FINANCE FINANCE, COMPETITIVENESS & INNOVATION EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT Not-so-magical realism: A climate stress test of the Colombian banking system © 2021 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved. This work is a product of the staff of The World Bank 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. The World Bank does not guarantee the accuracy of the data included in this work. 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 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 This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO), http://creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: Attribution—Please cite the work as follows: 2021. Not-so-magical realism: A climate stress test of the Colombian banking system. EFI Insight-Finance. Washington, DC: World Bank. Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third- party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to reuse a component of the work, it is your responsibility to determine whether permission is needed for that reuse and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; e-mail: pubrights@worldbank.org. Cover photography: Santiago La Rotta Cover design and typesetting: Diego Catto / www.diegocatto.com >>> Acknowledgments Report prepared by Henk Jan Reinders, Martijn Regelink, Pietro Calice, and Mariana Escobar Uribe. We are grateful to the Superintendencia Financiera de Colombia and the Banco de la República for fruitful discussions and for providing the data needed to perform the analyses. We are furthermore grateful to the Departamento Nacional de Planeación for providing estimates of the economic effects of abrupt climate policy scenarios and to the Instituto de Hidrología, Meteorología y Estudios Ambientales for sharing data on climate change and flood risk. We also thank three peer reviewers for their comments and suggestions: Stephane Hallegatte (World Bank), Guan Schellekens (European Central Bank), and Edo Schets (Bank of England). The vulnerability modeling was developed with input from Nepomuk Dunz, Faruk Miguel Liriano, Oliver Masetti, Carolina Rogelis, Hugo Alexander Rojas Romagosa, Jose Angel Villalobos, and Dieter Wang. >>> Contents Acronyms 5 Executive Summary 6 1. Climate Change and Financial Risks in Colombia 12 Natural disasters and climate change 12 The Colombian banking sector 14 Main climate-related risks to the banking sector 17 Cross-border risks 20 2. Physical risk assessment 21 Flood risk 22 Financial impact 25 Banking sector stress 26 3. Transition risk assessment 30 Transition risk scenarios 31 Macroeconomic impact 33 Banking sector stress 34 4. Conclusions and policy recommendations 37 Main risks and vulnerabilities 37 Financial impact 38 References and Suggested Readings 40 Appendixes 43 Appendix A. Flood regression model 43 Appendix B. Flood vulnerability 45 Appendix C. Stress test model 46 Appendix D. DNP CGE model results 47 >>> Acronyms BCBS Basel Committee on Banking Supervision BR Central Bank of Colombia CAR capital adequacy ratio CEPAL Economic Commission for Latin America and the Caribbean CGE computable general equilibrium Col$ Colombian peso DANE National Administrative Department of Statistics DID difference-in-difference DNB Dutch Central Bank DNP National Planning Department ESG environmental, social, and governance ETS emission trading system FSB Financial Stability Board G20 Group of Twenty IDEAM Institute of Hydrology, Meteorology and Environmental Studies IMF International Monetary Fund IPCC Intergovernmental Panel on Climate Change GDP gross domestic product GHG greenhouse gas LAC Latin America and the Caribbean MT medium term NDC nationally determined contribution NGFS Network of Central Banks and Supervisors for Greening the Financial System NPL nonperforming loan RCP representative concentration pathway RP return period S&P Standard & Poor’s SDG Sustainable Development Goal SFC Financial Superintendence of Colombia SSP shared socioeconomic pathway ST short term US$ United States dollar WRI World Resources Institute EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 5 >>> Executive Summary Losses associated with the materialization of climate-related risks are increasingly affecting economies and financial sectors globally, and Colombia is no exception. The Colombian bank- ing sector is hence potentially exposed to climate-related financial risks, particularly from floods and strong climate mitigation targets. Large scale riverine floods are the main climate-related disaster risk (that is, physical risk) in Colombia. In the past few decades, such events have led to major damages to real estate and other capital goods, with the 2010 and 2011 floods leading to combined damages of US$7.0 billion (equivalent to 2.0 percent of the 2011 gross domestic product [GDP]).1 Approximately 6.5 percent of banks’ total loan exposures are in municipalities that face high flood