Policy Research Working Paper 11259 Assessing the Macroeconomic Impacts of Stochastic Climate Shocks in Paraguay Heather Ruberl Pui Shen Yoong Diego Castillo Brent Boehlert Economic Policy Global Department A verified reproducibility package for this paper is November 2025 available at http://reproducibility.worldbank.org, click here for direct access. Policy Research Working Paper 11259 Abstract This paper examines the macroeconomic consequences of heatwaves and floods on agriculture and livestock pro- climate shocks in Paraguay, a small and open agricultural duction and labor productivity in Paraguay. The findings exporter that is highly exposed to droughts and floods. reveal significant vulnerabilities in agricultural outputs Utilizing a modified version of the World Bank’s Mac- and exports across a range of climate scenarios. The paper ro-Fiscal Model, the analysis integrates stochastic climate presents a comprehensive analysis across different climate shocks drawn from external biophysical models to assess scenarios ranging from mild to severe and quantifies the the impacts of extreme weather events such as droughts, range of potential impacts on key economic indicators. This paper is a product of the Economic Policy 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 psyoong@worldbank.org. A verified reproducibility package for this paper is available at http://reproducibility.worldbank. org, click here for direct access. RESEA CY LI R CH PO TRANSPARENT ANALYSIS S W R R E O KI P NG PA 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 Assessing the Macroeconomic Impacts of Stochastic Climate Shocks in Paraguay Heather Ruberl1 , Pui Shen Yoong2 , Diego Castillo3 , and Brent Boehlert4 World Bank∗ 1,2 3,4 Industrial Economics, Incorporated Authorized for distribution by Shireen Mahdi, Practice Manager, Economic Policy Global Department, World Bank Group JEL Classification: C10, C50, E37, Q54 Keywords: Paraguay; macro-structural model; economic modeling, droughts, floods, heatwaves, cli- mate change ∗ World Bank Economic Policy Department. This paper was produced under the guidance of Oscar Calvo- Gonzalez, Marianne Fay and Shireen Mahdi. The authors would like to thank Fernando Im, Daniel Navia Simon, Sebastian Miller, Arthur Bragan¸ ca and Mariana Conte Grand for their valuable feedback and advice. This study benefited from close collaboration with the Paraguayan Ministry of Economy and Finance (MEF). The authors thank the following MEF officials for their inputs: Felipe Gonzalez Soley, Rolando Sapriza, Elisa Vera. 1 Introduction Climate shocks are increasingly being recognized as “macro critical” risks that are essential to under- stand when considering a country’s growth and development prospects. Extreme weather events such as floods, droughts and heatwaves can have direct impacts on economic outcomes. These events could lead to large, unplanned reductions in government revenues due to disruptions in economic activity, and/or to increases in government spending for relief and reconstruction measures following a climate shock. By integrating climate risks into macroeconomic models, policymakers can better understand their potential short- and long-term impacts on economies and people, and thus prioritize the necessary policies and investments to cope with them. There are several approaches to modeling the physical impacts of climate shocks on the economy (see for Greening Financial Systems (2024) for a summary). This paper demonstrates one approach for Paraguay, a small, open economy in South America, using a customized version of the World Bank’s Macro-Fiscal Model or MFMod (Burns et al. (2019)). MFMod is a macro-structural, general equilibrium model where the short-run parameters are estimated econometrically using country-specific historical data, and the long run is determined by a set of identities and behavioral equations derived from neoclassical economic theory. To model the impact of climate shocks on the Paraguayan economy, we incorporate into MFMod “vectors of shocks” produced by a separate set of external biophysical models, which convert projected changes in climatic factors such as temperature and precipitation into impacts on specific sectors. In the case of Paraguay, we model the climate-change induced impacts of droughts and heatwaves on crop yields, livestock yields and labor productivity, and of floods on capital stock. This is similar to the approaches followed in the Sahel region (Abalo et al. (2025)), Uruguay (Giuliano et al. (2024)) and Pakistan (Burns et al. (2021)). Paraguay is the ideal setting for this study for several reasons. First, despite a small population of around 6 million people, its 406,752 square kilometers contain two distinct ecosystems: (i) a tropical savanna climate with lowland plains subject to floods in the Eastern region, and (ii) a hot and semi- arid climate in the sparsely-populated Western region, which is part of the ecosystem known as the Gran Chaco (World Bank (2021)). This landscape is prone to different types of extreme weather events. Data from the Emergency Event (EM-Dat) database indicates that Paraguay has suffered at least nine droughts1 and 17 riverine floods in the last century (Benitez and Domecq (2014) and World Bank (2021)). Second, it is a small, open economy that depends on exports of mostly unprocessed agricultural exports, and acts as a price taker in global agricultural markets. Climate and other external shocks thus have a direct and rapid impact on the Paraguayan economy through the real, trade and financial market channels. Third, Paraguay’s central institutions have generally maintained a stable macroeconomic framework, with low inflation and generally small fiscal and external imbalances over the last two decades. Climate patterns are therefore one of the main sources of risk to an otherwise stable outlook for growth and poverty reduction. We find that climate shocks could have larger effects on Paraguay’s economy than previously thought. A severe, once-in-a-century drought is estimated to cause a 6.9 percent reduction in GDP in the year of the shock, a worsening of the fiscal balance by 0.8 percentage point of GDP, and an increase in the debt-to-GDP ratio by almost 4 percentage points in the year of the shock. While the economy recovers quickly in the following year, it only returns to its baseline trajectory five years later. With climate change, these types of stochastic climate shocks are expected to increase in frequency and severity. In a combined scenario of droughts, heatwaves and floods, annual real GDP is expected to be between 2.8 and 5.4 percent lower by 2050, and the debt-to-GDP ratio is projected to increase 1 This may be an underestimate. Using the Standard Precipitation Index or SPI, Benitez and Domecq (2014) count 12 severe to extreme agricultural droughts (SPI-3) between 1967 and 2009 alone 2 by 4.5 percentage points. These estimates may be interpreted as upper bounds as they do not consider technological change, adaptation policies and other factors that are likely to offset some of the modeled impacts. This paper makes several contributions. First, it provides Paraguayan policymakers with a flexible framework to understand the macroeconomic impact of climate change through different channels. Second, compared to existing studies on the aggregate impact of climate change on the Paraguayan economy, it considers a wider range of climate models under a more plausible set of potential climate scenarios. Third, in addition to looking at the impact of climate change on crop yields and worker productivity, it examines the impact of heat stress on livestock productivity and the impact of floods on capital stock. These two channels are understudied in the available literature for Paraguay. Finally, in contrast to most existing studies which only use deterministic models, it simulates the impact of stochastic or random climate shocks on the economy, accounting for the inherent uncertainty of climate change. This paper is organized as follows: Section 2 provides a primer on the Paraguayan economy; Section 3 summarizes what we know about the impacts of climate shocks on the Paraguayan economy; Section 4 outlines the modeling approach; Section 5 discusses the resulting impacts of a once-in-a-century drought and, subsequently, Section 6 discusses the results of the stochastic climate shocks. Finally, Section 7 concludes with a discussion of the limitations of this approach and Section 8 discusses relevant policy recommendations. 2 Context Paraguay is an upper-middle-income country (UMIC) endowed with abundant natural resources such as fertile soils, rivers and forests. Renewable natural capital wealth accounts for a quarter of its per capita wealth, twice the proportion of the average LAC or UMIC country’s wealth (World Bank (2024a)). Buttressed by a stable macroeconomic framework and an open trade and investment regime, Paraguay has successfully leveraged these natural endowments to drive growth and poverty reduction over the last two decades. From 2002 to 2024, per capita income grew by an average of 2.9 percent annually, faster than many Latin American and Caribbean (LAC) countries. Robust growth was primarily driven by high commodity prices in the first half of the period, which, along with productivity improvements in commercial agriculture, helped propel Paraguay into a major global exporter of soy and beef. It also exports hydropower to Brazil and Argentina through jointly-owned dams on the Paraguay and Paran´ a Rivers. Despite some diversification, Paraguay’s productive structure still heavily relies on natural re- sources. Although the primary sector consisting of agriculture, livestock and forestry only accounts for 11 percent of GDP, this proportion increases to 29 percent of GDP when including hydropower and manufacturing sectors linked to agricultural inputs,2 and would be even higher if it included the back- ward and forward linkages from agriculture to services (such as transport and logistics). Paraguay’s small industrial sector relies heavily on domestic inputs from the agricultural sector. The 2014 social accounting matrix indicates that 40 percent of the industrial sector’s domestic intermediate inputs come from the agriculture sector. In turn, this industry sector represents 33 percent of exports, with primary agricultural goods representing an additional 27 percent of exports. The importance of these natural resource-linked sectors to growth can be seen in the high positive correlation between their total value added and Paraguay’s overall gross value added over time (Figure 2 This comprises the processing of meat, oils, dairy, cereals, sugar, other food, leather, wood, pulp, and paper products. 3 1. Natural resource-linked sectors are also important to the trade balance and external accounts, as 81 percent of Paraguay’s direct exports consist of agricultural goods (primarily soy), beef, and hydropower. The performance of these sectors is also critical to government revenues. An estimated 67 percent of the variation in tax revenues over the last two decades was associated with the variation in agricultural output (World Bank (2024b)). Figure 1: Real value added, year-on-year percentage change (%) Droughts have regularly had adverse impacts on Paraguay’s growth. According to the FAO, Paraguay experiences mild droughts every two years on average and at least one severe drought every decade. In the recent decade, such droughts appear to have become more frequent and intense, occurring in 2012, 2019 and again in late 2021 and early 2022. In the latter episode, soybean production fell by nearly 70 percent, translating into an estimated US$3 billion loss in foregone export earnings. Real GDP only grew by 0.2 percent in 2022. The economy has since recovered, but it remains vulnerable to climate shocks. Although real GDP expanded by 4.2 percent in 2024, uneven rainfall along the major rivers led hydropower production to halve compared to the previous year and reduced growth by 0.4 percentage point. In addition, large swaths of the country’s sparsely populated Western region continue to experience dry conditions, affecting livestock production. While some producers have adapted by moving cattle to other pastures and in feedlots, the persistence of dry conditions has led the cattle herd size to decline by 1 to 2 percent annually on average since 2020 and reduced longer-term investments the sector (USDA (2024)). Extreme weather events have also slowed the pace of poverty reduction. Droughts have increased logistics costs as grains producers and other exporters had to rely on land transport rather than major rivers, which typically carry the bulk of Paraguay’s domestic trade. In 2022, the severe drought, along with higher global fuel prices, prompted a rise in inflation and hence a temporary increase in extreme poverty. As for heatwaves, Janz et al. (2024) estimate that these could increase poverty levels by 4.2 percentage points in rural areas (which tend to depend on subsistence agriculture) and by 1.7 percentage points in urban areas, on average. While floods are less common, they have also had catastrophic effects. Floods in March 2019, for example, damaged 347 schools, forced 19,500 families to evacuate, and killed 16 people. Such floods disproportionately affect the poor, as nearly a third of the poor live in floodplains. In urban areas, the share of the poor population exposed to flooding is 1.2 times higher than those not exposed to flood hazards (Janz et al. (2024)). 