FINANCE FINANCE EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT From extreme events to extreme seasons: Financial stability risks of climate change in Mexico Authors: Michaela Dolk, Dimitrios Laliotis, Sujan Lamichhane © 2023 International Monetary Fund and The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org Disclaimer This work is a product of the staff of The World Bank and the IMF with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent, or the views of the IMF, its Executive Board, or IMF management. The World Bank does not guarantee the accuracy, completeness, or currency of the data included in this work and does not assume responsibility for any errors, omissions, or discrepancies in the information, or liability with respect to the use of or failure to use the information, methods, processes, or conclusions set forth. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be construed or considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org >>> Contents Acknowledgments 4 Abstract 5 Brief Summary 6 I. Introduction 7 II. Climate-related disasters in Mexico 11 III. Scenarios 14 IV. Methods 17 IV. I. Direct damages model 18 IV. I. I. Exposure module 19 IV. I. II. Hazard module 19 IV. I. III. Vulnerability module 21 IV. II. Macroeconomic impact 21 IV. III. Financial sector impact model 22 V. Results 24 VI. Discussion and Conclusion 28 VI. I. Uncertainties, limitations, and interpretation 28 VI. II. Moving from individual events to sequences of events 29 Appendix I. Climate-related disasters in Mexico 30 Appendix II. Description of individual event scenarios 32 References 34 >>> Acknowledgments This report was prepared by Michaela Dolk (Crisis and Disaster Risk Finance Unit, The World Bank), Dimitrios Laliotis (International Monetary Fund), and Sujan Lamichhane (International Monetary Fund). This work developed from joint IMF and World Bank work on climate risk analysis for the 2022 Mexico FSAP. We would like to thank Vikram Haksar (IMF) and Ilias Skamnelos (WB) (FSAP mission chiefs) and FSAP team members for their helpful comments and suggestions. We are also grateful to the members of Mexican Financial Authorities for a highly productive engagement, discussions, and feedback. We wish to thank colleagues for their valuable review comments and insights, including Mark Roland Thomas, Jean Pesme, Nepomuk Dunz, Faruk Liriano, Florent McIsaac, Mariza Montes de Oca Leon, Sneha Thube, and Charlotte Gardes-Landolfini. We would like to acknowledge Dorra Berraies and Cristina Stefan for their analytical support. >>> Abstract This paper explores the financial stability implications of acute physical climate change risks using a novel approach focusing on a severe season associated with a series of tropical cyclone and flood events. Our approach was recently applied to study physical risks in the Mexican financial sector, but the framework also pertains to other countries. We show that even if the scale of individual climate events may not be material at an aggregate national scale, considering a sequence of events could lead to potentially significant macro-financial impacts in the short term. This could occur even if none of the individual events affect the particular region(s) with the highest concentrations of banking sector exposures. Our results indicate the potential for even greater effects in the future, given the increasing severity and frequency of extreme events from climate change. Thus, this paper highlights the importance of considering series of extreme physical risk events driven by climate change, rather than just individual extreme events, to better understand financial stability implications and design effective policies. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 5 >>> Brief Summary Extreme events do not necessarily occur in isolation – rather, it is possible for a series of events to occur within a relatively short timeframe (e.g., over the course of a season). This may be particularly relevant when considering the financial stability implications of acute physical climate change risks. Whereas impacts of localized individual extreme events may appear relatively benign at an aggregate national scale (particularly for large countries with geographically diversified exposures), a series of extreme events affecting multiple different parts of a country could result in more material impacts to the financial system. This paper outlines a novel approach to consider the potential implications of such series of events for financial stability analyses of acute physical climate change risks. An extreme but plausible series of extreme events is defined, and subsequently analyzed using a model of direct damages, combined with macroeconomic and financial sector impact modeling. The approach builds upon recent methodologies developed jointly by the IMF and the World Bank in the Financial Sector Assessment Program (FSAP) context. The approach was applied to study physical risks in the Mexican financial sector, with the analysis considering a sequence of tropical cyclone and flood events occurring during an extreme season. A sequence of geographically non-overlapping events was defined and analyzed under both a current climate conditions scenario and a potential future climate change conditions scenario to explore the potential impacts of climate change. We show that even if the scale of individual climate events may not be material at an aggregate national scale, considering a series of events could lead to potentially significant macro-financial impacts in the short term. This could occur even if none of the individual events directly affect the region(s) with the highest concentrations of banking sector exposures (e.g., Mexico City). Our results indicate potential for even greater effects in the future, given the increasing severity and frequency of extreme events from climate change. This paper highlights the importance of considering a series of extreme events, rather than just individual extreme events, to better understand the financial stability implications of acute physical climate change risks and design effective policies. While the approach was applied to study physical risks in the Mexican financial sector, the framework is potentially applicable to other countries as well. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 6 1 >>> I. Introduction The geographical diversification of firms’ exposures (relative to the often more localized nature of disasters) is cited as a key factor for only moderate/ contained systemwide financial stability impacts of acute physical climate risks1 in several preliminary climate risk assessments (NGFS, 2022). It may be reasonable for large countries to expect relatively mild national-level impacts of geographically concentrated individual climate-related extreme events (since an individual event would only impact a subset of total exposures). This is particularly true if exposures are not geographically concentrated in a single location subject to such climate-related extreme events. Yet it is important to recognize that extreme events do not always occur in isolation. Rather, it is possible for a series of extreme events affecting multiple different parts of a country (or indeed the globe) to occur within a relatively short time window (e.g., over the course of a season).2 The concept of series or clusters of extreme events is not new: there is an increasingly robust body of scientific literature on the topic, and it is also already embedded in many of the catastrophe modeling approaches used by the (re) insurance industry. Sequences and clusters of extreme events have been studied within the growing field of compound weather and climate events science and statistics (Zscheischler et al., 2018; Towe et al. 2020; Bevacqua et al., 2021). Compound events are “the combination of multiple drivers and/or hazards that contributes to societal or environmental risk” (Zscheischler et al., 2018). The interaction of such shocks could amplify impacts (with the overall impacts being greater than the sum of the parts; Kopp et al., 2017), potentially exceeding the coping capacity of firms, households, the government, and banks, who might have been able to cope with those events if they had occurred separately. The IPCC Sixth Assessment Report included a discussion of compound events, concluding that there is high confidence that “concurrent extreme events at different locations, but possibly affecting similar sectors in different regions, will become more frequent with increasing global warming, in particular above 2°C of global warming” (IPCC, 2021). Indeed, some regions already have evidence of increasing sequential tropical cyclone hazards (Xi et al., 2023). 1 - Physical risks from climate change are typically categorized into acute physical risks (associated with increased frequency and severity of extreme events – e.g., tropical cyclones, floods, and droughts) and chronic physical risks (associated with gradual changes – e.g., rising sea levels). Acute physical risks associated with tropical cyclones and floods are the focus of this paper, while noting that other acute physical risks and chronic risks may also be substantial for Mexico. 2 - For example, within an individual tropical cyclone season, a country might be subjected to multiple tropical cyclones, such as those experienced by Mexico in 2010 (Hurricane Alex, Hurricane Karl and Tropical Storm Matthew). FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 7 In the (re)insurance industry, clusters of events are floods to the same extent as other parts of the country). captured in catastrophe models that are built using Outside of Mexico City, exposures are distributed stochastic event sets with multiple events each amongst several regions that are not geographically year. Such models enable the estimation of not only proximate due to the large size of the country. These occurrence exceedance probabilities (which look less geographically concentrated exposures are less at probabilities of individual event losses – e.g., a likely to be impacted by a single tropical cyclone or 1-in-100-year flood event loss) but also aggregate flood event due to the comparatively localized nature exceedance probabilities (which consider total of these perils (for example, relative to perils such as aggregated losses from multiple events in a given drought, which can impact the entire country in a single year – e.g., a 1-in-100-year aggregate loss) (Homer event – see Stahle et al., 2016). and Ming, 2017). These metrics are important from a portfolio management perspective, and (re)insurers The modeling approach used in our analysis builds with portfolios across multiple regions are increasingly upon recent methodologies developed jointly by looking to understand how global-scale correlations the IMF and the World Bank in the context of the of extreme events can impact their risk diversification Financial Sector Assessment Program (FSAP). strategies (Carozza and Boudreault, 2021). The A set of disaster scenarios are analyzed for current annual exceedance probability metric has also been climate conditions and potential future conditions used for analyzing general insurance liabilities in the considering climate change. The risks from these Bank of England’s 2021 Biennial Exploratory Scenario scenarios are analyzed using a combination of a for testing resilience to climate change risks (Bank of model of direct damages (based on catastrophe risk England, 2021). modeling approaches), a macroeconomic model for generating macro-financial scenarios, and a In the context of physical climate risk assessments, financial sector impact model to analyze the impact the implications of spatial or temporal clusters of on the financial sector (based on the standard IMF events could be substantial. Failure to account for solvency stress testing method used in FSAPs). the potential for series of events, and to recognize The damage estimates from the extreme season that individual extreme events are not spatially and/ scenarios were used as input in generating scenario- or temporally independent from one another, could dependent macro-financial paths. These scenarios are lead to an underestimation of the materiality of further translated into impact on the financial sector. acute physical risks (Zscheischler et al., 2018). For Additionally, to account for other indirect effects that individual financial institutions, more widespread or could compound the effects of initial direct damages, correlated extreme weather events under climate we also considered three additional channels via change could undermine the effectiveness of banks’ shocks to total factor productivity (TFP), higher diversification strategies (BIS, 2021; ECB, 2021). unemployment, and effects from financial markets. A The implications for a domestic financial system as a similar approach was also used to model typhoon risks’ whole could be substantial if a series of shocks affects impacts on the Philippines (Hallegatte et al., 2022a). many institutions and/or jurisdictions, creating longer- The novel aspect of the Mexico analysis is to extend lasting impacts and triggering negative feedback loops this approach to consider a sequence of disaster between bank lending and the real economy (FSB, events occurring during a season rather than looking 2020). at individual