risk.2 Moreover, the flood hazard is expected to increase between 25 and 65 percent between 1980 and 2080 because of climate change. In general, only a minor frac- tion (between 2 to 4 percent) of economic damages after disaster events are insured, leaving a large share of the burden on government, households, and firms, thereby increasing the credit risk on banks that finance them. Furthermore, to prevent further climate change, a broad-based change in the Colombian economy would be needed to achieve its greenhouse gas (GHG) emission reduction targets. Recently, the Colombian government increased its GHG emission reduction targets from 20 percent to 51 percent by 2030, constituting one of the most ambitious targets in the Latin America and the Caribbean (LAC) region. About 20 percent of Colombian banks’ corporate loans are in sectors that are highly sensitive to transitions, whereas a broader set of sectors and assets is vulnerable through value-chain effects (that is, links between sec- tors). Besides domestic climate policies, transition risks also originate from climate regulation abroad, technological change, and shifting consumer preferences. Colombia is also exposed to transition risks from abroad, due to relatively high fossil fuel export revenues and a high carbon intensity of manufacturing exports compared to other countries in the region. This report identifies and assesses climate-related risks in the banking sector and develops two innovative approaches to conduct basic climate risk stress tests in emerging markets. For physi- cal risks, we develop a stress test at the municipal level to investigate the vulnerability of banks toward severe riverine floods. We introduce three innovations (a) to model climate risk in absence of nationwide probabilistic disaster scenarios; (b) to estimate the effects of flood-related economic damages on banks using spatial panel data on loan provisions; and (c) to extend a basic stress test model with spatially disaggregated credit risk and sovereign credit risk channels. These in- novations allow us to make first estimates on the impacts of floods on banks’ profitability and solvency. For transition risks, we develop a stress test at the two-digit sectoral level using a locally available computable general equilibrium (CGE) model to estimate the effects of severe but plau- 1 Figures for flood damages and insurance penetration based on EM-DAT. 2 High flood risk based on more than 10 percent of the land area flooded during flood events in the past. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 6 sible transition scenarios on value added in 83 economic sec- align the banking sector’s risk management and capital alloca- tors. This approach allows us to account for interdependencies tion with the national NDC targets. between sectors as well as for the partial compensation of firms through increased government revenues.3 Transition scenarios Results show differences in climate-related vulnerabilities in are tailored to Colombia and in line with the recently updated credit portfolios between banks, underscoring the importance GHG reduction targets as part of the Colombian nationally de- of risk-based supervision. Three banks are substantially (about termined contribution (NDC). We finally estimate the impact on two to three times) more vulnerable to flood hazards than most banks, through non-performing loans, using a credit risk model others owing to high exposures in the more rural areas or rela- of the Central Bank of Colombia (BR). tively large sovereign exposures. Also, there is significant het- erogeneity in the exposure of individual banks to sectors that Our flood risk assessment shows that severe flood scenarios are highly sensitive to transitions, ranging between 1 and 26 can lead to declines in capital adequacy that could moreover percent of their credit portfolios. These banks could suffer rela- coincide with other shocks. We look at riverine floods occur- tively high losses in the broad set of potentially affected sec- ring owing to heavy rainfall, which in the most extreme cases tors compared with their total assets. They also have relatively has historically been related to La Niña.4 We investigate three high loan concentrations in the sectors that are most vulner- flood scenarios, one based on the 2010 and 2011 floods re- able in our transition scenarios: fossil fuels, waste collection, lated to La Niña and two more severe floods with return peri- agriculture, and electricity supply. These observations support ods of once in 500 years. For those three scenarios we find a risk-based approach to addressing climate-related risks in mi- an average decline in the capital adequacy ratio (CAR) for croprudential supervision while also noting that the banks with Colombian banks between 0.3 and 1.1 percentage points. A the highest vulnerability potentially have a high potential to con- fourth scenario, investigating a