4 3 Literature Review Despite the importance of the topic, few studies have provided quantitative estimates of the economic impact of climate shocks in Paraguay. In 2014, CEPAL had projected that Paraguay could lose 0.6 to 1 percent of its annual GDP until 2100 due to the effects of climate change on crops yields, disease prevalence, and the damage to capital stock and losses of pastures and forests caused by droughts and floods (CEPAL (2014)). This figure was derived from bottom-up dollar estimates of losses from historical events to which various discount rates were applied, and a series of qualitative assumptions about the intensity of climate damages in the future. Following IPCC convention at the time, the study described the results for two climate scenarios, A2 and B2, which applied simplified assumptions on future temperature and precipitation increases. Banerjee et al. (2021) uses a computable general equilibrium model to evaluate the economic im- pacts of climate change in 20 Latin American countries including Paraguay on three outcome variables: agricultural yields, labor productivity in agriculture, and economy-wide labor productivity. The paper finds that Paraguay is expected to suffer among the largest GDP losses (-1.3 percent from the baseline) from 2021 to 2050 due to the combined effects of reduced crop yields and lower economy-wide labor productivity. Livestock exports are estimated to increase by 1.6 percent in Paraguay, but the paper offers no explanation for this result. The World Bank (2025) Paraguay Country Climate and Development Report, which also uses MfMOD, estimates that the country’s GDP will decrease by 0.1 to 1.1 percent annually until 2050 from the baseline, assuming a more optimistic future with lower emissions (SSP1-2.6). Should the more pessimistic scenario with higher emissions prevail (SSP5-8.5), GDP could be reduced by 0.5 to 3.1 percent from the baseline. Labor productivity losses and crop yields are the main channels of impact through which these losses are expected to occur, whereas the impact of climate shocks on hydropower generation is expected to be economically insignificant. The latter finding reflects the high uncertainty regarding the direction of impact of climate change on hydropower production (see, for example,Bravo et al. (2014) and World Bank and OLADE (2023)). It may also reflect the fact that prices of hydropower exports from Paraguay to Brazil are bilaterally negotiated and do not reflect market conditions, thus attenuating the negative and positive impacts of climate shocks. More recently, the Government of Paraguay has also begun to contribute to literature on the impact of drought shocks using macroeconometric models. Using a structural vector autoregressive (SVAR) model, researchers at the Central Bank estimate that an increase in the drought index by one standard deviation causes significant peak declines in output and investment two quarters after the shock occurs (B´ aez et al. (2023)). Despite the importance of agriculture and hydropower in the export basket, they do not find significant impacts on the trade balance. Separately, the Ministry of Economy and Finance used a Growth-at-Risk model to show that a drought shock negatively impacts agricultural GDP projections, while electricity and water GDP projections remain stable (MEF (2024)). They surmise that hydropower plants maintain consistent production in the face of droughts due to efficient management and preventive maintenance. This result, however, is not aligned with historical experience. In 2001, 2014, 2017, 2019 and 2021, the volume of hydropower exports fell by between 19 and 26 percent compared to the previous year due to dry conditions. In 2024, hydropower export volumes halved due to uneven rainfall along the waterways. Several other studies provide quantitative estimates of the impact of climate change on particular sectors without translating them into aggregate macro impacts. For example, Ervin and Gayoso de Ervin (2019) estimate that a 1 percent increase in average maximum temperatures could lead to a 5 per- cent reduction in agricultural productivity, translating to a 1 percent decrease in caloric consumption and hence potentially threatening food security. Conversely, Benitez Rodriguez et al. (2024) suggests 5 that Paraguay’s production and exports of wheat and soybeans may increase due to climate change as other regions of the world experience larger decreases in crop productivity compared to Paraguay, giving it a competitive advantage. On labor productivity, Janz et al. (2024) find that rural outdoor workers suffer 15-30 percent productivity declines during heatwaves, disproportionately affecting the 23 percent of the labor force that work in agriculture. This analysis, however, did not examine the aggregate economic impacts of these losses. The literature indicates three gaps, which we try to address in this paper. First, few papers use modeling approaches that connect both the sector-specific and economy-wide impacts of climate shocks. This approach is particularly relevant from a policymaking perspective not only to quantify the amount of expected damages from these shocks and plan for contingency financing if needed, but also to sequence and prioritize investments in adaptation in specific sectors. Second, the macroeconomic impacts of droughts on livestock and of flooding have been understudied in the Paraguayan context. Third, most of these papers only model deterministic shocks, and mostly under unlikely optimistic and pessimistic climate futures (SSP 1-2.6 and SSP5-8.5, respectively). By also modeling stochastic climate shocks, our paper captures the inherent uncertainty in climate change and hence provides a more realistic view of the potential outcomes. It also focuses on a set of shared socioeconomic pathways or SSPs that are more realistic (SSP2-4.5 and SSP3-7.0). 4 Modeling approach The modeling approach can be roughly described in four steps: First, we run a number of General Circulation Models or GCMs3 to produce a set of climate change impacts for two representative scenarios. Second, we estimate the impact of droughts on crop and livestock production and the impact of heat waves on labor productivity in Paraguay using biophysical models. Third, we extract exceedance curves from these models to produce the shocks. Fourth, we incorporate these shocks into a macroeconomic model, MFMod, to estimate the economy- wide impacts of climate change relative to a ‘no additional climate change’ future. The latter exercise has two parts: we first analyze the effects of a single drought on relevant macroeconomic variables and their recovery dynamics. We then run 1,000 stochastic simulations using the exceedance curves to illustrate the range of potential macroeconomic impacts of these extreme weather events. These steps are described in detail below and in the appendix. Figure 2: Integrating climate and macro modeling 3 GCMs are advanced computer models that simulate Earth’s climate system. They use mathematical equations to represent physical processes in the atmosphere, oceans, and land surface, dividing the planet into a three-dimensional grid of cells. GCMs are used to understand climate dynamics, forecast weather, and project future climate changes under various scenarios. Despite limitations, they are considered as the best tools available for predicting climate responses to increased greenhouse gases. 6 4.1 Climate modeling 4.1.1 Future climate conditions The potential future impact of climate change on the frequency and intensity of climate shocks is highly uncertain. This is because the trajectory of global emissions over the next few decades depends on many factors such as the evolution of low carbon technologies, policies that encourage decarbonization, social attitudes towards climate change, and so on. Variations in global climate sensitivity across different GCMs are also a major source of uncertainty. To capture both inherent climate variability and uncertainty of future projections, we assess future climate scenarios using a wide range of GCMs. We consider results from 29 different GCMs under two climate scenarios, namely SSP2-4.54 and SSP3-7.05 . In total, 55 GCM runs are available, so not all GCMs have results for both climate scenarios. To establish a counterfactual, ’no additional climate change’ reference scenario, we extrapolate observed climate data and modeled hindcasts across all 29 GCMs through to 2050. We then isolate the impact of climate change by measuring the deviation of each of the 55 climate projections from this baseline scenario. Once the impacts have been modeled for each GCM run, we calculate annual averages for the 2031-2050 period, and then select the top and bottom ten annual average impacts from the GCM runs to present a range of possible climate change impacts. The top ten GCM runs can be thought of as a more optimistic end of the range of climate outcomes, while the bottom ten GCM runs can be thought of as a more pessimistic end. In general, the pessimistic future is hotter and/or drier than the optimistic future, although the sensitivity to changes in climate variables differs across biophysical models. Figure 3: Left panel: Projected mean temperature; Right panel: Projected precipitation. Source: World Bank Climate Change Knowledge Portal using CMIP6. As shown in Figure 3, the multi-model ensembles show higher average temperatures for Paraguay, 4 SSP2-4.5 is a climate change scenario that assumes moderate emissions reductions and socioeconomic trends that follow historical patterns. It is sometimes called the “middle of the road” scenario because it is consistent with what countries have pledged to do so far about climate change. Under this scenario, carbon dioxide emissions remain around current levels until the middle of the century, then start to fall but do not reach net zero by 2100. The scenario projects a global average warming of around 2.7°C by the end of the 21st century, with a likely range of 2.1°C to 3.5°C. 5 SSP3-7.0 is a medium to high reference scenario for climate change that assumes no additional climate policy efforts. It is a “regional rivalry” scenario where countries prioritize national and food security over other issues. This scenario assumes high non-CO2 emissions, including high aerosol emissions. This scenario projects that by the end of the century, average global temperatures rise by 3.6°C. 7 predicting almost 4°C higher temperature by 2100 under the SSP3-7.0 scenario and almost 3°C in SSP2- 4.5. While these same models show average annual rainfall remaining broadly at historical levels, the variance of rainfall is projected to increase under climate change. The timing of rainfall (shown in Figure 4 will also shift, with lower precipitation expected in September and higher precipitation in November (relative to historic rainfall patterns). The annual average precipitation does not capture the change in frequency and intensity of droughts, extreme rainfall and flooding events. Figure 4: Projected precipitation variance by month. Source: World Bank Climate Change Knowledge Portal using CMIP6. 