events. Consequently, this paper also contributes to the growing literature on climate-related Against this background, we developed an analysis disasters and crises (Wolbers and others, 2021), their of acute physical climate risks, focusing on a macroeconomic impacts (Hallegatte, 2015), and the sequence of extreme events affecting different parts analysis of their impacts on the financial system (see of the country within a single season. The analysis Ranger and others, 2022 and Adrian et al. (2022) for takes Mexico as an example. Mexico is exposed to more details along with discussion of climate-related many acute physical climate risks, including tropical policy works by central banks and policy institutions cyclones, floods, droughts, and heat waves. Given around the world), and ongoing efforts by the Network the substantial impact of these events, the analysis for Greening the Financial System to strengthen focuses specifically on tropical cyclones and floods. methodologies for the assessment of physical climate Considering a series of extreme events comprised of risks. tropical cyclones and floods is particularly pertinent to Mexico. This is largely due to the concentration of To date, most analyses of acute physical risks have banking sector credit exposures in a few geographical adopted one of two approaches. They have either: (i) regions relative to the hazard distribution of tropical considered scenarios of individual events (e.g., the two cyclones and floods. Banking sector credit to firms is individual flood scenarios considered in the analysis highly concentrated in Mexico City (which historically of climate-related risks for the Dutch financial sector has not experienced tropical cyclones and widespread (Regelink et al., 2017), and the individual typhoon and FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 8 flood scenarios considered in the analysis of climate- banking system appear relatively modest, the results related risks in Japan (Bank of Japan, 2022)); or (ii) indicate that some individual financial institutions could taken an approach that is based on hazard maps that face material risks. For example, while the aggregate show the potential severity of a given peril at a given capital adequacy ratio (CAR) depletion is about 1.2 location for a specific return period3 without considering percent under the high-emissions climate change potential plausible scenarios of individual events or scenario, it declines by more than four percentage sequences of events (rather, they essentially assume points at some banks. Thus, vulnerabilities exist in that all locations are affected simultaneously, even the financial system. These effects indicate potential for very large countries) (e.g., analysis of exposures for severe effects in the future given the potential to flood risk in Denmark (Danmarks Nationalbank, increasing severity and frequency of extreme events 2021) and Norway (Haug et al., 2021)). In the second due to climate change. approach, the lack of consideration of potential scenarios may limit the interpretability of the analyses The implication is sharp: climate risks could have and their application in scenario-based stress-testing significant financial stability risks. This holds true exercises. In the first approach, focusing on individual not just for Mexico but for other countries as well. events while ignoring potential sequences of events Even if a country may not face systematic risk due may risk underestimation of the potential materiality of to any single climate-related extreme event (e.g., an risk. individual tropical cyclone or flood) due to large/diverse geography (such as Mexico, United States), there is a The contribution of our work is the extension of possibility that a sequence of events occurring within the event-based approach to consider sequences a short timeframe (e.g., over the course of a season) of events where we highlight their importance in could negatively affect the entire economic and understanding the potential materiality of possible financial systems. Smaller countries may be less likely climate scenarios on financial stability. The to be affected by multiple events in a short timeframe approach challenges previous studies (e.g., Bos and than some larger countries, but this possibility remains. others, 2018) that make a case for focusing on large- Given that just a single event can already devastate scale disasters only. Even if the scale of individual the economy (e.g., Hurricane Maria in 2017 led to events may not be material at an aggregate national estimated losses equivalent to 226 percent of the scale, by considering a sequence of events we Dominican Republic’s GDP; IMF, 2021), a sequence of demonstrate that it is possible to have a substantial multiple events could generate even greater economic macro-financial impact. This could occur even if none impacts. of the individual events affect the region(s) with highest concentrations of banking credit and other financial The insight from this paper is relevant for exposures (e.g., Mexico City in our application). While informing climate-related policy. Despite the approach used in our analysis does not explicitly coveringonly limited channels of risks and data capture the potential for non-linear amplifications often limitations, our study was able to detect the potential characteristic of compound events (Dunz and others, for material risks from climate-related extreme events. 2021; Ranger and others, 2021), these amplifications While the results give insights into the potential would only serve to potentially further increase the impacts of acute physical risks from tropical cyclones materiality of the modeled scenarios. and flooding on the Mexican financial system, the analysis is regarded as exploratory, and more work In particular, the application of our approach to is needed to refine the model before it can be used the Mexican economy shows that the impact of for regulatory applications. A further extension to other extreme seasons involving floods and tropical climate-related risks faced by Mexico (e.g., drought) cyclones could be material in the short-term following could help to uncover the potential scale of physical these disaster events, as highlighted by the modeled risks to the Mexican economy. In addition, applying our effects on key macro-financial variables. For example, approach to other countries with additional data, and within the first year following a series of extreme events further exploring non-linear compounding effects that occurring under end-of-century climate conditions may arise from a series of events in these countries in a high-emissions climate scenario in Mexico, the and potentially across borders, could help deepen our country’s gross domestic product (GDP) could decline understanding of the nature of compounding physical by up to two percent (relative to the baseline without climate risks. Such a risk assessment could help climate-related extreme events). And, while the short- develop better policy frameworks. term system-wide aggregate impacts on the Mexican 3 - A return period is a measure of the frequency of an event and is defined as the expected time (usually expressed in years) between events exceeding a particular extreme threshold. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 9 The remainder of this paper is structured as component – the direct damages model, the follows: Section II provides an overview of climate- macroeconomic model, and the financial sector impact related disasters in Mexico. Section III presents the model, respectively. Section V presents the results, scenarios that are considered in this paper. Section and Section VI discusses these findings, including IV describes the methodology for each modeling uncertainties and limitations, and concludes. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 10 2 >>> II. Climate-related disasters in Mexico Mexico is highly exposed to a range of disasters associated with acute physical climate change risks, including tropical cyclones and floods. During the last two decades, floods and tropical cyclones have constituted a substantial hazard burden among hydrometeorological disasters (Figure 1), both in terms of frequency of occurrence and the number of people impacted. These perils are the focus of this analysis. However, other acute risks, including drought and heatwaves, as well as chronic risks, may nonetheless be substantial4, along with non-climate-related risks, including earthquakes. Past tropical cyclone events have included cases where some federal states were hit twice in a very short period (e.g., Hurricane Karl and Tropical Storm Matthew in Veracruz in 2010, which caused 25 billion MXN of economic losses and affected more than half a million people) or the storm had two landfalls (e.g., Hurricane Alex in 2010, which caused 25 billion MXN of economic losses and affected 650,000 people), significantly exacerbating financial losses and the impact on populations. Amongst the most impactful recent flood events is the October 2007 floods in Tabasco and surrounding states, which resulted in economic losses of more than 35 billion MXN and affected more than 2 million people. The frequency and severity of floods and tropical cyclones in Mexico are likely to be impacted by climate change, though there are substantial uncertainties in projected changes (as discussed further in Appendix I). 4 - For example, for heatwaves there is evidence of a relationship between extreme heat and credit performance in Mexico (Aguilar-Gomez and others, 2022). For droughts, although direct impacts of droughts on the banking sector may be limited, for example due to low credit exposures to agriculture, they can nonetheless have substantial impacts on communities and the economy. For example, 80 percent of farmers in Mexico are smallholder farmers, with limited access to credit and insurance, and a reliance on rainfed production – hence high vulnerability to drought risk. Recent drought and heatwave conditions in Mexico have highlighted the importance of these risks. For chronic risks such as sea level rise, modeling suggests the potential for substantial economic losses, for example due to loss of coastal ecosystem services (Fernández-Díaz and others, 2022). It is also important to acknowledge that other nature-related risks, including biodiversity-related risks, may interact with climate physical risks. While an analysis of nature-related risks more broadly is beyond the scope of this paper, an initial analysis of exposures to biodiversity loss has been completed by Banco de Mexico (Martínez-Jaramillo and Montanez-Enriquez, 2021). FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 11 >>> FIGURE 1. - Hydrometeorological Disasters in Mexico Floods and tropical cyclones are the most frequent hydrometeorological …and affect the highest number of people disasters in Mexico… Note: The estimates presented in Figure 1 are based on data from CENAPRED. Different methodologies for defining events and for estimating the affected population may result in different estimates of these variables, potentially resulting in higher estimates for some disaster types. The period of the analysis (2000-2020) does not include the recent severe drought and heatwave events occurring in 2021 and 2022. Furthermore, Figure 1 does not include an estimate of the direct and indirect damages and economic losses associated with different hazard types and does not consider chronic risk. Due to the relatively localized nature of floods for only 28 percent of total credit exposure. The and tropical cyclone events, the impacts of these concentration of exposures in Mexico City is important perils on the macroeconomy and financial system when considering the physical risk profile of Mexico, depend on the geographic distribution of economic particularly since the tropical cyclone and flood hazard and banking sector exposures relative to that of profile of Mexico City is relatively mild compared with physical hazards. GDP and credit exposure in Mexico other parts of the country (see Appendix I for further are heavily concentrated in a few states. In 2020, six details). This informs the selection of scenarios states contributed collectively to around 50 percent of modeled as part of our analysis, as outlined in the national GDP: Mexico City (17.51 percent), Mexico Section III. However, it is important to note that the State (9.15 percent), Nuevo León (7.74 percent), low concentration of credit exposure in several federal Jalisco (6.92 percent), Veracruz (4.53 percent), and states known for their active economic activity may in Guanajuato (4.00 percent) (Figure 2). Credit exposure part be due to the recording of credit at the domicile in Mexico is also heavily concentrated geographically of a firm’s headquarters (often in Mexico City), even in a few states, broadly reflecting the geographic though the operations financed by the credit may be distribution of GDP. More than 44 percent (1,513 billion located elsewhere in the country. Bressan and others MXN) of total credit (3,434 billion MXN) is recorded in (2023) have recently started developing a methodology Mexico City, aligning with its contribution to GDP. Other to examine credit on an asset level for a sample of federal states that account for a relatively high reported listed firms in Mexico (taking into account the location credit exposure are Nuevo Léon, Jalisco, State of of individual industrial plants rather than relying solely Mexico, and Sinaloa, which contribute respectively 13 on headquarter location), an approach which, data- percent, 6.4 percent, 4.9 percent, and 3.2 percent of permitting, could be further explored in future. total credit. The remaining 27 federal states account FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 12 >>> FIGURE 2. - Geographic Distribution of GDP and Credit Exposure in Mexico GDP is highly concentrated in a few states… …Credit exposure is too. State contribution to national GDP State contribution to national credit exposure (in percent) (in percent) Sources: author calculation using GDP data from statistica.com and credit data from Banco de Mexico (end of 2021) FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 13 3 >>> III. Scenarios The relative geographical distributions of hazard and exposure in Mexico suggests that an individual severe tropical cyclone or flood event is unlikely to cause material impacts at an aggregate national scale, unless it impacts Mexico City with sufficient intensity (which is relatively unlikely based on the hazard profile of the city). However, it is possible that a sequence of events impacting different parts of the country within a relatively short time period (i.e., within a single tropical cyclone/flood season), could result in a large, combined impact. There is existing evidence for the occurrence of such clustering of events in the region. For example, Atlantic tropical cyclones counts have been shown to be temporally clustered, with periods of high activity interspersed with periods of relatively low activity, in part linked to large-scale patterns, such as those of sea surface temperatures (Villarini et al., 2010; Mumby et al., 2011; Dominguez et al. 2021). To analyze the potential scale of impacts from a sequence of events, five scenarios are included in the analysis. Scenarios TC1, TC2 and TC3 are individual tropical cyclone events, Scenario F1 is an individual flood event, and Scenario ES1 is an extreme season comprised of a sequence of several extreme events (i.e., TC1, TC2, TC3, and F1 all occurring within a few months of each other during one season) (Figure 3). Each of the scenarios is an extreme (but plausible) event and is estimated to result in direct damages with a return period greater than 50 years at a national scale. Each of the individual event scenarios is based either on similar events that have occurred historically or on events drawn from catalogs of synthetic events (generated using statistical resampling algorithms). The scenarios were selected to cover a range of different flood and tropical cyclone-prone regions with financial sector exposures. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 14 >>> FIGURE 3. - Tropical Cyclone and Flood Scenarios The scenarios analyzed are a tropical cyclone …two tropical cyclones impacting Veracruz with two landfalls (Yucatan Peninsula as within approximately 10 days (TC2), … Category 4 hurricane and northeast Mexico with heavy precipitation) (TC1), … …a tropical cyclone making landfall …a river flood in Tabasco and surrounding in Colima impacting Guadalajara with states (F1), … Category 1-2 winds (TC3), … …and an extreme season scenario combining all these events within a single season (ES1). Source: Author illustration. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 15 Two versions of each scenario have been defined these scenarios considers potential future increases to analyze the impacts of climate change on in hazard severity associated with climate change the severity of these scenarios. The first has (i.e., increased windspeed and precipitation for tropical been defined based on the estimated distribution cyclone scenarios and increased streamflow for flood of extreme events under historical/current climate scenario), but assumes constant vulnerability and conditions (“historical conditions”; Scenarios TC1- exposure (i.e., potential future growth of capital stock hist, TC2-hist, TC3-hist, F1-hist, and ES1-hist). The is not considered). Each scenario is intended to have second has been defined based on the estimated the same frequency of occurrence (return period) distribution of extreme events under future climate for the historical and climate change conditions, with change conditions corresponding to RCP8.5 5 by the climate change adjustment applied to the hazard the middle of the century and the end of the century severity only (keeping the frequency fixed).7 A more (“climate change conditions”; Scenarios TC1-cc, TC2- detailed description of the scenarios is available in cc, TC3-cc, F1-cc, and ES1-cc).6 The modeling of Appendix II. 5 - Representative concentration pathway (RCP) 8.5 represents an extreme climate change scenario, corresponding to 8.5W/m2 of total radiative forcing (a cumulative measure of greenhouse gas emissions from all sources) by 2100. 6 - We also analyzed some scenarios using RCP4.5 but these results are not included in this paper. The estimated direct damages for these scenarios fell between those of the “historical conditions” scenarios and the RCP8.5 scenarios. 7 - Analyses of climate change impacts can consider the change in frequency for an event (or series of events) of a given magnitude/severity, and/ or the change in the magnitude/severity for an event (or series of events) of fixed frequency. The two approaches are related due to the relationship between the frequency and severity of extreme events, and the decision regarding which approach to use should be tailored based on the specific application. For this analysis, the frequency was kept fixed to explore the potential impacts of climate change on the magnitude/severity of extreme events. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 16 4 >>> >>> IV. Methods To estimate the financial sector impacts for each of the defined scenarios, the direct damages and economic losses are first estimated (Figure 4). The direct damage estimates, expressed as the percentage of physical capital stock that is damaged for a given scenario, are used as an input to the macroeconomic impact modelling, the output of which is then used to estimate financial sector impacts. FIGURE 4. - Method for Direct Damages, Economic Impact, and Financial Sector Impact Modelling To estimate the financial sector impacts for each scenario, the direct damages and economic losses are first estimated Identification of climate hazards + related data/model Scenarios for “historical conditions” and “climate change conditions Specifically, the direct damage estimation combines an analysis of historical data with a model-based approach. Bottom-up model- based approach Exposure modeling (spatial disaggregation of capital stock data based on population and Scenarios Analysis of loss data (e.g., CENAPRED data) Physical direct damages to settlements) for comparable capital stock historical events, Hazard modeling with adjustments (flood based on Fathom applied for change hazard maps, wind based in exposure, on CLIMADA; cimate inflation, etc change adjustments approximated from Economic losses (productivity literature) shocks, etc.) Vulnerability modeling (based on published regional vulnerability curves) Impact on financial sector Physical direct damages to capital stock Source: author illustration. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 17 IV. I. Direct damages model estimation methodology results for individual events are summed. Note that a simple addition is plausible in the case of the direct damages for ES1 The methodology used to estimate direct damages8 since the spatial extents of the direct damages of the for each scenario combines (i) an analysis of constituent scenarios (Scenarios TC1, TC2, TC3 and historical loss data for comparable events; and (ii) F1) do not overlap.10 However, for indirect economic a model-based approach using exposure, hazard, impacts, compounding effects may be expected to and vulnerability data (Figure 4). Historical loss data occur due to transmission channels that can result analysis results are used to inform and validate the in impacts beyond the original spatial extent of the model-based approach, with the final direct damage area directly affected by an event. Yet, as explained estimates for each scenario based on a combination further below, in this analysis, compounding effects of the results of these two perspectives. are only captured as an economic general equilibrium effect to the extent that the direct effects of damages For the analysis of historical CENAPRED9 loss generate consequent endogenous effects on TFP, data, historical losses for events comparable to unemployment, and financial markets channels via the each scenario (where available) are scaled to macro model. This is the extent to which compounding a 2020 loss estimate based on exposure growth can be captured in the model, given modelling and and inflation assumptions. Historical loss data were data limitations. extracted from detailed CENAPRED reports for events like the defined scenarios. CENAPRED provides While several commercial catastrophe models economic loss data divided into direct damages and are available for Mexico, the simplified approach indirect losses by state. In cases where the split developed for our analysis is less resource between direct and indirect impacts is unavailable, this intensive and can be implemented when outputs is estimated using the average ratio for other events of from more sophisticated models are not available the same peril type (flood or tropical cyclone). These (as is often the case in many emerging markets data were then scaled to estimated “as-if” 2020 losses, and developing economies, where locally validated using exposure growth and inflation estimates. These catastrophe models may not be available). The government data are considered reliable historical loss catastrophe models that are available for Mexico (e.g., datasets and were compared with other historical loss those developed by Universidad Nacional Autónoma databases, including EMDAT. de México, Evaluación de Riesgos Naturales, Verisk, RMS, and other commercial vendors) are widely The model-based approach combines three used in the insurance industry and for regulatory modules: (i) exposure, which characterizes the purposes. We did not have access to these models spatial distribution of capital stock; (ii) hazard, which for this analysis, and it is important to note that the characterizes the flood depths and/or windspeeds loss estimates could potentially be substantially associated with a scenario; and (iii) vulnerability, which improved using such models. There are limitations and estimates the damage for a given flood depth and/or uncertainties associated with each of the modules of windspeed for the exposure at risk. A comparison of the approach we used. The model results are sensitive the results from the approach based on historical loss to each of these components and should thus be data and the model-based approach was conducted, interpreted with full recognition of the limitations of to further tune the parameters of the model-based the analysis. Further analysis based on more granular approach, and to determine the final loss estimates. exposure information and leveraging the latest climate science and catastrophe modeling expertise in Mexico To estimate the direct damages for the extreme would allow damage estimates to be refined. season scenario (Scenario ES1), the direct damage 8 - Direct damages (first-round effect) refer to the physical damage to assets caused directly by an event, with the losses occurring at the time of the disaster or shortly thereafter. Examples of direct physical damages include the destruction of residences, productive capital, infrastructure, and crops. 9 - Centro Nacional de Prevención de Desastres 10 - Were the individual event scenarios to overlap spatially, it is possible that the direct damages of the extreme season scenario would not be equal to the sum of the direct damages of the individual events. For example, there might be overlapping destruction between climate events (e.g., the same building being destroyed by two events separately), making the damage sub-linear. Or on the contrary, a scenario could weaken some infrastructure that is then destroyed by a subsequent event, potentially resulting in super-additive damages. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 18 IV. I. I. Exposure module which provide population and built-up area densities. The rationale is that capital stock (e.g., machinery and equipment) is more likely to be in populated or The exposure module involves two steps: (i) the built-up areas. The maximum of the two normalized scaling of total physical capital stock data by state GHS_POP and GHS_SMOD densities per pixel is used to an estimate of 2020 exposure; and (ii) the spatial to estimate capital stock density. The capital stock disaggregation of this data. This process was used density map is then classified into two sectors using to generate a map of exposures that can subsequently agriculture and built-up land cover classes identified be overlaid with hazard maps. from the Copernicus Global Land Service Land Cover data set (Buchhorn and others, 2020). Finally, the The scaling of physical capital stock data used capital stock data is distributed by means of its density. variables from the World Development Indicators While this methodology provides a useful approach database to adjust the 2013 capital stock data from for disaggregating the available capital stock data to the Instituto Nacional de Estadística y Geografía a more granular geographic dataset, it is important (INEGI) to estimates for 2020. INEGI capital stock to note the limitations of this technique. For example, dataset for 2013 (INEGI, 2013) includes both public the spatial population distribution in Mexico City may and private capital stock, including fixed assets (e.g., not reflect the spatial distribution of capital stock (e.g., real estate, infrastructure), and movable assets (e.g., due to inequalities and the non-homogenous spatial machinery and equipment), reported at a state level. distribution of wealth and poverty). From the dataset, the ratio between total capital stock and annual GDP was extracted. Data from the IV. I. II. Hazard module World Development Indicators database (WB, 2022), including data regarding inflation in Mexico between 2000-2020 as well as the nominal GDP values for The approach to model hazard is based on the the period 2000 -2020 and the gross fixed capital perils considered in each scenario, namely wind formation, was then used to estimate the value of the and/or flood. For some tropical cyclone scenarios, capital stock in 2020. wind is the main driver of losses, whereas, for others, the wind hazard is considered negligible, with most State-level exposure estimates for 2020 were of the losses driven by flooding caused by heavy spatially disaggregated to a resolution of precipitation associated with the tropical cyclone. approximately 100m and split into agricultural and Depending on the nature of the scenario, the modeling built-up (industrial, commercial, and residential) approach was adjusted accordingly (i.e., for some classes. The spatial disaggregation utilizes the Global tropical cyclones, the wind was modeled, while for Human Settlement Layer – Population (GHS_POP; others, only flood was modeled). Storm surge is not Schiavina and others, 2019) and the Settlement modeled. Model Grid (GHS_SMOD; Pesaresi and others, 2019), FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 19 For the “historical conditions” tropical cyclone are a global product and may not have been locally scenarios, the approach used to generate a calibrated for Mexico. Using different flood hazard windspeed map to model the wind hazard differed maps can result in large differences in risk estimates. between scenarios: To define the flood hazard for a given scenario event, it was necessary to select a hazard map with an • For Scenario TC1-hist, the wind component of appropriate return period and to delineate the area tropical cyclone scenarios was modeled based on in which this hazard map is to be applied. Here, historical IBTrACS track data for similar historical return periods for spatial units were defined as river events, with wind fields calculated from the track catchments within a state. The approach to defining data using CLIMADA’s (Aznar-Siguan and Bresch, the appropriate return period for each state/river basin 2019) implementation of Holland’s (2008) method. varies between the scenarios, based on a combination While several historical wind fields were modeled, of historical flood footprints for similar historical events, it was decided to use the Hurricane Wilma 2005 event descriptions, precipitation maps for similar track as a reference event for Scenario TC1-hist, historical events, precipitation and streamflow return since its landfall path and intensity is considered period analyses available in the literature, and a representative of the scenario’s first landfall, which calibration based on the comparative analysis outlined is associated with the wind-driven loss component below. of the scenario. In the case of “climate change conditions” • The wind component of Scenario TC2-hist is scenarios for tropical cyclone-driven flooding, negligible, with most damages driven by flooding a change in rainfall is estimated by Knutson caused by precipitation associated with the tropical and others (2015). The estimate from Knutson and cyclone. Hence, the wind hazard was not modelled others (17.3 percent increase in rainfall rate) is only in detail for this scenario. available for one RCP and time horizon (RCP4.5 late 21st Century). To derive factors for RCP8.5 and • For Scenario TC3-hist, for which there is no similar other time horizons, the adjustment factor is scaled historical event, a similar track was selected from according to relative radiative forcing estimates based the STORM stochastic track dataset (Bloemendaal on the methodology implemented in CLIMADA for wind and others, 2020), with wind fields subsequently adjustment factors. The rainfall adjustments are then simulated using the CLIMADA implementation of translated to an estimated change in streamflow based Holland’s (2008) method. on an assumption that the precipitation-streamflow elasticity for extreme events approaches unity (Breinl In the case of the “climate change conditions” and others, 2021). Finally, the change in streamflow scenarios, for the wind hazard, an intensity is converted to an estimate of the change in return adjustment based on Knutson and others (2015) period based on published streamflow return period is applied using CLIMADA. Since the intensity estimates for rivers in Mexico (Isela and Zarco, 2014, adjustment factor from Knutson and others (4.5 percent Neri-Flores and others, 2019). This approach uses increase in intensity for the North Atlantic basin) is several simplifying assumptions, and the results should only available for one RCP and time horizon (RCP4.5 be interpreted recognizing the limitations. For example, late 21st Century), the CLIMADA implementation the results highly depend on the elasticity assumption, scales this adjustment factor for RCP8.5 and other which has not been locally proven for Mexican rivers. time horizons by interpolating them according to their In addition, published streamflow return periods are relative radiative forcing. It is important to note that only available for some gauges along the rivers and tropical cyclone projections for future climates are may not be representative of the flood frequency subject to substantial uncertainty, and different studies relationships elsewhere in a catchment. at basin or sub-basin scales have varied in their findings, as outlined in Section II. In the case of “climate change conditions” scenarios for non-tropical cyclone-driven flooding, For flood, the hazard module is based on the a change in streamflow for a 1-in-100-year is Fathom hazard maps (Sampson and others, 2015), estimated using Di Sante and others (2020). The available for a range of return periods. The riverine change in streamflow is converted to an estimate of the (defended) hazard maps are used to model flood risk change in return period based on published streamflow in Scenarios TC1, TC2, and F1, based on a literature return period estimates (Mora and others, 2008). This review highlighting that riverine flooding was the main approach is subject to similar limitations to those driver of flood-related losses for similar historical outlined above for the tropical cyclone-relating flooding events. It is important to note that the hazard maps climate change adjustments. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 20 While the approaches for climate change asset from a statistical distribution. Due to an inability adjustments outlined above result in increases to differentiate between exposure data, the same in flood risk, it is important to note that several vulnerability curves have been used for all exposures papers indicate that the risk may decrease for within a given class (built-up or agriculture), rather some rivers in Mexico (see, for example, Haer than differentiating between individual assets based on and others, 2018, Hirabayashi and others, 2013, factors such as their construction type, age, number of Hirabayashi and others, 2021). This may be a result stories, and occupancy. Using a different vulnerability of potentially drier antecedent conditions in the river curve, adjustments based on modifiers related to catchments due to the influence of climate change on detailed characteristics of the exposed assets, or a other hydrologically important climate variables (e.g., probabilistic sampling method may substantially impact increased temperature and evaporation). Thus, the the analysis results. climate change scenario damage estimates presented in this analysis may be more representative of a “bad case” change in flood risk, yet potentially not the “worst IV. II. Macroeconomic impact case” given that some model projections show even more severe increases in rainfall and/or streamflow The estimates of direct damages and destruction with climate change than the increases assumed in to the physical capital from hazards constitute the this analysis. key link between climate-driven risks and macro- financial outcomes that can be further translated IV. I. III. Vulnerability module into impact on the financial sector. These damages are first approximated at the regional level and then aggregated to generate the damage at the national Published vulnerability curves were used to level. As such, the estimated direct damage rates to calculate the damage for a given hazard for the the physical capital stock can be interpreted as an modeled exposures. The European Commission Joint immediate direct shock (equivalently, a depreciation Research Centre vulnerability curves (Huizinga and shock) to the capital stock in an aggregated others, 2017) were used for flood damage estimates, macroeconomic model. However, it is important to and the Eberenz and others (2021) regional tropical contextualize the macro-financial impact of climate cyclone impact functions were used for wind damage risk-driven estimated damages in terms of a country’s estimates. The flood vulnerability curves relate flood specific physical risk characteristics. In this regard, the depth to damage. The tropical cyclone vulnerability overall size, geographical location of the country, and curves relate wind speed to damage. Separate curves exposure of various locations of the country to various were used for agriculture and built-up (combination physical hazards play a fundamental role in driving the of residential, commercial, and industrial) exposures. aggregate damage estimates as mentioned above. It is important to note that other vulnerability curves The limitations of the methodologies and models as are available for Mexico (e.g., CENAPRED, 2006) discussed above also apply. and that model results are sensitive to the selection of vulnerability curves. The damage estimates from the extreme season scenarios discussed above constitute the key input Substantial uncertainty is associated with in generating scenario-dependent macro-financial modeling vulnerability (Kaczmarska and others, paths. One of the key objectives of the standard stress 2018). This includes uncertainty in both: (i) the testing analysis (such as the ones used in FSAPs) is development of the vulnerability model, including to use extreme but plausible risk scenarios. This is aleatory uncertainty (due to randomness of the because the analysis primarily attempts to quantify processes governing the relationships between hazard channels and mechanisms of risk propagation and severity and damage) and epistemic uncertainty (due their implications due to potential materialization of to a lack of knowledge or data); and (ii) the use of the tail risk events. Given the objective of quantifying vulnerability model when modeling a given scenario impacts from tail risks and considering the materiality (Trendafiloski and others, 2017). In this analysis, of the estimated damage rates, the natural choice vulnerability has been modeled as a deterministic is to explore macroeconomic impact of the extreme function of wind speed or flood depth. However, it season scenarios. Thus, the results discussed is important to note that other methods may better below will correspond to these set of scenarios and capture uncertainty. For example, vulnerability may corresponding damages. The horizon of the analysis be modeled by sampling the damage ratio for each is three years.11 11 - The short term three years horizon is consistent with the analysis in the stress testing exercises by central banks and regulators around the world, with the historical and future climate change conditions modeled as added risks to this underlying framework (which was calibrated using recent data). Although the damage estimates from future climate change scenarios explore mid-century and end-of-century physical risk (from tropical cyclones and floods), it is possible that the damages from these risks could materialize within the short horizon, albeit with a substantially lower probability. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 21 The macroeconomic impacts conditioned on the quickly for the output to recover. One of the estimated damages are modeled using the same drivers of TFP reduction after a disaster could underlying Global Macrofinancial Model (GFM) model come from the complementarity of infrastructure as in the 2022 Mexico FSAP stress test, to enable and non-infrastructure capital, meaning that if a comparability of the results relative to baseline without disaster damages infrastructure capital, then non- a climate risk overlay (i.e., isolating the impacts of infrastructure capital also becomes unproductive, the potential changes to tropical cyclone and flood which can magnify the impact of the disaster hazard due to climate change from other potential (Hallegatte and others, 2022b). The shock was future macroeconomic changes). This is a global multi- calibrated to be twice as much as the damage country DSGE model of the world economy, covering rate and assumed to be highly persistent. This is 40 countries, featuring macro-financial linkages and consistent with a general finding in the literature a rich set of transmission channels. The model is that overall productivity tends to be significantly regularly used to generate scenarios in IMF’s FSAP lower after disaster events (e.g., Dieppe and others risk analysis such as solvency stress tests. The model (2020)). allows for generating general equilibrium paths of macro-financial variables given various shocks, such • Impact on unemployment: Since capital stock as those arising from labor market frictions, productivity and labor also complement each other during etc. Some of these shocks, the primary one being production, it is economically intuitive to consider the immediate destruction of physical capital stock a concurrent shock to unemployment in the event (analogous to shocks to capital depreciation), were of large/immediate destruction to capital stock. applied in this analysis as discussed above. Thus, it For example, as factories and other productive forms the starting point of the more comprehensive infrastructures are damaged, the workers that stress testing exercise.12 complement the use of this capital stock are likely to be displaced as well. The shock is informed It is well known that the indirect effects via other by the elasticity of unemployment to changes in channels accompany and could potentially capital stock in Mexico. amplify the initial direct effects resulting from physical hazards. In this regard, the team considered • Effects from financial markets: Given the three additional channels and amplifiers to account sequence of climate events considered in the for impact that is not captured by immediate direct extreme season scenario, it is highly plausible that damages. Note that there could be multiple other equity markets might also experience negative channels in addition to these. However, given the data shocks. Since markets are forward-looking, such and model limitations, these shocks lend themselves disaster events could imply a potential negative to easy integration into the existing macro modeling impact on the cash flow generating capacity of framework (GFM) while also capturing relevant the firms, thereby lowering the market value of the economic dynamics one would likely observe in the equity of the firms. The shock is informed based on immediate aftermath of extreme events. the elasticity of changes in overall equity market returns to changes in capital stock in Mexico.13 • Direct destruction of physical capital: The direct damages to the existing physical capital were estimated at the regional state level and IV. III. Financial sector impact aggregated to arrive at the damages to the national model capital level. This constitutes the direct channel of risk transmission from climate events, leading to an immediate impact or shocks to the physical capital The approach follows a standard stress testing in the macro model. methodology where the CAR of the entire financial system was projected, although this is not a standard • Impact on TFP: The analysis considered shocks stress testing exercise. Standard stress testing practice to the TFP arising from the direct damages to is generally designed to assess the resilience of banks the capital stock. Disasters reduce productivity and the financial system against pass-fail criteria, since it may not be possible to rebuild the capital based on the impact on CARs relative to the regulatory minimum requirements. 12 - Since the GFM is a commonly used model across various FSAPs at the IMF, it is beyond the scope of this paper to discuss the intricate details of this model. We refer the readers to Vitek (2018) that documents the empirical and theoretical features of the model, available at: https://www.imf. org/en/Publications/WP/Issues/2018/04/09/The-Global-Macrofinancial-Model-45790 13 - While other macroeconomic and financial variables (including interest rates, probabilities of default of loans/debt) could also be impacted after materialization of climate events, their impact (other than the endogenous impact already captured by the scenario model) is not considered in this stage of macro- financial path generation for simplicity given data limitations. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 22 Importantly, the standard stress tests generally rely the banking sector under different climate scenarios on historical relationships between macro-financial were obtained from aggregating up the granular bank- tail risks and the consequent impacts on banks and portfolio level impact on credit impairments, net interest the financial sector. The corresponding adverse risk income, net trading income, risk-weighted assets, non- scenarios are extreme but plausible based on historical interest expense, and so on. episodes. In the case of climate risks, the scenarios are similarly defined to be extreme but plausible, The translation from the macro-financial scenarios considering potential future changes in the likely to the bank level and ultimately aggregate capital severity of extreme events due to climate change. impact is non-linear but a comprehensive, granular exercise that captures effects from The team applied the fully-fledged scenarios- various segments of banks’ portfolios -- credit based macro-financial solvency stress testing impairments, net interest income, market effects, framework to quantify the impact of climate- risk-weighted assets etc. In this regard, the related risks in the Mexican financial sector. This scenarios represent starting point inputs to the stress internally developed model is also used in the 2022 testing framework that generates heterogeneous and Mexico FSAP solvency stress testing analysis. The non-linear impact across banks and risk segments/ full details of the methodology are available in the portfolios given the intermediate estimated satellite published technical note14, and here we summarize models that link the scenario outputs to bank CAR level the key features to highlight various kinds of effects impact, ultimately generating impact on systemwide the framework captures. CAR. The analysis considered a full top-down risk The impact on CAR is represented as deviation analysis that includes various risk components from baseline CAR projections for the baseline in the banking books: credit risk, interest rate plus current/historical condition scenario and risk, impact on risk-weighted assets, market risk, baseline plus climate change conditions scenario. etc. Analysis used granular bank-level regulatory/ The baseline of the exercise is the same one used in supervisory data as of the end of 2021, covering the 10 the risk analysis of the 2022 Mexico FSAP and based largest commercial banks by asset size (including six on available 2022 World Economic Outlook projections. domestic systemically important banks), representing Given that the two climate related scenarios are more than 80% of the banking sector assets. Such considered as additional risks on top of the baseline, granularity allows for exhaustively quantifying the the impact on CARs can also be interpreted analogous impact on the aggregate capital ratio (CAR) in the to that in the baseline versus adverse scenario(s) banking system. 15 The aggregate CAR impacts in setting of the standard stress test exercise.16 14 - Mexico FSAP TN on Systemic Risk Analysis and Stress Testing (December 2022) available at: https://www.imf.org/en/Publications/CR/ Issues/2022/12/08/Mexico-Financial-Sector-Assessment-Program-Technical-Note-on-Systemic-Risk-Analysis-and-526751 15 - Further, since the macro-financial scenarios, consisting of pathways of various macro-financial variables were used as input to the stress testing framework, the spillover from the aggregate macro-financial channels to bank specific impact is already implicitly captured in the framework. However, second-round effects from financial sector to the economy and the feedback loops may not be fully captured. Modelling of such effects is a highly complicated exercise and thus, an active area of research in stress testing. 16 - These estimates are for an extreme season scenario with a high return period (i.e., they are not expected to occur at this level of severity every year). The estimated damages are the total damages for the entire season. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 23 5 >>> V. Results Direct damages for the most severe scenario, namely the extreme season scenario, are estimated to be approximately 0.9 percent of the total national capital stock for “historical conditions” (ES1-hist) and 1.9 percent for “climate change conditions” (ES1-cc end-of-century) (Table 1). 17 Direct damages for individual scenarios vary from 30 billion MXN (approximately <0.1 percent of capital stock) to 600 billion MXN (approximately 0.7 percent of capital stock) under “historical conditions”. The return period of the direct damages at a national scale is broadly estimated as 50 years for TC1-hist, 50-100 years for TC2-hist, 500 years for TC3-hist, and 50-100 years for F1-hist, based on comparisons with wind event modelling from the Global Assessment Report 2015 analysis (UNISDR, 2015) and a statistical analysis of historical flood event data from EMDAT using the World Bank Financial Risk Assessment Tool (WB, 2021b). However, there is substantial uncertainty associated with these estimates due to the lack of access to a full catastrophe model, and limitations of the datasets used for comparison. A comparison with the “climate change conditions” scenarios indicates that damages may increase by 40-50 percent by mid-century under RCP8.5, and by more than 100 percent by end-of-century, yet these results are highly dependent on the modelling assumptions and are subject to substantial uncertainty. 17 - A fully-fledged solvency stress test exercise is assumed under the physical risk scenario where shocks from economic losses is fed into the same scenario generation framework (as discussed earlier). This allows a full top-down exercise to be performed that includes all components (credit, market, net interest income, risk weighted assets, etc.). For more details on the methodology please see Mexico FSAP Technical Note on Systemic Risk Analysis and Stress Testing (December 2022). FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 24 >>> TABLE 1. - Direct Damage Estimates for Physical Risk Scenarios Direct damage Climate change impact Scenario Climate conditions1 rate (percent of (relative to “historical capital stock) conditions” scenario) Historical conditions 0.14 percent TC1 Climate change conditions mid-century 0.21 percent 43 percent Climate change conditions end-century 0.23 percent 59 percent Historical conditions 0.03 percent TC2 Climate change conditions mid-century 0.05 percent 54 percent Climate change conditions end-century 0.08 percent 124 percent Historical conditions 0.67 percent TC3 Climate change conditions mid-century 1.01 percent 51 percent Climate change conditions end-century 1.45 percent 118 percent Historical conditions 0.05 percent F1 Climate change conditions mid-century 0.08 percent 54 percent Climate change conditions end-century 0.13 percent 145 percent Historical conditions 0.90 percent ES1 Climate change conditions mid-century 1.35 percent 50 percent Climate change conditions end-century 1.89 percent 110 percent Note: 1/ “Climate change conditions” are based on RCP8.5. The impact of physical hazards from floods and effect on GDP rises to as much as 2 percent in 2022 tropical cyclones on the overall macroeconomy (Figure 5). These results confirm that increased climate could be significant as observed from the impact risks could have significant impacts going forward and on GDP. Under the extreme season scenario, the the economic costs of climate driven risks are non- impact on GDP in growth terms (in deviation from the trivial. Additionally, the initial impact of the risk could baseline) under the “historical conditions” (ES1-hist) is persist well into the future from additional indirect about 1.4 percent in 2022. However, considering the channels as discussed above.18 impact of climate change (ES1-cc end-of-century), the 18 - Since our climate risk methodology was first applied in context of 2022 Mexico FSAP, we used banking sector data as of end 2021 as the starting point of macro-financial impact projections up to three years. Further, these scenario conditional projections, as in general stress testing- based approach, concern extreme but plausible tail risks and thus, do not constitute forecasts. Thus, the projection horizon of three years from 2022 to 2024 in the results largely serve to illustrate the applicability of our approach while also emphasizing the potential materiality and importance of extreme seasons for financial stability in the short term. This means that whenever most recent regulatory and other macro-financial data required for the analysis are available in a given country, our approach can be well adapted and applied there as well. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 25 >>> FIGURE 5. - Impact of Extreme Season Scenarios on Macroeconomic Variables Note: The extreme season scenarios shown are ES1-hist (historical conditions) and ES1-cc end-of-century (climate change conditions based on RCP8.5). The current physical risk analysis suggests that These estimates show that significant vulnerabilities impacts on the banking system might occur in the do exists among some individual banks, even if the near term, but not at a scale to generate a systemic system-wide aggregate impact is relatively modest. stress event. Figure 6 shows that under the extreme season scenario under historical conditions (ES1-hist), While the system CAR would remain above the aggregate CAR of the Mexican banking system regulatory requirement, the depletion of 1.2% would decline by about 1 percentage point versus capital relative to baseline is indeed non-trivial. the baseline. Considering risk under climate change This shows that climate risk could indeed pose conditions (ES1-cc end-of-century), the CAR depletion challenges going forward, because while the starting is approximately 1.2 percentage points. While these CAR in the system was relatively higher at end 2021, are relatively modest effects, they nevertheless in the future this might not be the case. Hence, the suggest non-trivial capital impacts that increasingly analysis highlights that the banking system could in severe climate events could generate in the future. fact face pressures in the future if the prevailing state Furthermore, there is also significant heterogeneity of of the economy and banking sector is weaker. In this capital depletion across banks. For example, at some context, and bearing in mind the implicit constraints banks the CAR could decline by almost 4 percent due to modelling assumptions, the CAR depletion does under historical conditions and over 4 percent under highlight potential future vulnerabilities. climate change conditions, relative to the baseline. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 26 >>> FIGURE 6. - Impact of Extreme Season Scenarios on Capital Adequacy Ratio Note: The extreme season scenarios shown are ES1-hist (historical conditions) and ES1-cc end-of-century (climate change conditions based on RCP8.5). FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 27 6 >>> VI. Discussion and Conclusion VI. I. Uncertainties, limitations, and interpretation The results of the physical risk analysis (focused on acute risks associated with tropical cyclones and floods) highlight the potential for non-trivial impacts of these hazards on the financial system in Mexico. Material impacts due to flood and tropical cyclone disasters may be realized both under historical/ current climate conditions and considering the impacts of climate change, albeit with greater severity. Although the results of our analysis are not directly comparable with those of other studies (due to differences in the methodology, analyzed risks and scenarios, country specificities, and reported metrics), the magnitude of the estimated impacts is within the range of estimated impacts found in other analyses of physical risk in the region.19 For example, whereas our analysis estimated a CAR depletion of 1 percentage point under ES1-hist, a study of severe flood scenarios in Colombia estimated average declines in the CAR between 0.3 and 1.0 percentage point for domestic banks (Reinders et al. 2021). It is important to recognize that the analysis is only limited to tropical cyclones and floods, and that the materiality of the physical risks associated with climate change could be much larger if other risks (e.g., droughts, heatwaves, and chronic risks) are also considered. Furthermore, while our analysis focuses on financial sector impacts, it is important to acknowledge that the risks analyzed may also have more widespread impacts on livelihoods and communities, some of which may not be directly reflected in financial sector impacts, due in part to financial inclusion challenges, particularly for poor and vulnerable households (WB, 2019). The analysis is subject to uncertainty at multiple levels, and the results should thus be regarded as a preliminary “first estimate” using the data and tools available. Uncertainties exist in all layers of the physical risk analysis, including the definition of climate and disaster scenarios, the model used to estimate direct damages, the transmission channels in the macro-financial model, and the financial sector impact model. However, although substantial uncertainty exists on the quantitative side (i.e., strength of climate change risks), the qualitative effects (i.e., the sign of risk materialization) are increasingly well understood. 19 - There is a growing number of physical risk analyses in Latin America, including at a regional level (Calice and Miguel, 2021), for floods in Colombia (Reinders et al., 2021), for drought in Brazil (Banco Central do Brasil, 2022), and for heavy rainfall and droughts in Peru (Romero et al., 2022). In Mexico specifically, Aguilar-Gomez et al. (2022) analyzed the impacts of extreme heat events on non-performing loans. However, the results of the analyses are not directly comparable due to differences in the methodology, analyzed risks and scenarios, country specificities, and reported metrics. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 28 The modeling approach is limited in its ability to to better inform future financial stability risks from capture indirect impacts and non-linear effects climate change. Further research will be needed but nonetheless highlights that the order of before drawing strong prudential policy implications. magnitude of potential impacts may be significant Improved data collection and further modeling could and relevant from a financial stability policy support a more granular analysis and enable a more perspective. While modeling direct damages captured comprehensive understanding of the macro-financial impacts to physical capital stock by considering the risks shown in this study. geographical distribution of hazard and exposure, this geospatially explicit modeling did not include modeling of potential indirect effects, including those associated VI. II. Moving from individual with business interruption and large-scale disruptions events to sequences of events in supply chains. While the macroeconomic model attempted to capture some of these indirect effects, important limitations exist. Given the aggregated Future research on climate physical risk could national-scale macro model used to arrive at the help to further develop the methodologies to impact on GDP and other macro-relevant variables, understand the potential for sequences or clusters the regional variations and interconnected dynamics of extreme events. Although our analysis has already across different parts of the country could not be uncovered potential for significant future financial captured sufficiently granularly. Additionally, given the stability risks from an extreme season scenario linearized dynamics of the model, the significant non- consisting of a simple sum of individual events, linearities that could be associated with amplification additional work is required to develop approaches effects are not fully captured. Further, the analysis to capture potential non-linear compounding or also does not account for potential future adaptation amplifying impacts of clusters of events which could policies that could mitigate the impact of extreme further increase the materiality of modeled impacts. climate events and thereby reduce the severity of the While our analysis was only limited to events occurring impact on the banking sector. Nevertheless, despite within Mexico, future work could also consider the obvious limitations due to the coverage of a smaller the potential for sequences or clusters of extreme set of risk propagation mechanisms, the non-negligible events occurring internationally (e.g., international impact on the CAR and the potential for significant breadbasket failures driven by regional drought events future risks show that our approach can capture occurring simultaneously with national-scale disasters), and sharply quantify the materiality of physical risks considering potential teleconnections and other drivers to the financial system. Thus, our approach can be of correlations in the occurrences of disasters. integrated into existing policy assessment frameworks FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 29 7 >>> Appendix I. Climate-related disasters in Mexico Mexico is highly exposed to disasters associated with acute physical climate change risks, such as tropical cyclones, floods, droughts, and heat waves. During the last two decades, floods and tropical cyclones have constituted a substantial hazard burden among hydrometeorological disasters, both in terms of frequency of occurrence and the number of people impacted, with more than 15 million people affected by these hazards over the last two decades. These perils are the focus of this analysis. However, other acute risks, including drought and heat waves, and chronic risks, may nonetheless be substantial.20 It is important to note that Mexico is also subject to substantial risk from other natural disasters that are not climate-related, including earthquakes. Mexico is exposed to tropical cyclones from both the Atlantic and Pacific basins. Similar numbers of major hurricane landfalls originate from these two basins (AIR, 2018). These tropical cyclones impact the coastline with strong winds, heavy precipitation, and storm surges. Mexico’s mountain ranges along the coast impact the rate at which landfalling storms dissipate and result in orographic lifting, which increases rainfall, meaning that tropical cyclones in Mexico are often accompanied by heavy rainfall and substantial flooding. Precipitation often extends inland, where it can cause extensive flooding. Eight of the ten topmost impactful tropical cyclone events in Mexico since 2020 impacted states overlooking the Gulf of Mexico, according to data from CENAPRED. There are cases where some federal states were hit twice in a very short period (e.g., Hurricane Karl and Tropical Storm Matthew in Veracruz in 2010, which caused 25 billion MXN of economic losses and affected more than 500,000 people) or the storm had two landfalls (e.g., Hurricane Alex in 2010, which caused 25 billion MXN of economic losses and affected 650,000 people), significantly exacerbating financial losses and the impact on populations. Due to its inland location, the potential occurrence of tropical cyclones impacting Mexico City with high windspeeds is relatively low compared with the tropical cyclone hazard experienced by coastal regions. 20 - For example, for heatwaves there is evidence of a relationship between extreme heat and credit performance in Mexico (Aguilar-Gomez and others, 2022). For droughts, although direct impacts of droughts on the banking sector may be limited, for example due to low credit exposures to agriculture, they can nonetheless have substantial impacts on communities and the economy. For example, 80 percent of farmers in Mexico are smallholder farmers, with limited access to credit and insurance, and a reliance on rainfed production – hence high vulnerability to drought risk. Recent drought and heatwave conditions in Mexico have highlighted the importance of these risks. For chronic risks such as sea level rise, modeling suggests the potential for substantial economic losses, for example due to loss of coastal ecosystem services (Fernández-Díaz and others, 2022). It is also important to acknowledge that other nature-related risks, including biodiversity-related risks, may interact with climate physical risks. While an analysis of nature-related risks more broadly is beyond the scope of this paper, an initial analysis of exposures to biodiversity loss has been completed by Banco de Mexico (Martínez-Jaramillo and Montanez-Enriquez, 2021). FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 30 Mexico is subject to several different types of frequent under RCP 8.5. In comparison, Bloemendaal, flooding, including river, pluvial, and coastal and others (2022) found only minor changes in 1-in- floods.21 River floods and pluvial floods typically occur 100-year and 1-in-1000-year maximum tropical following extreme rainfall. Major drivers of rainfall in cyclone wind speeds for a future climate (SSP 5-8.5 Mexico include tropical cyclones and easterly waves ; 2015-2050) compared with a past climate (1980- during summer, and cold fronts during winter (Magaña 2017). Sea level rise may exacerbate storm surge risk and others, 2003), with the rainy season running associated with tropical cyclones. The precipitation from June to November. Coastal floods are also associated with tropical cyclones is projected to common, due to the occurrence of tropical cyclones increase, with projected increases ranging from ~5 that generate storm surges. The most impactful