severe flood coinciding with a tribute to greening the Colombian economy by stimulating GHG recession, finds an average decline in the CAR of 3.2 percent- mitigation in these sectors. This contribution could, for example, age points. Results differ strongly per bank, with loan losses be achieved by engaging with current clients and by including for individual banks ranging between 0.2 percent of total as- green considerations into a bank’s origination practices or by sets for the least vulnerable bank to 2.2 percent for the most increasing the offering of green products (for example, green vulnerable one in the most severe flood scenario. Finally, we corporate loans and green mortgages). find that a worst-case climate change scenario (RCP 8.5) could add an additional 0.1 to 0.6 percentage points in CAR Our analysis is explorative and often based on aggregated impact—per severe flood event—compared with a scenario data, hence the results of our assessment should be inter- with limited climate change (RCP 2.6) depending on the bank.5 preted with caution. Several limitations in data and model availability require us to make assumptions that lead to po- Our transition risk assessment shows that severe decarbon- tential over- or underestimation of outcomes. For example, ization scenarios can lead to substantial losses in the banking we improve on macroeconomic approaches by disaggregat- sector. However, the most severe scenarios can be avoided ing exposures into sectors and municipalities but do not have by managing the transition well. Relevant decarbonization detailed information at firm and household level. Limitations scenarios can already materialize in the medium run given are described throughout the report and should guide the in- Colombia’s high GHG reduction targets for 2030. We estimate terpretation of results. Specifically, we note that estimates for that in an adverse scenario, with a high GHG reduction tar- the effects of transition risks on value added per sector are get and delayed implementation of policies, aggregated loan obtained under adverse or stressed conditions and do not rep- losses for Colombian banks may range between 0.2 percent resent a forecast of any kind. The employed scenario does not of total assets for the least vulnerable banks to 2.7 percent account for all potential adaptive processes, including techno- for the most vulnerable ones. This scenario could materialize logical change, and hence may overestimate effects on the when climate policies are introduced late (from 2026) and no economy—especially in cases where the economy has more other measures are taken to allow the Colombian economy to time to adjust, that is, the “smooth” transition scenarios. When timely adapt. The estimates may be conservative since they climate policies are introduced gradually, the economy is ex- cover a two-year time frame and consider only credit risk, with pected to adjust and adverse effects are more likely limited more losses that could accumulate before and after the shock than estimated in this study. Given the explorative nature of period and through other risk channels. The possibility of sub- the stress testing, results are not intended to identify capital stantial losses warrants action to be taken in the short term to shortfalls or to provide pass or fail outcomes for banks. 3 Hence, we are not only looking at direct emissions from owned sources (scope 1 emissions), but also at emissions from the generation of purchased electricity and other emissions that occur in the value chain (scope 2 and 3 emissions), and we account for revenue recycling to alleviate macroeconomic effects. 4 La Niña refers to a recurring weather phenomenon in the Pacific Ocean that causes heavy rainfall, flooding, and landslides in Colombia. 5 A Representative Concentration Pathway (RCP) is a greenhouse gas concentration trajectory adopted by the Intergovernmental Panel on Climate Change (IPCC). EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 7 > > World Bank Recommendations: This report is a collaboration between the World Bank and the Financial Superintendence of Colombia (SFC). Based on this report, the World Bank has identified several short-term (ST) and medium-term (MT) actions that the SFC can consider to improve climate risk identification and mitigation in the banking sector. These actions include that the SFC could adopt risk-based supervision for climate-related risks and continuously improve information disclosures (both by nonfinancial corporates and by financial institutions) and data availability. The latter is also important to assess expo- sures through affiliated entities abroad, for which the SFC could collaborate with host supervisors to conduct a cross- country climate vulnerability assessment. For physical risks, the SFC could perform deep dives at the most vulnerable institutions, promote capacity building throughout the sector, and encourage the