4.2 Linking future climate conditions to specific impact channels In the next step, we model climate-induced impacts on rainfed crop yields, livestock yields, and labor productivity by 2050 using biophysical models, which are used to convert changes in temperature and precipitation in the country. The impacts on each of these channels are then modeled at the resolution of the climate data and aggregated nationally to develop the exceedance probability functions or exceedance curves. Exceedance curves graphically represent the relationship between the frequency of events (hori- zontal axis: return period in years) and their associated losses or damages (vertical axis: percentage impact of event on variable of interest). The return period reflects the expected annual frequency of an event of a certain magnitude, with shorter periods indicating more frequent, less severe events and longer periods representing rarer, catastrophic events (e.g., a 10-year event is expected to occur once every 10 years, or alternatively, an event with an annual probability of 1/10=10%). The curve shows the expected loss for different probabilities, allowing us to model the impact of climate change by linking event likelihoods to potential economic impacts. Following the methodology described in Appendix A, we construct the relevant exceedance curves by pooling annual productivity results for each GCM run between 2031-2050 and deriving probabilities. Exceedance curves were constructed for four sets of results: • (1) a no-climate change reference scenario, • (2) results considering the entire set of 55 GCM runs for SSP2-4.5 and SSP3-7.0, 8 • (3) results from the 10 GCM runs (from the same pool of 55) that results on the highest mean annual productivity levels, and • (4) results from the 10 GCM runs that results in the lowest mean annual productivity levels. While (2) represents a likely exceedance curve in the future, (3) and (4) represent “what if” scenar- ios in which climate change effects result in the most pessimistic or optimistic impacts on Paraguay. The selected 10 GCM runs are different for crops, livestock, and labor. The results from the corre- sponding exceedance curves are discussed below. 4.2.1 Droughts Climate change may impact crop yields through changes in rainfall patterns, water availability, increas- ing evaporative demands, and extreme heat as temperatures rise. The analysis focuses on crop-level effects, for representative crops that cover 95 percent of the harvest area and sector revenues. Crop ar- eas and production at the department level come from the 2022 Paraguay Agriculture Census, provided by the World Bank. Figure 5 presents the impacts of droughts on rainfed crop yields and livestock yields across a ‘no additional climate’ change future (i.e., the hindcasts curve) and the top 10 GCM (highest yields) and bottom 10 GCM runs (lowest yields) in terms of average impacts. These impact figures represent the difference between a maximum attainable yield and the actual yields obtained due to climatic conditions in Paraguay. Historically, we can see that droughts already have had a large impact on crop and livestock yields in Paraguay. The biophysical models indicate that a 1-in-100 year drought event has historically been associated with a 40 percent reduction in crop yields and a 21 percent reduction in livestock yields. Under climate change, crop yield impacts from droughts are generally expected to be more severe than without climate change. The average impact on climate change across all GCMs (shown by the medium green line) indicates that a drought with a given return period is expected to cause higher losses in crop outputs. Under the bottom 10 GCMs (more pessimistic future), which represent a more pessimistic future for crop production, yield impacts are even higher for a given return period. Under the top 10 GCMs (more optimistic future) there is an improvement in crop yields, with droughts becoming less severe at a given return period. Figure 5: Exceedance curves for drought impact on agricultural productivity by sector For livestock, the analysis considers increasing heat stress on animals, causing reductions in pro- ductivity, as well as changes in the availability of grazing pastures impacting animal feed intake. Our 9 analysis focused on the main species, namely cattle, chickens, and swine, for the three most important products: meat, milk, and eggs. In the case of livestock, the top 10 GCMs exceedance curve is similar to the hindcast curve. This indicates that, unlike crops, there is no anticipated increase in livestock yields even under the ‘best case’ scenario. This outcome reflects the reality that even the more optimistic GCM scenarios for Paraguay provide little to no benefit for livestock productivity. An overall negative impact from climate change on livestock yields is indicated, with projections from all GCMs and particularly the bottom 10 GCMs projecting more pronounced negative impacts than under historical conditions. 4.2.2 Heatwaves Climate change can impact labor productivity by increasing workday temperatures and decreasing the number of hours an individual can work. To estimate labor heat stress impacts due to climate change, we calculate workday wet bulb globe temperatures for indoor and outdoor environments as a measure of heat stress to derive labor productivity losses across three economic sectors, namely agriculture, industry, and services. These productivity estimates are expressed relative to an optimal baseline - i.e., conditions with no heat stress or heatwaves affecting worker output. Importantly, the analysis does not account for potential labor shifts across sectors or geographic regions in response to changing climate conditions, and instead assumes a fixed distribution of workers. Figure 6 summarizes the impacts of higher temperatures and more frequent heat waves on labor productivity for selected return intervals for each sector of the economy. Agricultural workers, who tend to conduct physical labor outside are the most exposed to climate change. Extreme temperature years will become more severe. Historically, a 1-in-100-years heatwave would reduce labor productivity in the agriculture sector by 16.5 percent in Paraguay in a ’no additional climate change’ scenario. With climate change (all SSP2-4.5 and SSP3-7.0 GCM run), this figure is expected to increase to almost 25 percent. Similar but more moderate trends can be observed for workers in other sectors, where the work is more likely to take place indoors, and to be less physical in nature. The services sector is expected to be the least affected. These estimates can be interpreted as a lower bound, as the analysis only considers impacts on the formal workforce. Figure 6: Exceedance curves for heatwave impact on labor productivity by sector 10 4.2.3 Floods For floods, we utilize global exceedance curves from the United Nations Office for Disaster Risk Re- duction (United Nations Office for Disaster Risk Reduction (UNDRR) (2015)), also known as the Global Assessment Report on Disaster Risk Reduction - or GAR15 dataset. As suggested in Myhre et al. (2019), for each degree Celsius increase in temperature, the frequency or probability of extreme precipitation and flooding events of any given size is assumed to double. For example, a once-in-a- century flood is expected to occur much more frequently as temperatures rise (as shown in Figure 7). Consequently, the expected damages to the capital stock associated with these events will happen more often. Figure 7: Exceedance curves for flood impact on the capital stock Since the baseline for the historical curve is 2015, when average temperatures in Paraguay were already 0.8°C higher than historical averages, we assume additional warming of 2.0°C by 2100 for the SSP2-4.5 scenario, and 2.8°C for the SSP3-7.0 scenario. This means that a flood that historically would only occur every 1-in-500 years is expected to be associated with damages of 4 percent of the capital stock. By 2100, this type of flood is expected to occur every 218 years under the SSP2-4.5 scenario. Under the warmer SSP3-7.0 scenario, this type of event is expected to occur more frequently, around every 125 years. 4.3 Macroeconomic modeling We build a model for Paraguay based on the workhorse macroeconomic projection model used by the World Bank, MFMod. For further detail on the macroeconomic modeling framework used to translate biophysical impacts into economic outcomes, and the constraints of this approach, see Abalo et al. (2025), including Appendix B. Macrostructural models such as MFMod make a concerted effort to estimate the economic and behavioral determinants of economic variables, developed to be both consistent with economic theory in the long run, and with the observed dynamics of the economy in the short term. As a result, the speed of adjustment of each country-specific model to its economically determined long-term equilibrium is estimated to reflect the historical behavior of the economy. Such models are thus particularly well fitted to reproduce the short-to-medium-term dynamics of an economy as a response to shocks, while 11 also providing fairly granular analysis. This contrasts with other approaches such as computed general equilibrium (CGE) models, which can provide a good amount of sectoral detail, but are not designed to assess the type of short-term dynamics that are relevant for this exercise. A stylized representation of the economic linkages in the MFMod fitted to the Paraguayan economy is shown in Figure 8. The modeling of GDP comprises three standard measurements: GDP from the (i) production expenditure side, (ii) expenditure production side, and (iii) income side (see Burns et al. (2019) for more details). Climate shocks like droughts, heatwaves and flooding are modeled as impacting on the production side of the economy, which then flows through to exports, and domestic incomes and expenditure. The standard MFMod models GDP from the production side of the economy as three sectors: agriculture, industry, and services. For the purposes of this analysis, the Paraguay MFMod disaggregates the production side of the economy further to allow for better analysis of the impact of climate shocks on different sectors. The agriculture sector is split into livestock and crops, and industry is split into agriculture related industry, electricity and water, and other industries. Figure 8: Paraguay MFMod simplified economic map Production in the crops and livestock and agribusiness sectors directly impacts on the volume of exports, which are assumed to be partly supply driven. The volume of exports in turn has an impact on the exchange rate, which appreciates when exports volumes increase, and depreciates when they decrease. We also incorporate a feedback mechanism whereby the level of the debt-GDP ratio is linked to the risk premium the government pays on its USD denominated debt. We estimate that for every 10 ppts increase in the debt-GDP ratio, the risk premium on government USD denominated debt increases by roughly 100 basis points.6 (See Appendix B for more detail) Potential GDP, measured by the production function, is the supply potential of the economy and anchors the real side of the model. It determines how much output can be produced when all resources in the economy are fully employed (given existing distortions, technology and preferences). Potential output (Yt∗ ) is a function of Total Factor Productivity (TFP) (At ), structural employment (Nt∗ ), the capital stock (Kt−1 ), and the wage share of income (α). These components are combined in a standard 6 This aligns well with historically experience: from 2011 to 2023, the Paraguayan debt-GDP ratio increased from 8 percent of GDP to 39 percent of GDP, or by around 30 percentage points. Over this same period, the implicit risk premium on external debt increased from an estimated -2.6 to 1.5 percent, an increase of roughly 300 basis points. 