective. Environment and Urbanization, 2(1), 159-182. Rentschler, J., Salhab, M., & Jafino, B. A. (2022). Flood exposure and poverty in 188 countries. Nature Communications, 13(1), 3527. Rentschler, J., Salhab, M., & Jafino, B. A. (2022). Flood exposure and poverty in 188 countries. Nature Communications, 13(1), 3527. 21 Rodriguez-Oreggia, E., De La Fuente, A., De La Torre, R., & Moreno, H. A. (2012). Natural Disasters, Human Development and Poverty at the Municipal Level in Mexico. The Journal of Development Studies, 49(3), 442–455. doi:https://doi.org/10.1080/00220388.2012 Rojas, R., Feyen, L., & and Watkiss, P. (2013). Climate change and river floods in the European Union: Socio- economic consequences and the costs and benefits of adaptation. Global Environmental Change, 23(6), 1737-1751. Samarasinghe, O., & Sharp, B. (2010). Flood prone risk and amenity values: a spatial hedonic analysis. Australian Journal of Agricultural and Resource Economics, 54(4), 457-475. Tesliuc, E., & Lindert, K. (2002). Vulnerability: A quantitative and qualitative assessment. Guatemala Poverty Assessment Program. World Bank. Ultima Hora. (2019, May 13). Mas de 60.000 familias afectadas por las inundaciones en todo el pais. Retrieved from Ultima Hora: https://www.ultimahora.com/sen-mas-60000-familias-afectadas-las- inundaciones-todo-el-pais-n2819102 Winsemius, H. C., Jongman, B., Veldkamp, T. I., Hallegatte, S., Bangalore, M., & Ward, P. J. (2018). Disaster risk, climate change, and poverty: assessing the global exposure of poor people to floods and droughts. Environment and Development Economics, 23. Wodon, Q., Liverani, A., Joseph, G., & Bougnoux, N. (2014). Climate change and migration: evidence from the Middle East and North Africa. Washington, D.C.: World Bank Publications. World Bank. (2023, November). Climate Risk Profile: Paraguay. Washington D.C.: The World Bank. Retrieved from Climate Change Knowledge Portal: https://climateknowledgeportal.worldbank.org/sites/default/files/2021-04/15726- WB_Paraguay%20Country%20Profile-WEB-2.pdf Yamazaki, D., Ikeshima, D., Tawatari, R., Yamaguchi, T., O'Loughlin, F., Neal, J., . . . Bates, P. (2017). A high‐ accuracy map of global terrain elevations. Geophysical Research Letters, 44(11), 5844-5853. Zhenshan, C., & Towe, C. (2024). Pricing Coastal Amenities and Flood Hazards. Land Economics, 100(1), 109-126. doi:https://doi.org/10.3368/le.100.1.102022-0085R 22 Appendix Table A 1: Household Geocoordinates (Grey Points) Overlayed with the 1,000-year Pluvial Flood Hazard Map Source: EPH 2021 and Fathom Flood Hazard Maps. 23 Table A 2: Variable Definitions Variable Definition Exposed to flood risk Indicator variable equal to 1 if the household is estimated to be exposed to any flood risk, 0 (FE) otherwise. Flood exposure index Continuous variable ranking households by their estimated likelihood and severity of exposure to (FEI) flood risk ranging from 0, no flood exposure, to 1, extreme flood exposure. Income per capita Household monthly income per capita (current PYG). Poverty status Indicator variable equal to 1 if household's income is below the poverty line, 0 otherwise. Receives social Indicator variable equal to 1 if household receives assistance through Tekoporã or Adulto Mayor, 0 assistance otherwise. Head age Age of the household head. Head years schooling Years of schooling completed by the household head. Head female Indicator variable equal to 1 if household head is female, 0 otherwise. Head language Guarani Indicator variable equal to 1 if household head speaks monolingual Guarani, 0 otherwise. Indicator variable equal to 1 if mother of household head lived in a rural area when household Head born rural head was born, 0 otherwise. Indicator variable equal to 1 if primary sector of work of household head that is active in the labor Head sector agriculture market is agriculture, 0 otherwise. Head inactive Indicator variable equal to 1 if household head is inactive in the labor market, 0 otherwise. Household size