further development of insurance markets since increased disaster insurance penetration mitigates risks for banks. Regarding transition risks, the SFC can work with the banking sector and other authorities to address risks in a timely and forward-looking manner. Such action requires dialogue between public and private stakeholders on how climate mitigation and adaptation measures will take shape, so that banks can address them early on (for example, in engagement with customers, origination practices, and pricing). It also requires capacity building in the banking sector, potentially including new tools such as scenario analysis and full bottom-up stress testing. This capacity building is particularly urgent with respect to domestic transition risks, as substantial efforts to decarbonize the Colombian economy can already be expected this decade. See table ES.1 for an overview of recommendations. > > > T A B L E E S . 1 . - World Bank recommendations to the SFC and other stakeholders Recommendation Timing Agency General Issue guidelines on governance, risk management, and climate risk disclosure to the banking sector. ST SFC Promote the development of forward-looking climate risk tools in the banking sector, including scenario ST/MT SFC analysis and stress testing. Incorporate climate risks as part of the risk-based supervision of banks. a MT SFC Continue to encourage climate risk disclosure by nonfinancial firms in Colombia, and support the MT SFC improvement and use of climate-risk data at the firm level. Conduct a cross-country climate vulnerability assessment with host supervisors, including improving data SFC, host MT collection on spatial and sectoral exposures through related entities. supervisors Physical risks Identify the information and data needed to carry out physical risk assessments tailored to individual ST/MT SFC institutions’ exposure. Promote technical capacity building in the banking sector to understand and manage physical risks, covering ST/MT SFC both disaster risks and gradual changes in climatic conditions (for example, through a platform). Promote the development and implementation of mechanisms to mitigate physical risks and their impacts, SFC, MT such as disaster risk insurance.b Government Transition risks Promote capacity building in the banking sector to understand and manage transition risks, specifically ST SFC focusing on ST and MT transition risks (that is, 2030 NDC targets). Promote further dialogue between climate policy makers and the financial sector including banks, other SFC, BR, ST/MT investors, and financial authorities. Government Provide more detailed guidance to the financial sector on how de-carbonization policies will be implemented Government, and on what is the time line for implementation (for example, a road map until 2030 to allow banks to better ST/MT SFC align their portfolios). Encourage better data collection to perform firm-level stress tests by authorities and banks. Specifically, this ST/MT SFC would require firm-level GHG emission data for nonfinancial firms in Colombia. Note: BR = Central Bank of Colombia; NDC = nationally determined contribution; SFC = Financial Superintendence of Colombia; ST = short term (within one year) and MT = medium term (within one to three years). a. This could include requiring increased supervisory reporting, addressing the risk in the banks’ Internal Capital Adequacy and Assessment Process, and addressing the risk in on-site supervision (including in board-level conversations). b. We note that sometimes the most cost-effective adaptation lies in preventing the creation of new risks by ensuring that land-use planning and infrastruc- ture regulations take disaster risks into account. Banks could play a role here by integrating disaster and flood risks into their loan origination process (and requiring alignment with land-use plans and infrastructure regulations). Development of insurance markets could include (a) strengthening the legal framework including on the use of parametric insurance and (b) improving the availability of disaster insurance data. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 8 >>> Introduction Globally, increasing attention is being paid to the effects of climate change and environmental risks and opportunities on financial sectors. Regulators and central banks—through the Network for Greening the Financial System (NGFS) among others—are warning on the effects of cli- mate change and environmental risks on the stability and soundness of financial sectors. These calls follow attention paid to this topic by the Financial Stability Board (FSB), its Task Force on Climate-Related Financial Disclosures, and the G20 (Group of Twenty) Green Finance Study Group. At the same time, there is global recognition of the importance of financial sectors in mobilizing capital for green objectives, including those related to the Paris Agreement and the Sustainable