12 Cobb-Douglas production function: 1−α Yt∗ = At Nt∗α Kt −1 (1) Adjustment mechanisms in the model serve to adjust demand in line with supply over the long run. In the short run, output is driven by demand, or GDP from the expenditure side (household consumption, government consumption, and investment). Here there is consistency in the sense that shocks to final demand will affect production, but production is constrained based on its factor use and factor costs (wages and cost of capital). Factor use and factor costs together determine GDP from the income side. Labor demand, wages, and output are thus jointly determined. Labor and wage outcomes affect consumption decisions of households and these in turn affect overall prices. Prices impact the user cost of capital, which affects investment. Investment and consumption determine final demand, which then has an impact on industry output thus closing the link. We further expand the MFMod for Paraguay to allow for losses and damages due to climate shocks. The standard Cobb-Douglas specification is modified to account for damages from climate shocks, including: i) reductions in agricultural TFP due to droughts, ii) reduction in sectoral labor productivity from heatwaves and iii) the impact of flooding on capital stock. Droughts may impact crop and livestock yields through changes in rainfall patterns, water avail- ability, increasing evaporative demands, and extreme heat as temperatures rise. This is a reduction in output given the same inputs (capital and labor is unaffected), so droughts are modeled as a sector- specific productivity shock. Heatwaves can impact labor productivity as higher temperatures during the work day may de- crease the number of hours an individual can work. Temperature directly affects the productivity of labor, where the effect intensifies for labor types that are outdoors and are conducting more intense physical work, such as agriculture and construction. Floods can cause significant damage to infrastructure and other capital assets. This damage must then be repaired, diverting spending away from investment into new productive capital. Floods are therefore modeled as a shock to the capital stock of the economy, with flow-on implications for investment spending. The following sections discuss the modeling of each of these climate shocks in turn. 4.3.1 Droughts The modeling connects livestock and crop yields to climate shocks and then maps them to economic activity. To account for different productivity shocks in each sector, we construct a production function for each sector, which is modified to account for losses due to climate shocks. The agricultural damage or losses due to the drought di,t is incorporated into the model as reduced TFP, because the same amount of capital and labor produces a reduced amount of output compared to previous years. ∗ 1−α Yi,t = (1 − di,t )At Nt∗α Kt−1 (2) Output of the agriculture sectors is then a function of their respective production functions. Dam- age or losses are allowed to have an autoregressive component, calibrated to the persistence of shocks on output, described in Appendix B. 4.3.2 Heatwaves To capture the impact of heatwaves on sectoral output, potential output is amended to capture labor productivity losses. As in Burns et al. (2021), we distinguish between work hours L and effective work 13 ˆ = (1 − h)L where h is the percentage reduction in hours worked due to the heatwave event. hours L However, since we have proxies for sectoral potential output, the labor productivity losses are incorporated into these sectoral production functions: ∗ 1−α Yi,t = At (1 − hi,t )α Nt∗α Kt−1 (3) Where hi,t is the percentage reduction in hours worked in each sector i. 4.3.3 Floods As described in Burns et al. (2021) the production function is amended to account for the damaged stock of capital (DSt ). The economic impact of a unit of damaged capital is assumed to be equal to the average productivity of capital (Y /K ) times the damaged capital. The level of damaged capital stock at time t is determined according to the following formula: REP DSt = DSt−1 + Dt − It (4) REP where Dt is new damages in time t and It is repairs carried out. Capital not affected by natural disasters (Kt ) grows with capital investment (It ) less investment spending on repairs due to flooding: REP Kt = (1 − δ )Kt−1 + It − It (5) Because capital repairs can take time, and there is not always budget or resources immediately available, it is assumed that reconstruction investment (both public and private) cannot exceed 25 percent of total investment spending in that year. We assume that the repairs that occur are carried out by the government instead of investment into new capital. REP It = min[DSt−1 , 0.25It ] (6) To reflect the idea that the economic impact of damaged capital exceeds that of marginal capital, the standard potential output equation is modified to explicitly account for the higher productivity of destroyed capital (see Hallegatte and Vogt-Schilb (2016) for an in-depth treatment): 1−α Y Yt∗ = At Nt∗ αKt −1 − DSt (7) K 4.4 Methodology for incorporating stochastic climate shocks into the model We iteratively construct and incorporate 1000 climate shocks one by one. For a more detailed descrip- tion of this methodology, see Appendix C. Monte Carlo simulations are used to model stochastic climate shocks by generating random events based on the probabilities of extreme climate occurrences, such as heavy rainfall or flooding, drawn from exceedance curves. These curves indicate the likelihood of events of different magnitudes. Random numbers between 0 and 1 are sampled, with each number corresponding to a cumulative probability, and are then mapped to specific event magnitudes using the inverse of the exceedance probability function. This process generates a range of possible climate events and the associated damage to the capital stock (Dt ), over multiple iterations. These shocks are incorporated into our MFMod for Paraguay, allowing and assessment of their potential impacts on critical economic variables such as GDP and fiscal outcomes. By running a large 14 number of simulations, the Monte Carlo method captures the uncertainty and variability of extreme climate events, thus providing a robust framework for understanding how rare but severe events could affect long-term economic outcomes. 5 Results: Model dynamics for a 1-in-100-years drought To test the model dynamics, we simulate a 1-in-100-years drought. The yield reductions are drawn from the drought exceedance curves described in section 4.2.1 for events with this historical frequency, i.e., reduction in crop yields by 33 percent, and a reduction in livestock yields by 21 percent. In the second year of the drought, crop and livestock outputs are projected to remain 5 percent below their baseline levels (Figure 9). This is based on econometric estimation of the persistence of shocks in each sector. For livestock, we estimate that 15 percent of a shock will persist from one quarter to the next, while for crops, just 5 percent of a shock will persist into the next quarter. This reflects the longer lasting impacts of droughts on herd health, while crops can be replanted year-to-year. Figure 9: Impulse response functions for selected variables - Demand and supply. Note: Percent deviation from baseline shown. In the long run of our model, the prices for both livestock and crops follow global commodity prices, since Paraguay is a relatively small exporter, and so is a price taker (rather than a price maker) in global markets. However, in the short run, events like droughts can impact on the domestic prices of livestock and crop production. For livestock, drought events and other domestic price pressures are the short-run driver of price changes, explaining an estimated 90 percent of price movements. For crops, non-traded crop prices drive an estimated two-thirds of short-run price movements, while domestic price pressures are less significant. This reflects the fungibility of most crops, which can be imported 15 more easily to replace local inputs of agroprocessing in the case of a drought. Livestock on the other hand is more difficult to import ‘live’ as an input into slaughtering and meat product manufacturing. The 1-in-100-years drought drives a 20 percent increase in livestock producer prices, reflecting the large reduction in productivity in the sector. Non-traded crop prices increase by even more, reflecting the larger productivity impact in that sector. However, this effect is moderated by traded-crop prices, which move in tandem with the exchange rate, resulting in a more moderate 14 percent increase in producer prices for crops. Industrial production related to agriculture (agroprocessing) is also impacted, both by higher livestock and crop output prices, and by a reduction in the availability of these intermediate inputs into production. We estimate that shocks to livestock production have a 15 percent flow-through to industrial production related to agriculture in the short run, while shocks to crop yields have a much smaller impact (only 5 percent flow through). This likely reflects the fact that industrial producers can import alternative crops in the short-run if there is a year of low production in Paraguay, as crops are an easily traded commodity. Livestock-related industrial production on the other hand is likely slaughtering and butchering, where it is more difficult to replace lost domestic inputs with imported livestock. Given these dynamics, in the year of the drought, industrial output related to agricultural contracts by 4 percent, with industrial output still trailing 3 percent below baseline levels in the following year. On impact, exports are projected to fall by 13.5 percent, while domestic demand falls by 2.9 percent. This is a strong response from exports and is based on the econometric estimation that in the short run, demand drives only 20 percent of movements in Paraguayan exports, with the remaining 80 percent of exports growth being explained by output of the crops, beef and agriculture-related industry sectors (i.e., supply side constraints). Domestic demand has a more muted response to the drought, which reflects a 5.2 percent reduction in real disposable incomes in the year of the drought. The overall impact on GDP is a reduction of 6.9 percent. The fall in real GDP puts downwards pressure on domestic prices (“cost push” prices). Consumer prices increase by just under 1 percent as a result of higher agricultural prices, partially offset by lower cost-push prices. After the drought, there is a period of stronger economic activity, as lower prices stimulate domestic demand, followed by a recovery in the next year with GDP 2.7 percent higher than it would’ve been otherwise. The economy has mostly returned to its baseline activity level five years after the drought. As output and exports contract, fiscal and external accounts deteriorate (Figure 10). Government revenues fall as a direct result of the deterioration in economic activity, although they are a slightly larger share of GDP since the tax base is mostly domestic expenditure (rather than exports). With the assumption of fixed nominal spending in government expenditure categories, this results in a worsening of the fiscal balance by 0.8 percentage points of GDP. This leads to increased government borrowing, which raises the level of government debt. Combined with balance sheet effects due to the depreciation of the currency and with a lower level of GDP in the year of the drought, this increases the debt-to- GDP ratio by almost 4 percentage points. This increase in the debt-GDP ratio leads to an increased risk premium on international debt by almost 50 basis points. As the economy recovers from the drought, a period of stronger economic growth and a return to normal levels of government revenue allows the debt-to-GDP ratio to stabilize at around 0.25 of a percentage point above its baseline level. The debt-to-GDP ratio is permanently higher due to the increased value of USD-denominated debt accrued during the period of the drought. External accounts deteriorate due to the strong fall in exports. Imports also fall in line with lower domestic demand, but by much less than exports (-3.2 percent in the year of the drought). 16 Figure 10: Impulse response functions for selected variables - Fiscal and monetary variables. Note: Percent deviation from baseline shown unless otherwise indicated. 