percent to 40 percent (Knutson and others, 2020), flood events in Mexico over the past two decades driven by increased tropical water vapor. Bruyère et occurred in Tabasco, and its surrounding states al. (2017) found likely future increased precipitation (Chiapas, Veracruz, Oaxaca and Puebla), generating for tropical cyclones in the Gulf of Mexico. However, significant economic losses and impacting millions of there is substantial uncertainty associated with these people, based on data from the Centro Nacional de projections. Prevención de Desastres (CENAPRED). For example, the October 2007 floods in Tabasco and surrounding Climate change is expected to affect the frequency states resulted in more than 35 billion MXN economic and severity of extreme precipitation in Mexico. losses and affected more than 2 million people. While The historical (1985-2014) 1-in-100-year maximum floods occur in Mexico City, they are typically pluvial 5-day cumulative precipitation for Mexico is expected flood events restricted to a few neighborhoods. to occur with a return period of less than 50 years by the end of the century under SSP5-8.5 (WB, 2021a). The frequency and severity of floods and tropical Similarly, the historical 1-in-100-year maximum 1-day cyclones in Mexico are likely to be impacted by cumulative precipitation is expected to occur with climate change, though there are substantial a return period of less than 40 years by the end of uncertainties in projected changes. Projections the century under SSP5-8.5. Such increases in the of future changes in tropical cyclone frequency and frequency of extreme rainfall events could in turn lead severity are subject to considerable uncertainty, to increased flood risk. Although increased extreme particularly when considering changes at a basin or precipitation may increase flood risk, it is important sub-basin scale. Considering basin-scale projections, to note that other factors may influence flood risk climate change is projected to impact tropical cyclone too, including changes in antecedent conditions, risk in Mexico, with frequency estimated to decrease due to increase temperature resulting in reduced in the North Atlantic and increase in the Northeast soil moisture. While some studies suggest increased Pacific under Representative Concentration Pathway flood risk in Mexico due to climate change, several (RCP) 4.5 (Knutson and others, 2015). Both basins, studies have found that the risk may decrease in some however, are expected to experience an increase in regions, including in Tabasco and other southeastern the frequency of tropical cyclones of at least Saffir- states (Haer and others, 2018, Hirabayashi and Simpson Category 4 intensity (Knutson and others, others, 2013, Hirabayashi and others, 2021). There 2015). Results at a sub-basin scale vary between is substantial uncertainty in flood risk projections for studies. Focusing on landfalling tropical cyclones in Mexico (Alfieri and others, 2017). the Yucatan Peninsula, Appendini and others (2019) estimated that intense hurricanes would be more 21 - River floods occur when water in a river, lake or other waterbody overflows onto adjacent land. Pluvial floods occur when extreme rainfall results in inundation independent of an overflowing waterbody. 22 - SSP5-8.5 represents an extreme climate change scenario, corresponding to Shared Socioeconomic Pathway (SSP) 5 with a total radiative forcing level by 2100 (cumulative measure of greenhouse gas emissions from all sources) of 8.5W/m2. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 31 8 >>> Appendix II. Description of individual event scenarios Scenario TC1 is defined as a North Atlantic tropical cyclone event with two landfalls in Mexico. The first landfall is on the Yucatan Peninsula as a Category 4 hurricane. The second landfall is in Tamaulipas in northeast Mexico, with lower wind speeds (Tropical Storm / Category 1 hurricane) but heavy rainfall impacting Nuevo León, including Monterrey. The initial landfall may be compared with other hurricanes impacting the Yucatan Peninsula, including Janet 1955, Gilbert 1988, Wilma 2005, and Dean 2007 (see Appendix II for map showing these events). The second landfall may be compared with other tropical cyclones that made landfall in northeast Mexico associated with Monterrey flooding, including the 1909 Monterrey Hurricane, Hurricane Beulah 1967, Hurricane Gilbert 1988, and Hurricane Alex 2010. Scenario TC2 is defined as two tropical cyclones impacting the state of Veracruz 10 days apart with heavy precipitation. This scenario is based on a similar sequence of tropical cyclones that impacted Veracruz in 2010 (Hurricane Karl and Tropical Storm Matthew), which resulted in substantial damages, mainly due to flooding caused by heavy precipitation. Scenario TC3 is defined as an Eastern Pacific tropical cyclone making landfall at Manzanillo, Colima as a Category 4 hurricane and impacting Guadalajara, Jalisco as a Category 1-2 hurricane.23 The event may be compared with other Category 4+ hurricanes impacting the Jalisco/Colima region of Mexico, including Hurricane Patricia 2015. However, the estimated losses from this event are higher than those of Hurricane Patricia, since Hurricane Patricia did not directly hit the exposure concentrations of Manzanillo and Guadalajara (AIR, 2015). Scenario F1 is defined as a river flood event in Tabasco and surrounding states. This scenario is based on several historical floods that have impacted the region, including the 2007 Tabasco floods. Key factors that contributed to this flood event include heavy rains associated with the combination of a cold front and tropical storm Noel and the release of water from the Peñitas dam. This event was estimated to have a 1-in-100-year return period in the region (Ramos and others, 2009). Tabasco is an area of concern for flooding, having experienced frequent flood events over the past decades (Haer and others, 2017), including floods in 1993, 1999, 2007, 2008, 2009, 2010, 2011 and 2020. 23 - This event was selected from the STORM global synthetic tropical cyclone hazard dataset (Bloemendaal and others, 2020), which consists of 10,000 years of synthetic tropical cyclone tracks, generated using historical data from IBTrACS. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 32 Scenario ES1 is defined as an extreme season Scenario TC1), Hurricane Karl and Tropical Storm comprised of a sequence of several extreme Matthew (similar events for Scenario TC2), and tropical cyclone and flood events, namely TC1, flooding in Tabasco (albeit much less severe than TC2, TC3, and F1, all occurring within a few in 2007, a similar event for Scenario F1) occurred. months of each other during one season. The While none of these events individually were the extreme season scenario was included to analyze most expensive disaster to impact Mexico historically the potential for substantial economic and financial (e.g., the 2017 earthquake resulted in economic impacts if Mexico is impacted by a series of extreme losses almost double the costliest 2010 event), in the events affecting different parts of the country in close aggregate, the economic losses from the sequence succession. Such seasons with a series of extreme of disasters in 2010 were higher than any other year events have occurred historically, as illustrated by an in the period 2000-2020, based on CENAPRED data. analysis of event frequency (Table 2). While multiple This highlights the importance of such sequences of years have had high numbers of events (e.g., 2001, events for the risk profile of Mexico, though the total 2005, 2009, 2010, 2011, 2014, 2021), these events economic losses from the 2010 sequence of events vary in severity. A recent example of a year with were still only estimated at less than one percent of multiple severe events is 2010, when Hurricane Alex GDP, due in part to the size of Mexico’s economy. (one of the similar events for the second landfall of >>> TABLE 2. - Historical Frequency of Tropical Cyclone and Flood Events in Mexico1 2000 2000 2002 2003 2004 2005 2006 2007 2008 2009 2010 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2011 Mean Flash flood 1 1 1 1 1 0 0 0 1 0 0 0 0 0 0 0 3 0 1 0 0 2 0.55 Riverine flood 0 0 1 1 1 2 1 1 0 5 2 3 0 1 0 1 0 1 0 0 0 0 0.91 Tropical cyclone 1 5 3 2 0 4 3 4 1 2 4 5 2 4 6 1 2 4 2 3 5 5 3.09 Total 2 6 5 4 2 6 4 5 2 7 6 8 2 5 6 2 5 5 3 3 5 7 4.55 Note: 1/ Number of events meeting EMDAT event criteria 2000-2021. At least one of the following criteria must be met for an event to be included in the EMDAT database: (i) 10 or more deaths; (ii) 100 or more people affected; or (iii) declaration of state of emergency or appeal for international assistance. Source: author analysis of EMDAT data (CRED, 2022). FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 33 >>> References Adrian, Tobias, Pierpaolo Grippa, Marco Gross, Vikram Haksar, Ivo Krznar, Sujan Lamichhane, Caterina Lepore, Fabian Lipinsky, Hiroko Oura, and Apostolos Panagiotopoulos. 2022. Approaches to Climate Risk Analysis in FSAPs. IMF Staff Climate Note 2022/005. Aguilar-Gomez, S., Gutierrez, E., Heres, D., Jaume, D. and Tobal, M., 2022. Thermal Stress and Financial Distress: Extreme Temperatures and Firms’ Loan Defaults in Mexico. Available at SSRN: https://ssrn.com/abstract=3934688 or http:// dx.doi.org/10.2139/ssrn.3934688 AIR, 2015. Hurricane Patricia makes landfall, Mexico escapes major damage. Boston. https://www.air-worldwide.com/news-and-events/press-releases/hurricane- patricia-makes-landfall-mexico-escapes-major-damage/ AIR, 2018. The Air Tropical Cyclone Model for Mexico. https://www.airworldwide. com/SiteAssets/Publications/Brochures/documents/AIR-Tropical-Cyclone-Model- for-Mexico Alfieri, L., Bisselink, B., Dottori, F., Naumann, G., de Roo, A., Salamon, P., Wyser, K. and Feyen, L., 2017. Global projections of river flood risk in a warmer world. Earth’s Future, 5(2), pp.171-182. Appendini, C.M., Meza-Padilla, R., Abud-Russell, S., Proust, S., Barrios, R.E. and Secaira-Fajardo, F., 2019. Effect of climate change over landfalling hurricanes at the Yucatan Peninsula. Climatic Change, 157(3), pp.469-482. Aznar-Siguan, G. and Bresch, D. N., 2019. CLIMADA v1: a global weather and climate risk assessment platform, Geosci. Model Dev., 12, 3085–3097, https://doi. org/10.5194/gmd-12-3085-2019 Banco Central do Brasil (2022). Financial Stability Report. Vol. 21, no. 2. ISSN 2176-8102. https://www.bcb.gov.br/content/publications/ financialstabilityreport/202211/fsrFullRep.pdf Bank for International Settlements. (2021). Climate-related risk drivers and their transmission channels. https://www.bis.org/bcbs/publ/d517.pdf Bank of England (2021). Guidance for participants of the 2021 Biennial Exploratory Scenario: Financial risks from climate change. https://www.bankofengland.co.uk/-/ media/boe/files/stress-testing/2021/the-2021-biennial-exploratory-scenario-on-the- financial-risks-from-climate-change.pdf Bank of Japan (2022). Pilot scenario analysis exercise of climate-related risks based on common scenarios. https://www.fsa.go.jp/en/news/2022/20220826/03. pdf FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 34 Bevacqua, E., De Michele, C., Manning, C., Couasnon, A., Ribeiro, A.F., Ramos, A.M., Vignotto, E., Bastos, A., Blesić, S., Durante, F. and Hillier, J. (2021). Guidelines for studying diverse types of compound weather and climate events. Earth’s future, 9(11), p. e2021EF002340. Bloemendaal, N., de Moel, H., Martinez, A.B., Muis, S., Haigh, I.D., van der Wiel, K., Haarsma, R.J., Ward, P.J., Roberts, M.J., Dullaart, J.C. and Aerts, J.C., 2022. A globally consistent local-scale assessment of future tropical cyclone risk. Science advances, 8(17), p. eabm8438. Bloemendaal, N., Haigh, I.D., de Moel, H. and others, 2020. Generation of a global synthetic tropical cyclone hazard dataset using STORM. Sci Data 7, 40. https://doi. org/10.1038/s41597-020-0381-2 Bos, J., Li, R., & Sanders, M. (2018). Hazardous Lending: The Impact of Natural Disasters on Banks’ Asset Portfolio. Maastricht University, Graduate School of Business and Economics. GSBE Research Memoranda No. 021 https://doi. org/10.26481/umagsb.2018021 Breinl, K., Lun, D., Müller-Thomy, H. and Blöschl, G., 2021. Understanding the relationship between rainfall and flood probabilities through combined intensity- duration-frequency analysis. Journal of Hydrology, 602, p.126759. Bressan, G., Duranovic, A., Monasterolo, I. and Battiston, S., 2023. Asset-level climate physical risk assessment and cascading financial losses. Available at SSRN. https://dx.doi.org/10.2139/ssrn.4062275 Bruyère, C., Rasmussen, R., Gutmann, E., Done, J., Tye, M., Jaye, A., Prein, A., Mooney, P., Ge, M., Fredrick, S., 2017. Impact of climate change on Gulf of Mexico hurricanes. NCAR Tech Note. NCAR/TN535 Buchhorn, M., Smets, B., Bertels, L., De Roo, B., Lesiv, M., Tsendbazar, N.