Development Goals (SDGs). Several global financial bodies have recently stepped up their work on climate-related financial risks, including reports by the Basel Committee on Banking Supervision (BCBS) and the FSB (BCBS 2021; FSB 2020). Sustainable finance is also one of the priorities of the Italian G20 presidency in 2021. Within the Colombian banking sector, there is an emerging awareness of climate-related risks, but the understanding of the risks and their management are still at an early stage.6 Although the industry has a broad awareness of risks related to climate and sustainability, relatively little attention is paid to the financial risks that climate change trends pose to the Colombian banking system. Some banks have a management system that takes into account environmental, social, and governance (ESG) risks, which is typically focused on limiting any negative effects of bank activities on ESG-related factors. The financial risk side is explored to a lesser extent, including the effects of climate on credit risk, market risk, and other financial indicators. Also, the financial sector, including a group of Colombian banks, has been seeking to develop strategies to sup- port sustainable development in Colombia through the Protocolo Verde Ampliado (the expanded Green Protocol). The Financial Superintendence of Colombia (SFC) identified the better understanding of climate- related financial risk as a priority in its 2019 action plan related to climate change (SFC 2019). The SFC action plan focuses on four key areas, including (a) taxonomy, (b) ESG integration, (c) transparency on climate risks, and (d) capacity building. This report contributes to the third and fourth areas by expanding the knowledge base to identify, assess, and manage climate-related financial risks in Colombia. It also provides a foundation for future, more detailed risk assess- ments and aligns with activities of the Colombian financial authorities who are increasingly in- volved in the NGFS and other international forums. 6 Based on interviews and an SFC survey amongst banks. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 9 This report identifies and assesses relevant physical and tran- folio, the effects on the market value of government bonds, sition risks with focus on the banking sector. Banks constitute and the effects on exposures through investments in other fi- the largest segment of the Colombian financial sector with as- nancial institutions. These three channels represent the most set holdings of Colombian peso (Col$) 720 trillion (US$209 bil- important asset classes, covering 79 percent of total assets in lion or 78 percent of gross domestic product [GDP]) in 2020.7 the Colombian banking sector. However, in some of our quan- This includes both foreign and domestic banks. Some banks titative assessments, we limit our scope further owing to data are parts of larger conglomerates in which the related enti- limitations, including investments in related entities and non- ties could be exposed to similar climate-related financial risks, corporate loans (the latter only for transition risk).9 Because such as foreign banks, insurance companies, and asset man- data are not available for all potential channels that affect the agers. Insurance companies and asset managers are, how- financial sector, our outcomes can be conservative and lead to ever, not part of this report. The scope of our analysis includes an underestimation of the total effect of climate risks on banks. both physical risks (that is, those emanating from weather-re- lated events and gradual changes in climatic conditions) and We investigate a set of scenarios specific to the Colombian transition risks (that is, those emanating from decarbonization context that are designed to investigate events in the tail of of the global economy in line with targets in the Paris Agree- the probability distribution. The scenarios that we use for the ment). We note that we use a broad definition of physical risks, analysis investigate events that are specifically relevant in covering both climate-related disaster risks and the effects of Colombia, including a relatively high greenhouse gas (GHG) climate change on their probability distribution. In the remain- emission reduction target in the medium term (that is, by der of the report, we refer to the collection of these risks as 2030) and high and increasing flood risk. For both transition climate-related financial risks, or climate-related risks in short. and physical risks, we investigate orderly scenarios but also events that are less likely to occur but that would potentially The report also builds on quantitative data from a range of cause more stress to the Colombian banking sector. Looking sources to explore the vulnerability of banks in specific sce- at such severe but plausible scenarios is commensurate with narios. The report bases its analysis on data provided by the common