6 Results: The impact of stochastic climate shocks in Paraguay This section presents the results of Monte Carlo simulations which model the economic impacts of stochastic climate shocks under various climate scenarios. These scenarios differentiate between the frequency and intensity of shocks under historical conditions and those anticipated under climate change. By incorporating different GCMs and SSPs, the simulations provide insights into the poten- tial economic consequences of climate-induced disruptions for Paraguay. The analysis highlights how changes in the frequency, severity, and duration of climate shocks (droughts, heatwaves, and floods) could affect key sectors of the economy, including agricultural production, exports, and government fiscal balances. This section discusses the range of economic impacts (fan chart) using the average GCM results for droughts and heatwaves and using SSP2-4.5 results for flooding. Top 10 and bottom 10 GCM results for droughts and heatwaves, and SSP3-7.0 results for flooding can be found in Appendix D. 6.1 Droughts Over the next 25 years, the frequency and intensity of droughts is projected to increase due to climate change. As discussed in the Section 4, the impacts of climate change on crops and livestock are slightly different, so it is worth exploring these separately. As figure 11 shows, the impact of climate change on crops yields is not necessarily negative. While the average estimated impact is a small (0.4 percent) reduction in agricultural output, there may be years where climate change results in overall wetter conditions which are more favorable for crop production (in terms of water availability). On the other hand, the impact of climate change on livestock yields are projected to be only 17 Figure 11: Monte Carlo simulation results for crops (all GCMs). Note: Dark area represent 1 standard deviation (68% confidence interval), light area represents 2 standard deviations (95% confidence interval), solid line represents mean and dashed line represents median negative in all scenarios, with an average projected yield reduction of 0.8 percent. Figure 12: Monte Carlo simulation results for livestock (all GCMs). Note: Dark area represent 1 standard deviation (68% confidence interval), light area represents 2 standard deviations (95% confidence interval), solid line represents mean and dashed line represents median The combined impact of droughts on crops and livestock is a projected average annual reduction in agricultural output by approximately 1.2 percent in Paraguay (Figure 13). This estimate represents the impact over and above the damages from droughts occurring at historical frequency and intensity - that is, the impacts occurring due to climate change. We report the mean or average expected impact, which smooths through year-to-year variability. Under the lower bound of the 95 percent confidence interval, agricultural output could decline by as much as 2.8 percent annually. Conversely, under the upper bound, the impact may be limited to a 0.3 percent reduction in yields. A slightly larger mean impact compared to the median reflects asymmetric downside risk, indicative of a negatively skewed distribution of potential outcomes. The average impact on GDP is initially around -0.17 percent, although this impact reduces over time as structural change occurs in the economy and the agriculture sector’s contribution to GDP shrinks. The reduction in GDP is driven by exports, which are lower by around 0.4 percent on average, due to lower agricultural outputs, and lower industrial production related to agriculture. As output and exports contract, the fiscal and external accounts deteriorate. Government revenues fall as a direct result of the deterioration in economic activity, although they are a slightly larger share of GDP since the tax base is mostly domestic expenditure (rather than exports). With the assumption of fixed nominal spending in government expenditure categories, there is a small deterioration of the fiscal balance, which slowly raises the level of government debt to be 0.3 percentage points of GDP higher by 2050. This in turn leads to higher interest payments and further reduces the government fiscal balance. Should the government temporarily increase social transfers or other recurrent spending to support affected producers, the fiscal deficit would widen further. As outlined in the data section, the top 10 GCMs suggest potential positive impacts on crop 18 Figure 13: Monte Carlo simulation results for droughts (all GCMs). Note: Dark area represent 1 standard deviation (68% confidence interval), light area represents 2 standard deviations (95% confidence interval), solid line represents mean and dashed line represents median Figure 14: Comparison of mean results for drought impacts for the average of all GCMs, bottom 10 GCMs and top 10 GCMs, against a base case of no further climate change. yields, driven by generally wetter climatic conditions and more moderate temperature increases. Con- sequently, the average outcome from Monte Carlo simulations, utilizing exceedance curves derived from these models indicates a positive impact on GDP, with agricultural production projected to be 19 approximately 2 percent higher on average. In contrast, the bottom 10 GCMs forecast a nearly 5 per- cent reduction in agricultural output on average, leading to an initial decrease in real GDP of around 0.8%, with the impact moderating to a -0.4% reduction by 2050 (Figure 14). Under the top 10 GCMs, exports increase by 0.5 percent on average, and the debt-GDP ratio declines by more than 0.5 percentage points. Conversely, under the bottom 10 GCMs, exports reduce by almost 1.2 percent, and the debt-GDP ratio rises to be 1.2 percentage points higher by 2050. Note that these comparisons are looking at the mean outcomes under each set of climate exceedance curves. The lower bound outcomes will show a more extreme difference. 6.2 Heatwaves Over the next 25 years, heatwaves in Paraguay are projected to reduce economic activity by an average of 1.5%, primarily due to declines in labor productivity (Figure 15). The agriculture sector, where a significant proportion of labor occurs outdoors, is expected to experience the most pronounced impact, with agricultural output (attributable solely to labor productivity losses) projected to decrease by 2.6 percent by 2050. The agriculture-related industrial sector will also face reductions in output, driven by both lower labor productivity and diminished inputs from the agriculture sector, resulting in a comparable contraction. In contrast, the services sector is expected to be the least affected, as much of its activity takes place indoors, where heat impacts can be mitigated by shade and air conditioning. Figure 15: Monte Carlo simulation results for heatwaves (all GCMs). Note: Dark area represent 1 standard deviation (68% confidence interval), light area represents 2 standard deviations (95% confidence interval), solid line represents mean and dashed line represents median As output and exports contract, both fiscal and external accounts are expected to deteriorate. Government revenues will decline as a direct consequence of reduced economic activity, though the 20 share of revenues relative to GDP may increase slightly, as the tax base is more reliant on domestic consumption than on exports. Assuming fixed nominal government expenditure, the fiscal balance is projected to worsen, leading to higher levels of government debt. This, in turn, will result in increased interest payments, further exacerbating the fiscal imbalance. By 2050, the fiscal balance is expected to deteriorate by 0.3 percentage points, while the debt-to-GDP ratio is projected to increase by 4 percentage points. Across all GCMs, heatwaves are associated with negative impacts on productivity, resulting in a reduction in economic output. However, the average outcome derived from the top 10 GCM exceedance curves remains negative, albeit less pronounced than the aggregate average across all GCMs (Figure 16). This occurs because labor is exposed to invariably increasing temperatures due to climate change, as opposed to crops which can benefit from potentially higher rainfall. Figure 16: Comparison of mean results for heatwave impacts for the average of all GCMs, bottom 10 GCMs and top 10 GCMs, against a base case of no further climate change. Under the bottom 10 GCMs, the Monte Carlo simulations project an average reduction in output of 2.3 percent by 2050. This contraction is primarily driven by a 4 percent decline in agricultural output and a corresponding 2.5 percent reduction in exports. The resultant decrease in output and exports is anticipated to exacerbate the fiscal balance, leading to an estimated deterioration of 6 percent by 2050. 6.3 Floods Floods, while less frequent than droughts or heatwaves, still pose a risk to economic activity, though their average impact is relatively modest. Figure 17 illustrates the Monte Carlo simulation results for the SSP2-4.5 climate scenario. Over a 25-year horizon, the economic impact of floods becomes more pronounced as government expenditure is redirected from new productive capital investment to repairs. This diversion leads to a reduction in the capital stock, resulting in a 2% decline in real output on average by 2050. In this simulation, the government finances repairs through its existing capital investment budget, avoiding additional borrowing. However, the resulting decrease in output negatively affects both the fiscal balance and the debt-to-GDP ratio, with the latter projected to rise by approximately 1.7 percentage points by 2050. 21 Figure 17: Monte Carlo simulation results for floods under SSP2 climate scenario. Note: Dark area represent 1 standard deviation (68% confidence interval), light area represents 2 standard deviations (95% confidence interval), solid line represents mean and dashed line represents median In the worst-case scenario, represented by the lower bound of the 95 percent confidence interval, the capital stock could be depleted by 10 percent, leading to an output reduction of approximately 5 percent. This could cause the debt-to-GDP ratio to increase by almost 5 percentage points by 2050. There is little variation between the results under the SSP2-4.5 and SSP3-7.0 exceedance curves. This outcome can be attributed to the fact that the primary difference between these scenarios lies in the frequency and severity of floods, which occur infrequently—less than once every 100 years. Given that the focus is on a 25-year time horizon, these rare but very damaging events are infrequent in the Monte Carlo simulations and, as a result, have a limited impact on the overall results. 6.4 Combined impacts Droughts, floods, and heatwaves can occur concurrently or consecutively within the same year, leading to compounding or cascading impacts. While our modeling approach does not fully capture these impacts, we can model the combined impact of these three types of climatic shocks on the Paraguayan economy. To address some of the complexities of multiple simultaneous shocks, our model integrates liquidity constraints by applying a higher risk premium on government debt as the debt-GDP ratio increases. This adjustment results in higher interest payments following consecutive or concurrent shocks, thus compounding the impact on government fiscal balances. Additionally, we account for reconstruction constraints by limiting the amount of reconstruction activity that can occur in a single year, which suggests that economic recovery from consecutive or concurrent shocks will be slower than recovery from a single shock. The Monte Carlo simulation results presented in Figure 18 illustrate the average climate scenario for droughts and heatwaves, along with the SSP2-4.5 climate scenario for floods, providing a snapshot of potential future challenges. Figure 18 presents the Monte Carlo simulation results for the average climate scenario for droughts and heatwaves, along with the SSP2-4.5 climate scenario for floods. Over a 25-year projection period, the cumulative economic impact of these shocks is significant, with annual real GDP expected to be 4% lower by 2050, and the debt-to-GDP ratio projected to 22 Figure 18: Monte Carlo simulation results for combined climate shocks (average climate scenario for droughts and heatwaves, SSP2-4.5 climate scenario for floods). Note: Dark area represent 1 standard deviation (68% confidence interval), light area represents 2 standard deviations (95% confidence interval), solid line represents mean and dashed line represents median increase by 4.5 percentage points. The agriculture sector is particularly vulnerable, facing combined effects from droughts reducing crop and livestock yields, heatwaves decreasing labor productivity, and floods damaging capital stock. As a result, agricultural output is projected to decline by 6% annually by 2050 due to these climatic changes. This is expected to reduce Paraguay’s annual exports by 4 percent by 2050. However, imports also fall due to the depreciation of the exchange rate and lower levels of economic activity, so that the current account balance deteriorates by 1 percentage point of GDP by 2050. A worst-case scenario encompasses the bottom 10 GCMs for droughts and heatwaves, alongside the SSP3-7.0 climate scenario for flooding. This scenario forecasts a reduction in real GDP by 5.4% by 2050, with agricultural output declining by 10.7 percent and exports decreasing by 5.8% (Figure 19). Conversely, the best-case scenario integrates the top 10 GCMs for droughts and heatwaves with the SSP2-4.5 scenario for floods. This scenario predicts a slight initial increase in agricultural production, attributed to more favorable climatic conditions enhancing crop yields. However, these gains are quickly offset by declines in livestock production, labor productivity, and capital stock due to flooding impacts. The aggregate effect on GDP under this optimistic scenario is an annual decline of 2.8%, with the debt-to-GDP ratio projected to increase by 2.7 percentage points by 2050. 