Number of people that live in the household. Child under age of 15 Indicator variable equal to 1 if a child or children 14 years old and under are present in the present household, 0 otherwise. Adult over age of 65 Indicator variable equal to 1 if adult or adults 65 years old or over are present in the household, 0 present otherwise. Indicator variable equal to 1 if any room in the household is occupied by more than 2 people, 0 Overcrowded otherwise. Dwelling floor made of Indicator variable equal to 1 if dwelling floor made of tile, granite, porcelain, or parquet, 0 if made tile, granite, porcelain, of dirt, wood, brick, or low-quality cement. or parquet Dwelling wall made of Indicator variable equal to 1 if dwelling wall made of brick, cement, stucco, or adobe, 0 if made if brick, cement, stucco, wood, palm trunk, carton/rubber, or others. or adobe Urban Indicator variable equal to 1 if household is in an urban area, otherwise 0. Rural Indicator variable equal to 1 if household is in a rural area, otherwise 0. Asuncion Indicator variable equal to 1 if household is in the capital city of Asuncion, 0 otherwise. San Pedro Indicator variable equal to 1 if household is in the department of San Pedro, 0 otherwise. Caaguazú Indicator variable equal to 1 if household is in the department of Caaguazú, 0 otherwise. Caazapá Indicator variable equal to 1 if household is in the department of Caazapá, 0 otherwise. Itapúa Indicator variable equal to 1 if household is in the department of Itapúa, 0 otherwise. Alto Parana Indicator variable equal to 1 if household is in the department of Alto Parana, 0 otherwise. Central Indicator variable equal to 1 if household is in the department of Central, 0 otherwise. Indicator variable equal to 1 if household is in one of the departments not previously listed Other department excluding Boquerón and Alto Paraguay, 0 otherwise. 24 Table A 3: Mean Characteristics of Urban Households Exposed to Flood Risk Compared to Households Not Exposed to Flood Risk Urban Households Exposed to Not exposed Variable flood risk to flood risk Income per capita* 1,842,890.75 2,162,041.33 Poverty status* 0.21 0.16 Receives social assistance 0.14 0.13 Head age 47.54 47.56 Head years schooling* 9.55 10.23 Head female 0.43 0.43 Head language Guarani 0.27 0.23 Head born rural 0.40 0.40 Head sector agriculture 0.04 0.03 Head inactive 0.20 0.21 Household size* 4.07 3.64 Child under age of 15 present* 0.61 0.54 Adult over age of 65 present 0.22 0.19 Overcrowded* 0.27 0.20 Dwelling floor made of tile, 0.69 0.77 granite, porcelain, or parquet* Dwelling wall made of brick, 0.92 0.92 cement, stucco, or adobe Note: * on variable name indicates difference between households exposed to flood risk and not exposed to flood risk are statistically significant. 25 Table A 4: Mean Characteristics of Rural Households Exposed to Flood Risk Compared to Households Not Exposed to Flood Risk Rural Households Exposed to Not exposed to flood Variable flood risk risk Income per capita 1,252,606.63 1,204,034.15 Poverty status 0.24 0.28 Receives social assistance 0.16 0.17 Head age 48.18 48.19 Head years schooling 7.00 6.80 Head female* 0.37 0.31 Head language Guarani 0.67 0.64 Head born rural 0.87 0.89 Head sector agriculture 0.40 0.44 Head inactive 0.18 0.16 Household size 3.76 3.78 Child under age of 15 present 0.62 0.62 Adult over age of 65 present 0.21 0.22 Overcrowded* 0.31 0.25 Dwelling floor made of tile, 0.36 0.39 granite, porcelain, or parquet Dwelling wall made of brick, 0.71 0.67 cement, stucco, or adobe Note: * on variable name indicates difference between households exposed to flood risk and not exposed to flood risk are statistically significant. Table A 5: Regression Analysis of Urban Household Exposure to Flood Risk and Flood Exposure Severity (Flood Exposure Index) Exposed to flood risk (FE) (Logistic regression marginal effects) Flood exposure index (FEI) (OLS regression cond. on exposure) VARIABLES (1) (2) (3) (4) (5) (6) (7) (8) Poverty status 