practice in analysis of financial sector scenarios and SFC, the Central Bank of Colombia (BR), the National Plan- stress testing but should not be interpreted as investigating ning Department (DNP), and the Institute of Hydrology, Me- the most likely outcome. For floods, the report investigates the teorology and Environmental Studies (IDEAM), and further full disaster risk—which includes both the baseline risk and desk research.8 In general, the report focuses on three main potential impacts from climate change. Figure I.1 provides a channels through which climate-related risks affect Colombian mapping of the investigated scenarios to the NGFS classifica- banks’ balance sheet: the effects on credit risk in the loan port- tion that are relative to the new set of NGFS scenarios.10 7 US$1 = Col$3,439 on December 31, 2020. 8 IDEAM is Colombia’s national meteorological institute. 9 For the effect of climate-related risks through foreign subsidiary and affiliate exposures, we provide a separate and more high-level analysis in the section titled “Cross-border risks,” in chapter 1. 10 The new NGFS scenarios are expected to be published in May 2021 and provide more variables for a wider set of scenarios than those currently available. To our understanding, no specific data will be available for Colombia, but aggregates will be provided for the Latin America and the Caribbean region and a few of its larger countries. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 10 > > > F I G U R E I . 1 . - Mapping of investigated scenarios to the NGFS classification Disorderly Too little, too late HIGH Delayed NDC 2030 Divergent net-zero (1.5oC) Delayed (2oC) Transition Risks Net-zero Smooth 2050 NDC (1.5oC) 2030 Well bellow (2oC) Floods NDCs Floods in RCP in RCP Current 2.6 8.5 policies Orderly Hot house world LOW LOW Physical Risks HIGH Source: Staff illustration based on NGFS (2020b). Note: NDC = nationally determined contribution; RCP = representative concentration pathway. The exercise is aimed to explore and improve the under- form further deep dives at the institution to confirm the initial standing of the impact of climate-related financial risks to the hypothesis and to get a more detailed understanding of the banking sector, both on an aggregated and per bank level. specific bank’s exposure. Globally, central banks and financial supervisors are develop- ing methodologies and tools to identify and mitigate climate- The report is structured in four chapters. The first chapter related financial risks. These practices are however emerging presents an overview of the climate change in Colombia, the and under continuous development. Our analysis is the first structure of the Colombian banking sector, and the main cli- of its kind in Colombia and is explorative, based on the avail- mate-related risks that are relevant for banks. It also provides able data in the country. Results should be interpreted with a high-level analysis of climate-related risks to countries that some caution because both our analyses and the data used the Colombian banking sector has (indirect) exposures to. for their input are based on different types of modeling (for ex- The second chapter focuses on physical risks, setting out a ample, climate, macroeconomic, financial), hence, leading to vulnerability analysis related to severe riverine floods that are the potential compounding model error. Our results, therefore, in severe cases often linked to La Niña episodes. The third should be interpreted in their order of magnitude and are not chapter focuses on transition risks, estimating the potential intended to identify capital shortfalls or to provide pass or fail impact of delayed decarbonization scenarios on the Colom- outcomes for banks. Besides aggregate results, we also pro- bian economy and banking sector. Finally, the fourth chapter vide outcomes for individual banks to inform microprudential concludes and puts forward policy recommendations to the supervision. With respect to this, we stress the need to per- SFC and other public stakeholders. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 11 1. > > > T A B L E 1 . 1 . - Most damaging natural disasters in Colombia between 1970 and 2020 Type Earthquake Volcanic activity Date January 1999 November 1985 >>> Climate Change and Financial Risks in Colombia Natural disasters and climate change The most economically damaging natural disaster events that occurred in Colombia in the past decades were earthquakes and floods. Colombia suffers from both geophysical and hydrological disasters, which led to economic damages following earthquakes, volcanic activity, floods, insect infestations, and landslides. In today’s currency, those that were most damaging were volcanic activity in 1985 (Col$7.8 trillion), earthquakes in 1999 (Col$9.4 trillion), and the floods resulting from the strong La Niña in 2010 and 2011 (combined damages of Col$12.6 trillion). Of those