23 Figure 19: Comparison of mean results for combined impacts for the average of all GCMs, bottom 10 GCMs and top 10 GCMs, against a base case of no further climate change. (note: for floods, SSP2-4.5 is combined with average and top GCMs, and SSP3-7.0 is combined with bottom GCMs) 7 Limitations of this analysis Although the methodology for simulating stochastic climate shocks offers valuable insight into po- tential future outcomes, it is not without limitations. Understanding these constraints is crucial for interpreting the results and acknowledging the boundaries of the analysis. First, the results are constrained by the design of the MFMod macrostructural model. The func- tional form of equations within the MFMod, along with any imposed coefficient restrictions and identity equation specifications, shapes the response of the model to the shocks. As a result, the analysis is bounded by the assumptions and parameterization of the MFMod, which simplifies certain dynamics of the economy and may not fully capture all real-world complexities, such as sectoral interdependencies and non-linear feedback effects. Second, the analysis relies on modeled exceedance curves, which are based on historical data and GCM projections. These curves, while informative, are inherently limited in their ability to account for extreme or unforeseen climate events or climate change that may lie outside the scope of the biophysical models used. For example, in the crop yield analysis, the modeling does not take into account the implications of deteriorating water quality and increasingly saline soils on future water demands, which could further reduce water reuse practices or impact yields. Similarly, the livestock production modeling does not consider potential reductions in water availability, which can further decrease the productivity of animals, nor the impact of wildfires. Third, the simulations assume a static policy environment and do not incorporate potential future adaptation strategies or technological innovations that could mitigate the economic impacts of climate shocks. For example, the crop yield analysis assumes that the start date and length of the crops’ growing seasons are static, rather than reflecting the fact that farmers tend to adjust growing patterns based on short-term weather forecasts. In Paraguay, for example, the major soy planting season occurs between September and November, but farmers can make adjustments according to weather patterns. Similarly, increased mechanization in the agriculture sector, along with increased air-conditioning coverage may help lessen the impacts on labor productivity. In the livestock sector, how farmers choose to cope with the scarcity of feed (i.e., by moving to feedlots, purchasing alternative feed, or 24 reducing intake per head) is also not taken into account. Fourth, the paper does not consider the interactions between consecutive climate shocks. For example, infrastructure already stressed by drought conditions (such as cracked or weakened roads and bridges) may be less capable of withstanding the sudden stress of flooding. This dual stress can lead to greater damage from a flood following a drought than if each event occurred separately. The model also does not incorporate interactions with non-climate channels, like health, where climate shocks can cause the rise in waterborne diseases, heat-related illnesses, and stress-related health issues, which can have longer-term economic implications. Structural changes due to climate events can lead to population displacements, straining urban infrastructure and shifting labor markets. Moreover, future work could incorporate the variability in energy demand and supply; for instance, heatwaves can lead to spikes in electricity demand for cooling, while droughts can reduce hydroelectric power supply. Hydropower supply and prices are a crucial element of the Paraguayan economy, but they are not captured in the current model given that local electricity tariffs are fixed and do not respond to current conditions. Finally, economic shocks unrelated to climate events, like global financial crises or commodity price shocks, can also compound the effects of climatic shocks, leading to greater economic volatility (see for example Giuliano et al. (2024) for an exploration of this). 8 Conclusion and areas for future research The analysis underscores the importance of designing strategies to cope with climate risks. Currently, Paraguay’s standard response to droughts is to provide special financing facilities and loan moratoria to affected producers, assuming that the situation will revert in the following year and enable repay- ment of loans. The Ministry of Agriculture and Livestock and other agencies also provide temporary assistance to subsistence farmers, reallocating existing public budgets away from investment toward recurrent spending (World Bank (2024b)). However, these reactive strategies would not be sufficient to respond effectively to consecutive droughts or a combination of droughts and floods with other economic shocks that could cause more significant damage. Paraguay could instead adopt a more proactive and comprehensive disaster risk financing strategy which includes a combination of financial instruments that provide layered financial protection. For example, to cover high-frequency but low impact events such as mild droughts, the government could design and implement a national para- metric insurance policy, which pays out based on predefined parameters such as a composite drought index consisting of rainfall levels or temperature indices. For more low-frequency but high-impact events such as a severe drought, it could also establish a reserve fund specifically earmarked for disas- ter response. A third option would be to combine the parametric insurance and reserve fund with a contingent credit line would provide additional liquidity in the event of moderate droughts, floods or other climate shocks. These options are discussed in more detail in Valdivia Zelaya et al. (2023). Publishing a fiscal risk statement that quantifies the risks of droughts and other extreme weather events can significantly aid the authorities in planning for their unforeseen impacts. A fiscal risk state- ment provides a transparent and detailed account of the potential economic and fiscal risks associated with climate and other external shocks. This transparency helps in building trust with stakeholders, in- cluding investors, development partners, and the general public. Moreover, it enables the government to make informed decisions regarding emergency budget allocations for repairs or temporary social assistance, contingency planning, and the sequencing and prioritization of investments in climate re- silience. The above analysis indicates, for example, that adaptation investments are especially needed for livestock producers and workers exposed to heatwaves. By proactively identifying and quantifying the risks through a fiscal risk statement, the Ministry of Economy and Finance could lead the process 25 of developing a more comprehensive, cross-sectoral strategy for responding to droughts, heatwaves and floods. More broadly, the analysis shows the importance of structural reforms that would help Paraguay diversify its exports away from raw agriculture and energy commodities. The country has already started to make important strides in this regard, with exports of light manufacturing goods such as textiles and automotive parts showing important growth in recent years, especially to neighboring Brazil, but these still account for only about 6 percent of total exports. Addressing shortcomings in human capital and infrastructure is critical if Paraguay is to modernize its agribusiness sector beyond capital-intensive crops such as soybeans, and to encourage the expansion of other manufacturing and services sectors that are less exposed to climate shocks. 26 References Abalo, K., Boehlert, B., Bui, T., Burns, A., Castillo, D., Chewpreecha, U., Haider, A., Hallegatte, S., Jooste, C., McIsaac, F., Ruberl, H., Smet, K. and Strzepek, K. (2025). The macroeconomic impli- cations of climate change impacts and adaptation options: A modeling approach, Policy Research Working Paper 11133, World Bank, Washington, D.C. URL: http://hdl.handle.net/10986/43258 Banerjee, O., Cicowiez, M., Rios, A. R. and Lima, C. Z. d. (2021). Climate change impacts on agricul- ture in latin america and the caribbean: An application of the integrated economic-environmental modeling (ieem) platform, Technical report, Inter-American Development Bank (IDB). Accessed: 2025-02-24. URL: https://publications.iadb.org/en/climate-change-impacts-agriculture-latin-america-and- caribbean-application-integrated-economic Benitez, J. and Domecq, R. (2014). Analysis of meteorological drought episodes in paraguay, Climatic Change 127: 15–25. URL: https://doi.org/10.1007/s10584-014-1260-7 Benitez Rodriguez, A. M., Wolf, R., Trotter, I. M. and Gurgel, A. C. (2024). Assessing the economic implications of climate change impacts on the paraguayan agricultural sector, Climate and Devel- opment 16(1): 1–8. URL: https://doi.org/10.1080/17565529.2023.2176186 Bravo, J., Collischonn, W., da Paz, A. et al. (2014). Impact of projected climate change on hydrologic regime of the upper paraguay river basin, Climatic Change 127(1): 27–41. URL: https://doi.org/10.1007/s10584-013-0816-2 Burns, A., Campagne, B. P. M., Jooste, C., Stephan, D. A. and Bui, T. T. (2019). The world bank macro-fiscal model technical description, Policy Research Working Paper WPS 8965, World Bank Group, Washington, D.C. URL: http://documents.worldbank.org/curated/en/294311565103938951/The-World-Bank-Macro- Fiscal-Model-Technical-Description Burns, A., Jooste, C. and Schwerhoff, G. (2021). Climate modeling for macroeconomic policy: A case study for pakistan, Policy Research Working Paper WPS 9780, World Bank Group, Washington, D.C. URL: http://documents.worldbank.org/curated/en/747101632403308927/Climate-Modeling-for- Macroeconomic-Policy-A-Case-Study-for-Pakistan aez, J., Biedermann, G. and D´ B´ ıaz, V. R. (2023). Drought shocks and the paraguayan macroeconomy: A bayesian svar analysis, Technical report, Central Bank of Paraguay, Asunci´on, Paraguay. Banco Central del Paraguay - Economic Studies. URL: https://www.bcp.gov.py/en/web/instituto-bcp/w/drought-shocks-and-the-paraguayan- macroeconomy-a-bayesian-svar-analysis CEPAL (2014). La econom´ atico en el Paraguay, Colecci´ ıa del cambio clim´ on Documentos de Proyecto, CEPAL, Santiago, Chile. URL: https://repositorio.cepal.org/handle/11362/37326 27 Ervin, P. A. and Gayoso de Ervin, L. (2019). Household vulnerability to food insecurity in the face of climate change in paraguay, Working Paper 19-04, FAO. URL: https://ageconsearch.umn.edu/record/288947/files/CA3391EN.pdf for Greening Financial Systems, N. (2024). Climate macroeconomic modeling handbook, Technical report, Network for Greening Financial Systems. URL: https://www.ngfs.net/system/files/import/ngfs/medias/documents/ngfsc limate − macroeconomic − modelling − handbook.pdf Giuliano, F. M., Navia Simon, D. and Ruberl, H. J. (2024). The macroeconomic impact of climate shocks in uruguay, Policy Research Working Paper WPS 10740, World Bank Group, Washington, D.C. URL: http://documents.worldbank.org/curated/en/099400003272435890/IDU13bfbac77199f81489e18f2319846d83a3953 Hallegatte, S. and Vogt-Schilb, A. (2016). Are losses from natural