-E., Herold, M., Fritz, S., 2020. Copernicus Global Land Service: Land Cover 100m: collection 3: epoch 2019: Globe. Calice, P., Miguel, F. 2021. Climate-Related and Environmental Risks for the Banking Sector in Latin America and the Caribbean: A Preliminary Assessment. Policy Research Working Paper; No. 9694. World Bank, Washington, DC. http:// hdl.handle.net/10986/35764 Carozza, D.A., Boudreault, M. (2021). Diversification of Atmospheric Perils. AXA XL. https://axaxl.com/-/media/axaxl/files/pdfs/campaign/climate-risk/section-2/ mktresilience_4diversification-of-atmospheric-perils_axa-xl_re.pdf CENAPRED, 2006. “Guía Básica para la Elaboración de Atlas Estatales y Municipales de Peligros y Riesgos : Fenomenos Hidrometeorologicos”. https:// www.cenapred.unam.mx/es/Publicaciones/archivos/63.pdf Centre for Research on the Epidemiology of Disasters (CRED), 2022. EM-DAT Database, Université Catholique de Louvain, www.emdat.be Danmarks Nationalbank (2021). Flood risk can potentially affect a large share of credit institutions’ exposures. https://www.nationalbanken.dk/en/publications/ Documents/2021/06/ANALYSIS%20No.%2013%20-%20Flood%20risk%20 can%20potentially%20affect%20a%20large%20share%20of%20credit%20 institutions%E2%80%99%20exposures%20.pdf Dieppe, A., Kilic Celik, S., and Okou, C. (2020). Implications of major adverse events on productivity. Working Paper 9411, The World Bank. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 35 Dominguez, C., Jaramillo, A. and Cuéllar, P. (2021). Are the socioeconomic impacts associated with tropical cyclones in Mexico exacerbated by local vulnerability and ENSO conditions? International Journal of Climatology, 41, pp. E3307-E3324. Dunz, N., Essenfelder, A.H., Mazzocchetti, A., Monasterolo, I. and Raberto, M., 2021. Compounding COVID-19 and climate risks: the interplay of banks’ lending and government’s policy in the shock recovery. Journal of Banking & Finance, p.106306. Eberenz, S., Lüthi, S., and Bresch, D. N., 2021. Regional tropical cyclone impact functions for globally consistent risk assessments, Nat. Hazards Earth Syst. Sci., 21, 393–415, https://doi.org/10.5194/nhess-21-393-2021 European Central Bank. (2021). Climate-related risks to financial stability. https://www.ecb.europa.eu/pub/financial-stability/fsr/special/html/ecb. fsrart202105_02~d05518fc6b.en.html Fernández-Díaz, V.Z., Canul Turriza, R.A., Kuc Castilla, A. and Hinojosa-Huerta, O., 2022. Loss of coastal ecosystem services in Mexico: An approach to economic valuation in the face of sea level rise. Frontiers in Marine Science, p.1347. Financial Stability Board (FSB), 2020. The Implications of Climate Change for Financial Stability. https://www.fsb.org/wp-content/uploads/P231120.pdf Financial Stability Board. (2020). The implications of climate change for financial stability. https://www.fsb.org/wp-content/uploads/P231120.pdf Haer, T., Botzen, W. W., Zavala-Hidalgo, J., Cusell, C., & Ward, P. J. (2017). Economic evaluation of climate risk adaptation strategies: Cost-benefit analysis of flood protection in Tabasco, Mexico. Atmósfera, 30(2), 101-120. Haer, T., Botzen, W.W., Van Roomen, V., Connor, H., Zavala-Hidalgo, J., Eilander, D.M. and Ward, P.J., 2018. Coastal and river flood risk analyses for guiding economically optimal flood adaptation policies: a country-scale study for Mexico. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 376(2121), p.20170329. Hallegatte, S., 2015. The indirect cost of natural disasters and an economic definition of macroeconomic resilience. World Bank Policy Research Working Paper, (7357). Hallegatte, S., Jooste, C., Mcisaac, F.J. 2022b. “Macroeconomic Consequences of Natural Disasters: A Modeling Proposal and Application to Floods and Earthquakes in Turkey”. Washington, DC: World Bank. Hallegatte, S., Lipinsky, F., Morales, P., Oura, H., Ranger, N., Gert Jan Regelink, M., & Reinders, H. J. (2022a). Bank Stress Testing of Physical Risks under Climate Change Macro Scenarios: Typhoon Risks to the Philippines. IMF Working Paper No. 2022/163. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4210160 Haug, K.J., Reiakvam, L.K., Solheim, H., Turtveit, L.-T., Vatne, B.H. (2021). Cliamte risk and banks’ loans to firms. Norges Bank. https://www.norges-bank. no/contentassets/244023305b474ca4a7fc4f82d766b46f/staff-memo-7-2021_ en.pdf?v=12/17/2021133156&ft=.pdf Hirabayashi, Y., Mahendran, R., Koirala, S., Konoshima, L., Yamazaki, D., Watanabe, S., Kim, H. and Hirabayashi, Y., Tanoue, M., Sasaki, O., Zhou, X. and Yamazaki, D., 2021. Global exposure to flooding from the new CMIP6 climate model projections. Scientific reports, 11(1), pp.1-7. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 36 Holland, G., 2008. A Revised Hurricane Pressure–Wind Model, Mon. Weather Rev., 136(9), 3432–3445, doi:10.1175/2008MWR2395.1. Homer, D., & Li, Ming. (2017). Notes on using property catastrophe model results. Casualty Actuarial Society. https://www.casact.org/sites/default/files/2021-02/2017_ most-practical-paper_homer-li.pdf Huizinga, J., De Moel, H. and Szewczyk, W., 2017. Global flood depth-damage functions: Methodology and the database with guidelines, EUR 28552 EN, Publications Office of the European Union, Luxembourg, ISBN 978-92-79-67781- 6, doi:10.2760/16510, JRC105688. Instituto Nacional de Estadística y Geografía (INEGI), 2013. Capital stock by state. https://en.www.inegi.org.mx/programas/acervos/2013/ Intergovernmental Panel on Climate Change (2021) Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [Masson-Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K. Maycock, T. Waterfield, O. Yelekçi, R. Yu, and B. Zhou (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 2391 pp. doi:10.1017/9781009157896. p.1520. International Monetary Fund (2021) Dominica: Disaster Resilience Strategy. Country Report No. 2021/182. Isela, A. and Zarco, A., 2014. Memoria de trabajo en la Gerencia de Aguas Superficiales e Ingeniería de Ríos, CONAGUA. Kaczmarska, J., Jewson, S. & Bellone, E., 2018. Quantifying the sources of simulation uncertainty in natural catastrophe models. Stoch Environ Res Risk Assess 32, 591–605. https://doi.org/10.1007/s00477-017-1393-0 Kanae, S., 2013. Global flood risk under climate change. Nature climate change, 3(9), pp.816-821. Knutson, T., Camargo, S.J., Chan, J.C., Emanuel, K., Ho, C.H., Kossin, J., Mohapatra, M., Satoh, M., Sugi, M., Walsh, K. and Wu, L., 2020. Tropical cyclones and climate change assessment: Part II: Projected response to anthropogenic warming. Bulletin of the American Meteorological Society, 101(3), pp. E303-E322. Knutson, T.R., Sirutis, J.J., Zhao, M., Tuleya, R.E., Bender, M., Vecchi, G.A., Villarini, G. and Chavas, D., 2015. Global projections of intense tropical cyclone activity for the late twenty-first century from dynamical downscaling of CMIP5/ RCP4. 5 scenarios. Journal of Climate, 28(18), pp.7203-7224. Kopp, R., Easterling, D.R., Hall, T., Hayhoe, K., Horton, R., Kunkel, K. and LeGrande, A., 2017. Potential surprises–compound extremes and tipping elements. Publications, Agencies and Staff of the U.S. Department of Commerce. 578. Magaña, V.O., Vázquez, J.L., Pérez, J.L. and Pérez, J.B., 2003. Impact of El Niño on precipitation in Mexico. Geofísica internacional, 42(3), pp.313-330. Martínez-Jaramillo, S. and Montañez-Enríquez, R. (2021). Dependencies and impact of the Mexican banking sector on ecosystem services. Unpublished NGFS- INSPIRE Study Group Input Paper. FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 37 Mora, R.D., Elizondo, E.C., Juarez, M.L.A. and Garduno, G.E., 2008. Análisis estadístico de los gastos máximos anuales registrados en las cuencas del bajo grijalva y de las descargas de la presa peñitas. https://www.gob.mx/conagua/ documentos/plan-hidrico-integral-de-tabasco-phit Mumby, P.J., Vitolo, R. and Stephenson, D.B. (2011). Temporal clustering of tropical cyclones and its ecosystem impacts. Proceedings of the National Academy of Sciences, 108(43), pp.17626-17630. Neri-Flores, I., Moreno-Casasola, P., Peralta-Peláez, L.A. and Monroy, R., 2019. Groundwater and river flooding: the importance of wetlands in coastal zones. Journal of Coastal Research, 92(SI), pp.44-54. NGFS, 2022. Climate scenario analysis by jurisdictions: Initial findings and lessons. https://www.ngfs.net/sites/default/files/medias/documents/climate_scenario_ analysis_by_jurisdictions_initial_findings_and_lessons.pdf Pesaresi, M., Florczyk, A., Schiavina, M., Melchiorri, M., Maffenini, L., 2019. GHS settlement grid, updated and refined REGIO model 2014 in application to GHS-BUILT R2018A and GHS-POP R2019A, multitemporal (1975-1990-2000- 2015), R2019A. European Commission, Joint Research Centre (JRC) DOI: 10.2905/42E8BE89-54FF-464E-BE7B-BF9E64DA5218 PID: http://data.europa. eu/89h/42e8be89-54ff-464e-be7b-bf9e64da5218 Pescaroli, G. and Alexander, D., 2018. Understanding compound, interconnected, interacting, and cascading risks: a holistic framework. Risk analysis, 38(11), pp.2245-2257. Ramos, J., Marrufo, L., Gonzalez, F.J. (2009) ‘Use of Lidar Data in Floodplain Risk Management Planning: The Experience of Tabasco 2007 Flood’, in G. Jedlovec (ed.), Advances in Geoscience and Remote Sensing, IntechOpen, London. 10.5772/8322. Ranger, N., Mahul, O. and Monasterolo, I., 2021. Managing the financial risks of climate change and pandemics: What we know (and don’t know). One Earth, 4(10), pp.1375-1385. Ranger, N., Mahul, O. and Monasterolo, I., 2022. Assessing Financial Risks from Physical Climate Shocks: A Framework for Scenario Generation (English). Equitable Growth, Finance and Institutions Insight Washington, D.C.: World Bank Group. http://documents.worldbank.org/curated/en/760481644944260441/ Assessing-Financial-Risks-from-Physical-Climate-Shocks-A-Framework-for- Scenario-Generation Regelink, M., Reinders, H.J., Vleeschhouwer, M., van de Wiel, I. (2017). Waterproof? An exploration of climate-related risks for the Dutch financial sector. De Nederlandsche Bank. https://www.dnb.nl/media/r40dgfap/waterproof-an- exploration-of-climate-related-risks-for-the-dutch-financial-sector.pdf Reinders, H.J., Regelink, M., Calice, P., Uribe, M.E. (2021) Not-So-Magical Realism: A Climate Stress Test of the Colombian Banking System (English). Equitable Growth, Finance and Institutions Insight Washington, D.C.: World Bank Group. http://documents.worldbank.org/curated/en/957831635911537578/Not-So- Magical-Realism-A-Climate-Stress-Test-of-the-Colombian-Banking-System Romero, D., Salinas, J.C., Talledo, J. (2022) Climate risk stress test: Impact of climate change on the Peruvian financial system. Superintendencia de Banca, Seguros y Administradoras Privadas de Fondos de Pensiones. https://www.sbs. gob.pe/Portals/0/jer/DDT_ANO2022/DT%2003%202022.pdf FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 38 Sampson, C. C., Smith, A. M., Bates, P. D., Neal, J. C., Alfieri, L., and Freer, J. E., 2015. A high-resolution global flood hazard model, Water Resour. Res., 51, 7358–7381, https://doi.org/10.1002/2015WR016954 Schiavina, M., Freire, S., MacManus, K., 2019. GHS population grid multitemporal (1975, 1990, 2000, 2015) R2019A. European Commission, Joint Research Centre (JRC) DOI: 10.2905/42E8BE89-54FF-464E-BE7B-BF9E64DA5218 PID: http:// data.europa.eu/89h/0c6b9751-a71f-4062-830b-43c9f432370f Stahle, D. W., Cook, E. R., Burnette, D. J., Villanueva, J., Cerano, J., Burns, J. N., Griffin, D., Cook, B.I., Acuna, R., Torbenson, M.C.A., Sjezner, P., & Howard, I. M. (2016). The Mexican Drought Atlas: Tree-ring reconstructions of the soil moisture balance during the late pre-Hispanic, colonial, and modern eras. Quaternary Science Reviews, 149, 34-60. Towe, R., Tawn, J., Eastoe, E. and Lamb, R. (2020). Modelling the clustering of extreme events for short-term risk assessment. Journal of Agricultural, Biological and Environmental Statistics, 25(1), pp.32-53. Trendafiloski, G., Podlaha, A., Ewing, C., 2017. Understanding and managing damage uncertainty in catastrophe models. Impact Forecasting. https://oasislmf. org/application/files/8515/1030/3324/OASIS_IF.pdf UNISDR, 2015. Making Development Sustainable: The Future of Disaster Risk Management. Global Assessment Report on Disaster Risk Reduction. Geneva, Switzerland: United Nations Office for Disaster Risk Reduction (UNISDR). Villarini, G., Vecchi, G.A. and Smith, J.A. (2010). Modeling the dependence of tropical storm counts in the North Atlantic basin on climate indices. Monthly Weather Review, 138(7), pp.2681-2705. Vitek, F. 2018. The Global Macrofinancial Model. IMF working paper Working 2018/081. Wolbers, J., Kuipers, S. and Boin, A., 2021. A systematic review of 20 years of crisis and disaster research: Trends and progress. Risk, Hazards & Crisis in Public Policy, 12(4), pp.374-392. World Bank (WB), 2019. Mexico - Systematic Country Diagnostic (English). Washington, D.C.: World Bank Group. http://documents.worldbank.org/curated/ en/588351544812277321/Mexico-Systematic-Country-Diagnostic World Bank (WB), 2021a. World Bank Climate Change Knowledge Portal. https:// climateknowledgeportal.worldbank.org/country/mexico/extremes World Bank (WB), 2021b. World Bank Disaster Risk Financing and Insurance Program – Financial Risk Assessment Tool. https://www.financialprotectionforum. org/disaster-risk-financing-drf-analytics-tools World Bank (WB), 2022. Inflation, consumer prices. Mexico. World Development Indicators. Xi, D., Lin, N. and Gori, A., 2023. Increasing sequential tropical cyclone hazards along the US East and Gulf coasts. Nature Climate Change, pp.1-8. Zscheischler, J., Westra, S., Van Den Hurk, B. J. J. M., Seneviratne, S. I., Ward, P. J., Pitman, A., Aghakouchak, A., Bresch, D. N., Leonard, M., Wahl, T., & Zhang, X. (2018). Future climate risk from compound events. Nature Climate Change, 8(6), 469-477. https://doi.org/10.1038/s41558-018-0156-3 FROM EXTREME EVENTS TO EXTREME SEASONS: FINANCIAL STABILITY RISKS OF CLIMATE CHANGE IN MEXICO <<< 39