damages, only a minor fraction of losses was insured. See table 1.1. According to estimates provided by IDEAM, the sectors most affected by the floods in 2010 and 2011 include agriculture (37 percent of total damages), mining (29 percent of total damages), and transport (20 percent of total damages). See CEPAL (2012). Damage US$ thousand 2,850,664 2,376,735 Damage Col$ million 9,407,193 7,843,225 Insured damage Col$ million 506,480 — People affected 1,205,933 12,700 Flood September 2011 1,466,166 4,838,347 — 498,924 Flood April 2010 1,172,442 3,869,059 54,167 2,791,999 Flood April 2011 1,170,659 3,863,176 153,777 988,599 Earthquake March 1983 1,054,715 3,480,561 — 36,200 Flood November 1970 913,980 3,016,133 — 5,105,000 Insect infestation May 1995 174,483 575,796 — — Landslide March 2017 104,299 344,187 — 45,360 Flood March 2012 69,038 227,826 — 8,000 Source: EM-DAT database Note: Damages have been adjusted to reflect the 2019 consumer price index, using an exchange rate of US$1 = Col$3,300; — = not available.. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 12 Climate change over time alters the underlying probability dis- Colombia will likely get substantially hotter during the coming tribution of losses emanating from natural disasters caused by decades, with consequences for business activity and labor the weather, which could lead to stronger La Niña events, in- productivity. IDEAM has created a number of scenarios for creased precipitation, and increased flood hazard. In general, the effects of climate change in Colombia. These scenarios IDEAM expects heavy rains, droughts, and hailstorms in places include predictions for three time frames, that is, changes where these did not occur before, as well as changes in the between 2011–2040, changes between 2041–2070, and properties of these events (for example, mean, modal values, changes between 2071–2100. Scenarios cover temperature and dispersion measures). Increased precipitation is especially and precipitation and are available not only at national but expected in the center and west of the country, and it could lead also regional levels (figure 1.1). The International Monetary to a substantial (25 to 65 percent between 1980 and 2080) in- Fund (IMF) has estimated that a 1 degree increase in tem- crease in flood hazard over time (Winsemius and others 2013). perature in Colombia can lead to a decline in real per capita Moreover, the recurring weather phenomena El Niño and La output of 1.0 to 1.5 percent (IMF 2017). A warmer climate also Niña may increase in severity owing to a changing global cli- implies that the glaciers in Colombia will retreat further, after mate, thereby making episodes of extreme drought (El Niño) already losing 62 percent of the area present in the country and precipitation (La Niña) more likely (Cai and others 2015). before 2017 compared with the mid 20th century (Rabatel and Finally, sea level rise may affect the value of real-estate assets others 2018). and disrupt supply chains of businesses in coastal areas. > > > F I G U R E 1 . 1 . - IDEAM 2015 climate scenarios Difference in (a) Average temperature difference in (b) Average temperature difference in (c) Average temperature difference in 2011–2040 compared to 1976–2005 temperature 2041–2070 compared to 1976–2005 2071–2100 compared to 1976–2005 (degrees C) 0.0 • 0.5 0.51 • 0.8 0.81 • 1.0 1.01 • 1.2 1.21 • 1.6 1.61 • 1.8 1.81 • 2.0 2.01 • 2.1 2.11 • 2.2 2.21 • 2.3 2.31 • 2.4 2.41 • 2.5 2.51 • 2.6 2.61 • 2.7 2.71 • 3.0 (d) Precipitation difference in 2011– (e) Precipitation difference in 2041– (f) Precipitation difference in 2071– Difference in 2040 compared to 1976–2005 2070 compared to 1976–2005 2100 compared to 1976–2005 precipitation (%) Less than -40 -40 ­• -30 -30 • -20 -20 • -10 -10 • 10 10 • 20 20 • 30 30 • 40 More than 40 Source: IDEAM 2015. Note: IDEAM = Institute of Hydrology, Meteorology and Environmental Studies. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 13 The measures for climate mitigation and adaptation are tak- (Bancolombia) and Grupo Bolívar (Davivienda). Bancolombia ing shape. In 2016, the authorities introduced a carbon tax of is the largest bank in Colombia covering 25.4 percent of bank US$5 per ton of carbon dioxide emissions, which is expect- assets followed by Banco de Bogotá (14.5 percent) and Da- ed to reduce emissions by more than 4.3 million tons over vivienda (14.6 percent). Larger banks that do not belong to 13 years. Substantial future efforts to reduce emissions can these three groups include BBVA, owned by the international furthermore be expected in Colombia during the coming de- Grupo BBVA, and the government-owned Banco Agrario de cade owing to the Colombian government’s ambitions to re- Colombia, respectively holding 9.4 percent and 3.8 percent of duce GHG emissions. The Colombian government has long bank assets. recognized the challenges posed by climate change, starting with its