disasters more than just asset losses? the role of capital aggregation, sector interactions, and investment behaviors, Policy Research Working Paper 7885, The World Bank. URL: https://doi.org/10.1596/1813-9450-7885 Janz, T., Gassmann, F. and de Ervin, L. G. (2024). Weathering shocks:unraveling the effects of short- term weather shocks on poverty in paraguay, Working Paper WPS10970, World Bank. URL: http://documents.worldbank.org/curated/en/099331411062421620 MEF (2024). Informe de las finanzas p´ ıa y ublicas 2024, Technical report, Ministerio de Econom´ Finanzas. Accessed: 2025-02-27. URL: https://www.mef.gov.py/es/dependencias/viceministerio-de-economia-y- planificacion/informes-las-finanzas-publicas Myhre, G., Alterskjær, K., Stjern, C. W. and et al. (2019). Frequency of extreme precipitation increases extensively with event rareness under global warming, Scientific Reports 9: 16063. URL: https://doi.org/10.1038/s41598-019-52277-4 United Nations Office for Disaster Risk Reduction (UNDRR) (2015). Global assessment report on disaster risk reduction 2015. Accessed: 2023-10-17. URL: http://www.undrr.org/quick/11404 USDA (2024). Paraguay: Livestock and products annual, https://www.fas.usda.gov/data/ paraguay-livestock-and-products-annual-9. Accessed: 2025-03-10. ıa t´ Valdivia Zelaya, P., Bresnyan, E., Villalobos, J. and et al. (2023). Asesor´ ecnica no reembolsable: omo hacer frente a los riesgos hidrometeorol´ ¿c´ ogicos que afectan a la agricultura familiar en paraguay?, Technical report, World Bank. Accessed: 2025-03-10. URL: https://documents1.worldbank.org/curated/en/099062323115086993/pdf/P179114092115e0008e020dfdd82368b63 World Bank (2021). Climate risk profile for paraguay, Technical report, World Bank. Available online:https://climateknowledgeportal.worldbank.org/sites/default/files/country-profiles/15726- WBP araguay World Bank (2024a). The changing wealth of nations 2024 : Revisiting the mea- surement of comprehensive wealth, Technical report, World Bank. Available online:https://documents1.worldbank.org/curated/en/099100824155021548/pdf/P178446-7dfe6e92- 9cf0-4d24-ad25-20b1320904c2.pdf. 28 World Bank (2024b). From landlocked to land of opportunity: Paraguay country economic memoran- dum, Technical report, World Bank, Washington, DC. URL: https://documents1.worldbank.org/curated/en/099062524172518884/pdf/P17810818bf3c40bf1b2ab145d817d52d03.p World Bank (2025). Paraguay country climate and development report, Technical report, World Bank, Washington, DC. URL: https://openknowledge.worldbank.org/entities/publication/f4ffcb9a-a0e4-492a-ab8e- 17e86198ab7e World Bank and OLADE (2023). Evaluaci´ atico en la generaci´ on del impacto del cambio clim´ ectrica on el´ ıses del cono sur, Technical report, World Bank and OLADE. Accessed: 2025-02-27. en los pa´ URL: https://documents1.worldbank.org/curated/en/099706209062333677/pdf/IDU02f576bf50b31a040ac0a59e0e05e1eb3c 29 A Appendix A: Technical description of biophysical models to estimate the impact of droughts on crop and livestock yields and heatwaves on labor productivity A.1 Modeling climate impacts on rainfed crop yields Water availability impacts: To estimate the impacts of changing water availability on crop yields, we apply the methods documented in the FAO’s Irrigation and Drainage Papers 33 and 56 (Allen et al. 1998; Doorenbos and Kassam 1979). Water demand is calculated from potential evapotranspiration, calculated monthly following the modified Hargreaves approach (Droogers and Allen 2001), and crop- specific water demand coefficients (Kc), resulting in evapotranspiration requirements (ETc). Rainfed water supply is estimated as effective precipitation, and its ratio to water needs is then used to estimate a crop-specific annual yield response. Actual evapotranspiration is reduced based on available rainfall, which allows for the calculation of actual yields as a deficit from non-water-constrained yields. Heat stress impacts: Yield impacts from temperature are calculated based on daily maximum temperatures and tolerance thresholds following the methods applied in FAO’s AquaCrop (Salman et al. 2021). We gather optimum and maximum tolerance temperatures by crop from the FAO’s ECOCROP database (2015), which determines at which temperature a crop will start experiencing damage until it suffers full loss. We utilize the average 3-day maximum temperature, as effects are typically experienced after 3 consecutive days of exposure (Wahid et al. 2007; Nuttall et al. 2018; Hatfield and Prueger 2015; Gourdji et al. 2013). Temperature yield responses are estimated based on a logistic relationship between temperature and maximum attainable yields. The yield impacts depend on a crop’s optimum temperature above which yields start decreasing and the maximum daily temperature. A.2 Modeling climate impacts on livestock yields Heat stress on milk and egg yields: The direct effects of heat stress on animals are estimated using animal-specific impact functions for milk and eggs that use a temperature-humidity index (THI) as an indicator of heat stress, with productivity losses based on heat tolerance thresholds. THI measures the perceived temperature by the animal based on air temperature, wet-bulb temperature, and rela- tive humidity. Species-specific THI equations are available from Rahimi et al. (2021). THI is then translated into daily impacts on milk and egg production as either kilograms or liters per head, as it surpasses a threshold set to 28C, based on impact functions for cattle milk (Mauger et al. 2015), goat milk (Salama et al. 2014), and chicken egg (St-Pierre et al. 2003). Pasture availability impacts on meat yields: Regional feed source proportions from FAO GLEAM are applied to obtain the fraction of the herd that relies on grazing pastures. To estimate grazing pas- ture yields, we follow the approach from the FAO’s Irrigation and Drainage Paper 33, Yield Response to Water (Doorenbos and Kassam 1979) and AquaCrop’s heat stress model (Salman et al. 2021), as explained in the Rainfed Crops Yields section. Pastures are assumed to be grazed at their maximum carrying capacity on average in the reference period, with room for additional productivity gains in wetter years of up to 20 percent. All decreases in pasture production are assumed to result in a lack of feed and dry matter intake. Animal live body weight can increase from higher intakes of dry matter or from higher animal feed efficiency, which can be attributed to differences across breeds and feed sources (Lima et al. 2017; Mackle et al. 1996; Wen et al. 2018). We assume constant efficiency values and assess changes in live body weight only from estimated changes in pastures. Reductions in feed are then converted into shocks to dry matter intake and, in turn, translated into meat production losses. 30 A.3 Modeling climate impacts on labor productivity Heat stress estimates: Temperature directly affects the productivity of labor, where the effect intensifies for labor types that are outdoors and are conducting more intense physical work. Labor productivity impacts are estimated following the methodology applied by the International Labor Organization (ILO 2019), which has been applied globally and used in other studies, such as in Kjellstrom et al. (2018). The functional relationship between work ability and heat stress is quantified using wet bulb globe temperatures (measured in degrees Celsius). For outdoor conditions, we use the formulation by Moran et al. (2003) plus a bias correction to compensate for underestimation under high solar radia- tion conditions based on Liljegren et al. (2008) and Kong and Huber (2022). For indoor conditions, we utilize the Kjellstrom et al. (2018) application of the Bernard and Pourmoghani (1999) formula- tion. Both equations require air temperature projections and assume constant relative humidity and solar radiation. We assume that in a typical workday, one-third of hours are close to the minimum, maximum, and midpoint daily temperatures. Productivity effects: Labor productivity effects are estimated based on the percentage of an hour that an acclimatized worker can be engaged in work using work ability curves from ISO and validated through epidemiological studies, as presented in ILO (2019). Occupations with lower physical activity can tolerate higher levels of heat stress than occupations with higher physical activity. For each economic sector, the productivity shocks depend on the sector’s outdoor exposure and physical activity requirements. Employment by occupation and sector is obtained from the 2021 National Labor Force Survey reported in the International Labor Organization’s Labor Force Statistics. Outdoor exposure by occupation is approximated from the Occupational Requirements Survey from the U.S. Bureau of Labor Statistics (BLS 2022). For indoor workers, we exclude the fraction with air conditioning and assume they do not experience any heat stress. B Appendix B: Technical description of MfMOD model changes to incorporate climate change dynamics This technical appendix will describe the modeling of supply and changes to some demand-side equa- tions to better capture the impact of droughts in Paraguay. This is an expansion on both the standard MFMod approach, described in Burns et al. (2019) and the standard climate change version of the model described in Burns et al. (2021). Model changes to capture heatwaves and floods are simpler, and can be found in Section 4. As described in the main paper, there are six sectors on the supply side of the model: crops, live- stock, industrial production (agriculture related), electricity production, other industries, and services. In the standard version of the model, each sector is derived from final demand and connected using IO tables. In the standard climate change extension of MFMod, inter-sectoral change is not adequately ac- counted for. For example, a drought shock would impact on production in the whole economy, but not specifically on the agriculture sectors, or those sectors which use agricultural products as inputs to production (such as industry). It would also impact on aggregate factor prices, but not sectoral prices. These sector specific impacts are important to capture for Paraguay, and the modeling approach to do this is described below. This methodology builds on the model extension for Uruguay described in Giuliano et al. (2024). 31 B.1 Model expansion to capture droughts Sectoral Y* and P* The modeling connects livestock and crop yields to climate shocks and then maps them to economic activity. To account for different productivity shocks in each sector, we construct a production function for each sector, which is modified to account for losses due to climate shocks. The agricultural damage or losses due to the drought di,t is incorporated into the model as reduced TFP, because the same amount of capital and labor produces a reduced amount of output compared to previous years. ∗ 1−α Yi,t = (1 − di,t )At Nt∗ αKt −1 (8) ∗ Similarly, the marginal costs for each sector (Pi,t ) is impacted by sectoral specific productivity shocks through the same damage function: 1−α ∗ Wtα Ct Pi,t = 1−α 1 α (9) 1 (1 − di,t )At 1− α α This price reflects aggregate factor prices (Wt and Rt ), aggregate productivity (At ), and sector- specific productivity shocks (di t ). Together, these reflect the marginal costs of the firm. Sectoral Volumes Output of the agriculture sectors is modeled as a function of their respective production functions. Damage or losses are allowed to have an autoregressive component, calibrated to the persistence of shocks on output. In the crop sector (YCRP,t ), the auto-regressive component of output (β ) is estimated to be quite small (0.05), and the speed of adjustment (θ) is quite fast (0.84). ∗ ∗ ∆ ln YCRP,t = −θ · ln YCRP,t−1 − ln YCRP,t−1 + β · ∆ ln YCRP,t−1 + (1 − β ) · ∆ ln YCRP,t (10) Livestock output (YLV S,t ) is econometrically estimated to be slower to recover from a drought, with a larger autoregressive component (0.15) and a slower speed of adjustment (0.57) compared to the crop sector. ∗ ∗ ∆ ln YLV S,t = −θ · ln YLV S,t−1 − ln YLV S,t−1 + β · ∆ ln YLV S,t−1 + (1 − β ) · ∆ ln YLV S,t (11) Industrial production (agriculture-related) is modeled with the following adjustments. In the long run, output is a function of potential output and crop output. In the short run, output is a function of potential output, crop production, and livestock production. ∗ ∆ ln YIAG,t = −θ · ln YIAG,t−1 − ω · ln YIAG,t−1 + (1 − ω ) · ln YCRP,t ∗ PtLV S PtCRP + β1 · ∆ ln YLV S,t + β2 · ∆ ln YCRP,t + (1 − β1 − β2 ) · ∆ ln YIAG,t + δ1 · + δ 2 · (12) PtCP SH PtCP SH In the long run, crop production explains 20% of industrial production related to agriculture. In the short run, growth is estimated to be 19% driven by livestock output, 3% driven by crop output, with the remainder explained by potential output. In the short run, production is also impacted by price elasticities, which capture the cost of 32 intermediate inputs (crops and livestock) relative to the marginal cost of other factors of production. These are both estimated to be positive and statistically significant, i.e., an increase in the price of crops or livestock is associated with a decrease in industrial production related to agriculture. Sectoral Prices Livestock and crop prices (Pi,t ) are strongly driven by global prices for these commodities. For each sector, we construct a global price index (PtW ). For crops, this is a weighted average of the prices for maize, soybean meal, soybean oil, soybeans, and wheat, while for livestock, this is just the global price of beef. Since these prices are in USD, they must be adjusted for the exchange rate (exrt ), where exrt is Paraguayan guarani per USD. The domestic price of crop and livestock production follows an ECM functional form where it is ∗ driven by global prices adjusted for the exchange rate, and factor costs (Pi,t ): ∗ ∆ ln Pi,t = −θ · ln Pi,t−1 − ln PtW W −1 · exrt−1 + β · ∆ ln Pi,t + (1 − β ) · ∆ ln Pt · exrt (13) For crops, factor costs contributed to one-third of price fluctuations in the short run (β = 0.34), while for livestock, they contributed much more (β = 0.89). Export volumes Exports are of particular interest given the importance of agricultural exports to Paraguay’s econ- omy. The standard export demand equation is amended to directly incorporate changes in agricultural exports. ∗ Equilibrium exports (Xt ) are adjusted so that the non-agricultural component of exports (for M KT example, tourism, business services, manufacturing) is a function of export market demand (Xt ), but crop and livestock goods exports are a function of the output of those industries rather than demand: ∗ M KT Xt = (1 − αt ) · Xt + αt · YtLV S + YtLV S + YtIN DAGRI (14) Where αt is the econometrically estimated contribution of crop, livestock and agriculture related industrial production contribution to export volumes. In the long run, αt is estimated to be 0.52, and in the short run it is 0.79, demonstrating the strong importance of agricultural output to exports. Export volumes are estimated as error-correcting to their equilibrium level, with price elasticity accounted for in both the short and long run: ∗ PtX −1 ∗ PtX ∆ ln(Xt ) = θ ln(Xt−1 ) − ln(Xt−1 ) − β1 · ln + ∆ ln(Xt ) − β2 · ∆ ln + ϵX t (15) PtCON −1 PtW Exchange rate It is useful for the modeling of the exchange rate to reflect economic fundamentals. In the case of Paraguay, exports constitute a significant proportion of the economy, and therefore, changes in export volumes are pertinent to exchange rate behavior. The change in the export share of GDP ( X Yt ) is t incorporated into the standard MFMod exchange rate equation: iU t S Xt ln(exrt ) = β1 + ln(exrt−1 ) + ln + β2 · ∆ + ϵexr t (16) it Yt Where iUt S is the implicit US interest rate on government debt, and it is the domestic monetary policy rate. There is a negative and statistically significant estimate for β2 = 0.46, indicating that as exports fall as a share of GDP (as they would during a drought), the Paraguayan exchange rate depreciates 33 (an increase in the exchange rate in terms of Paraguayan guarani per USD is a depreciation). Government risk premium on debt The Paraguayan government pays a risk premium on debt borrowed in international markets EXT L (riskt ), which represents the differential between the effective interest rate on this debt (iEXT t L ) and the USA effective interest rate on government debt. iEXT t L EXT L = riskt + iU t S (17) This external risk premium is modeled as having a positive relationship with the government’s overall debt-GDP ratio: EXT L Dt riskt = β1 + β2 ( ∗ 100) (18) Yt We estimate an intercept of -3.6, with an increase in the risk premium of 0.11 for every 1 percentage point increase in the debt-GDP ratio. C Appendix C: Monte Carlo simulation of stochastic climate shocks To model climate events such as droughts, heatwaves, flooding, tropical cyclones or extreme rainfall, we use risk or exceedance curves in conjunction with Monte Carlo simulations to generate stochastic or random events. Exceedance or risk curves link the probability of extreme events occurring to the level of severity of the event. In the context of extreme rainfall or flooding, an exceedance curve shows the relationship between the frequency of the event occurring and the scale of damage cause (in % of capital stock destroyed). For instance, an exceedance probability of P = 0.01 (1%) corresponds to a 1-in-100-year flood event. This implies a 1% chance of such a flood occurring in any given year. Monte Carlo simulations rely on random sampling to compute results. In this case, we use them to model the occurrence and intensity of extreme climate events based on the exceedance curve. 1. Sampling from a Uniform Distribution: Monte Carlo simulations begin by generating random numbers, typically between 0 and 1 (de- noted as u ∈ [0, 1]). These random numbers are sampled from a uniform distribution and represent probabilities. 2. Linking Random Numbers to Event Probabilities: The random number u corresponds to a probability that an event of a certain severity will occur. Since the random number is between 0 and 1, it is interpreted as a cumulative probability. Using the inverse of the exceedance probability function (derived from the exceedance curve), we map this random number to an event magnitude. That is, for a given probability u, the corresponding magnitude of the event X (i.e., the damage to the capital stock from a rainfall or flooding event) is determined. 3. Generating the Event Magnitude: The exceedance curve is used in reverse to map from the random probability u to a specific event magnitude X . This is done using the inverse exceedance function: X = F −1 (u) 34 where F −1 is the inverse of the cumulative distribution derived from the exceedance curve. For example, if u = 0.05, this represents a 5% chance of a certain extreme event occurring. The corresponding event magnitude (from the exceedance curve for flooding) is 0.1% damage to the capital stock. 4. Repeating the Process: The Monte Carlo simulation repeats this process for 1000 iterations, each time generating a new random number and calculating the corresponding event magnitude. This produces a distribution of possible outcomes for the climate event. Monte Carlo simulations can be extended to simulate stochastic shocks over multiple time periods (e.g., yearly or monthly). For each period, a new random number u is drawn, which determines whether an extreme climate event occurs and its severity (Figure 20, left panel). Figure 20: Example Monte Carlo simulation. Left panel: random number draw 89. Right panel: correspond- ing crop damages (reduction in yield) from drought using historical exceedance curve Once the stochastic shocks (event magnitudes) are generated, the time series for the damages under each iteration (Figure 20, right panel) are incorporated into the MFMod model to estimate potential economic losses. After conducting 1,000 iterations for a given climate shock, we aggregate and analyze the resulting data to characterize the distribution of potential outcomes. Once 1,000 simulated pathways for a specific climate-related event have been incorporated into the model, we extract the corresponding 1,000 pathways for affected variables like agricultural output, exports, GDP, and fiscal balance. Figure 21, left panel, shows the impact of droughts from draw 89 on Real GDP in the model. Figure 21: Example Monte Carlo simulation. Left panel: GDP pathway for draw 89. Right panel: fan chart showing median and mean results with 1 and 2 confidence intervals 35 By collecting these outcomes, we perform statistical analyses to derive key percentiles and assess the variability of results. Specifically, we compute the median outcome over time, representing the 50th percentile of the results distribution. To capture the dispersion of outcomes, we determine the 68 percent confidence interval by identifying the 16th and 84th percentiles, corresponding to one standard deviation from the median. For a broader assessment, we calculate the 95 percent confidence interval by identifying the 2.5th and 97.5th percentiles, representing two standard deviations from the median. Figure 21, right panel, shows the range of outcomes for Real GDP given droughts occurring at historical frequencies and intensities. Historical events are already implicitly incorporated into historical data, including real GDP figures and estimates of productivity and potential output. These events are therefore embedded in our baseline projections, which serve as a reference point for assessing the potential impacts of climate change. Our analysis focuses on two key aspects: 1. The variability in economic projections due to climate change, and the associated variability of these outcomes, as illustrated by the confidence intervals in Figure 21, right panel. 2. The impact of different climate change scenarios on median economic outcomes, as illustrated in Figure 22. To quantify these impacts, we compare how the median results for key variables, such as real GDP, shift under different climate change scenarios. Specifically, we subtract the median baseline outcome for a variable of interest (projected using historical exceedance curves) from the median outcome under a given climate scenario. This difference represents the effect of climate change on the central tendency of economic projections, as shown in Figure 22. Figure 22: Change in median outcomes under different climate scenarios This approach allows us to clearly separate the expected impact of climate change from historical trends while providing a robust framework for evaluating the variability and uncertainty in economic outcomes. 36 D Appendix D: Additional results Figure 23: Monte Carlo simulation results for droughts (top 10 GCMs). Note: Dark area represent 1 standard deviation (68% confidence interval), light area represents 2 standard deviations (95% confidence interval), solid line represents mean and dashed line represents median 37 Figure 24: Monte Carlo simulation results for droughts (bottom 10 GCMs). Note: Dark area represent 1 standard deviation (68% confidence interval), light area represents 2 standard deviations (95% confidence interval), solid line represents mean and dashed line represents median Figure 25: Monte Carlo simulation results for heatwaves (top 10 GCMs). Note: Dark area represent 1 standard deviation (68% confidence interval), light area represents 2 standard deviations (95% confidence interval), solid line represents mean and dashed line represents median 38 Figure 26: Monte Carlo simulation results for heatwaves (bottom 10 GCMs). Note: Dark area represent 1 standard deviation (68% confidence interval), light area represents 2 standard deviations (95% confidence interval), solid line represents mean and dashed line represents median Figure 27: Monte Carlo simulation results for floods (SSP 3). Note: Dark area represent 1 standard deviation (68% confidence interval), light area represents 2 standard deviations (95% confidence interval), solid line represents mean and dashed line represents median 39 Figure 28: Monte Carlo simulation results for combined climate shocks (Top 10 GCMs climate scenario for droughts and heatwaves, SSP2-4.5 climate scenario for floods). Note: Dark area represent 1 standard deviation (68% confidence interval), light area represents 2 standard deviations (95% confidence interval), solid line represents mean and dashed line represents median Figure 29: Monte Carlo simulation results for combined climate shocks (Bottom 10 GCMs climate scenario for droughts and heatwaves, SSP3-7.0 climate scenario for floods). Note: Dark area represent 1 standard deviation (68% confidence interval), light area represents 2 standard deviations (95% confidence interval), solid line represents mean and dashed line represents median 40