0.0609** 0.0241 0.0636** 0.0153 (0.0286) (0.0296) (0.0322) (0.0310) LN(Income per capita) -0.0403*** -0.0224 -0.0454*** -0.00849 (0.0140) (0.0171) (0.0171) (0.0209) Receives social assistance -0.00199 0.000725 0.0480 0.0490 (0.0316) (0.0320) (0.0363) (0.0362) Head age -0.00152* -0.00147* -0.00168 -0.00167 (0.000869) (0.000874) (0.00108) (0.00110) Head years schooling -0.00544** -0.00444* -0.00788** -0.00766** (0.00251) (0.00263) (0.00311) (0.00332) Head female -0.00298 -0.00427 0.0339 0.0336 (0.0209) (0.0209) (0.0223) (0.0224) Head language Guarani 0.0248 0.0241 -0.0110 -0.0106 (0.0260) (0.0259) (0.0279) (0.0280) Head born rural 0.00260 0.00359 -0.0208 -0.0202 (0.0217) (0.0217) (0.0226) (0.0227) Head sector agriculture 0.0123 0.0133 -0.0927 -0.0941 (0.0522) (0.0522) (0.0639) (0.0641) Head inactive -0.00891 -0.00798 0.0222 0.0225 (0.0264) (0.0263) (0.0288) (0.0288) Household size 0.0131* 0.0126* -0.00910 -0.00922 (0.00674) (0.00682) (0.00593) (0.00604) Child under age of 15 present 0.00310 -0.00150 0.0347 0.0332 (0.0264) (0.0262) (0.0287) (0.0298) Adult over age of 65 present 0.0234 0.0215 -0.0200 -0.0211 (0.0332) (0.0332) (0.0331) (0.0332) Overcrowded 0.00494 0.00162 0.0277 0.0272 (0.0289) (0.0286) (0.0281) (0.0277) Dwelling floor made of tile, granite, porcelain, or -0.0187 -0.0167 -0.0160 -0.0164 parquet (0.0238) (0.0241) (0.0281) (0.0280) Dwelling wall made of brick, cement, stucco, or 0.0307 0.0307 0.00796 0.00728 adobe (0.0325) (0.0324) (0.0500) (0.0502) Constant 0.426*** 1.082*** 0.467*** 0.589** (0.0120) (0.244) (0.0908) (0.292) Observations 2,406 2,405 2,405 2,404 509 509 508 508 R-squared 0.014 0.022 0.169 0.169 Note: Robust standard errors in parentheses. Regressions are weighted with household weights. Regressions include a set of indicator variables for departments (not shown). *** p<0.01, ** p<0.05, * p<0.1 Table A 6: Regression Analysis of Rural Household Exposure to Flood Risk and Flood Exposure Severity (Flood Exposure Index) Exposed to flood risk (FE) (Logistic regression marginal effects) Flood exposure index (FEI) (OLS regression cond. on exposure) VARIABLES (1) (2) (3) (4) (5) (6) (7) (8) Poverty status -0.0350 -0.0492* -0.00347 0.0159 (0.0229) (0.0257) (0.0240) (0.0254) LN(Income per capita) 0.0236* 0.0425** -0.0117 -0.0384** (0.0129) (0.0167) (0.0133) (0.0155) Receives social assistance -0.0508 -0.0516 -0.0182 -0.0182 (0.0360) (0.0358) (0.0377) (0.0378) Head age 0.00129 0.00113 -0.00111 -0.000847 (0.000998) (0.000997) (0.000900) (0.000893) Head years schooling 0.00306 0.00212 -0.00241 -0.00104 (0.00332) (0.00333) (0.00304) (0.00305) Head female 0.0409 0.0419* -0.0145 -0.0168 (0.0250) (0.0250) (0.0215) (0.0214) Head language Guarani 0.0326 0.0391 -0.0107 -0.0169 (0.0255) (0.0257) (0.0242) (0.0243) Head born rural -0.0319 -0.0318 0.00578 0.00536 (0.0388) (0.0388) (0.0297) (0.0297) Head sector agriculture -0.00758 -0.00457 -0.0122 -0.0180 (0.0258) (0.0258) (0.0246) (0.0244) Head inactive 0.00140 0.00704 -0.0432 -0.0496 (0.0348) (0.0356) (0.0317) (0.0315) Household size -0.00734 -0.00675 -0.00156 -0.00294 (0.00737) (0.00734) (0.00691) (0.00688) Child under age of 15 present -0.00690 0.00294 -0.0251 -0.0389 (0.0288) (0.0288) (0.0254) (0.0259) Adult over age of 65 present -0.00587 -0.00124 -0.00991 -0.0184 (0.0393) (0.0396) (0.0381) (0.0384) Overcrowded 0.0871*** 0.0889*** -0.0222 -0.0262 (0.0337) (0.0338) (0.0240) (0.0240) Dwelling floor made of tile, granite, -0.0200 -0.0257 -0.0184 -0.0112 porcelain, or parquet (0.0235) (0.0239) (0.0226) (0.0229) Dwelling wall made of brick, cement, 0.0113 0.00929 -0.0175 -0.0140 stucco, or adobe (0.0252) (0.0253) (0.0236) (0.0236) Constant 0.479*** 0.638*** 0.613*** 1.146*** (0.0108) (0.184) (0.0728) (0.229) Observations 2,240 2,240 2,240 2,240 591 591 591 591 R-squared 0.000 0.002 0.090 0.100 Note: Robust standard errors in parentheses. Regressions are weighted with household weights. Regressions include a set of indicator variables for departments (not shown). *** p<0.01, ** p<0.05, * p<0.1