approval of the United Nations Framework Conven- The majority of bank assets in Colombia are related to lend- tion on Climate Change in 1994. Since then, the government ing, with a substantial fraction of the assets invested in debt has included climate change in the National Development and equity securities. In 2020, about two-thirds (63 percent) Plans since 2002, signed and ratified the Paris Agreement of the assets in Colombian banks were in the credit portfo- in 2015, established the nationally determined contribution lio, which consists primarily of corporate and consumer loans. (NDC), the creation of the National System of Climate Change Nearly a quarter of all assets (22 percent) is invested in trad- and the formation of an intersectoral commission in charge able securities, including sovereign debt and corporate debt of coordinating in this area. In December 2020, the Colom- and equity. About half (47 percent) of the investment portfolio bian government signed a new NDC with a GHG reduction is invested in government-related securities, whereas another target of 51 percent below a business-as-usual scenario by third (33 percent) is in other financial institutions and groups 2030 (NDC Colombia 2020). This NDC represents one of the (consisting of subsidiaries, branches, and associated entities). most ambitious targets in the Latin America and the Caribbean The remainder of the investments in local securities (10 per- region and is aligned with the country’s long-term objective of cent), foreign securities (2 percent), and others (7 percent). Of achieving carbon neutrality by 2050.11 To achieve this goal, a all investments in subsidiaries, branches, and associated enti- broad-based change in the Colombian economy can be ex- ties most are toward entities outside Colombia (71 percent). pected during the coming decade. Finally, the DNP estimated See figure 1.2. that until 2030 about Col$3.1 trillion (US$0.9 billion) is needed annually to reach the Colombian mitigation goals. DNP pub- Banks in Colombia vary in the way that they address climate- lished a national climate change adaptation plan in 2016, pro- related financial risks in their strategy and operations. A recent viding guidelines and tools to prioritize adaptation action to survey conducted by the SFC shows that 64 percent of all reduce risks. banks has some policies or strategies to explicitly incorporate climate change or is developing one. Most banks furthermore report that they identify and address climate risks and oppor- The Colombian banking sector tunities in their operational and administrative activities and as part of their credit operations. The way in which banks do this, however, varies: for some banks it is part of their corpo- Colombia has a relatively large banking sector with total as- rate social responsibility policies and has not yet been inte- sets in 2020 amounting to Col$730 trillion (US$209 billion or grated into financial risk management, whereas others also 78 percent of GDP).12 Three national financial groups own the identify and take measures for vulnerable sectors and regions majority of assets (66 percent). The biggest group in bank as- to manage financial risks. Some banks approach climate-risk sets is Grupo Aval, which holds about a quarter of all banking management from a business continuity perspective specifi- assets and consists of four banks: Banco de Bogotá, Banco de cally looking at natural disasters. Finally, some banks have Occidente, Banco Popular, and Banco AV Villas. The second publicly communicated that they will reduce their exposure to and third largest groups are Grupo Empresarial Antioqueño coal-related activities to zero over the coming decades.13 11 The 2015 NDC commitments for Colombia aimed for a reduction of GHG emissions by 20 percent below the business-as-usual scenario in 2030. 12 Figures are for December 2020. Bank assets excluding other credit institutions (corporaciones financieras [financial corporations], compañías de financiamiento [trade finance companies], and cooperativas financieras [financial cooperatives]). 13 See https://www.elperiodico.com/es/economia/20210305/bbva-dejara-financiar-actividades-relacionadas-11559688?utm_source=mail&utm_medium=social&utm_cam- paign=btn-share. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 14 > > > F I G U R E 1 . 2 . - Breakdown of assets in the Colombian banking sector in 2020 (a) Overall assets, Col$ trillion (b) Credit portfolio 800 3% Σ = 730 15% 56% 700 41 Commercial lending Consumer lending Housing 600 33% Microcredits 500 461 400 (c) Investment portfolio 7% 300 38% Local sovereign securities Other national government securities 15 200 Local securities 33% Foreign securities 159 100 Subsidiaries, branches and associated entities Other 54 2% 9% 0 10% Cash Investment portfolio Market operations Credit portfolio Other Sources: SFC, data for December 2020; staff calculations. Note: Σ = sum. In the nonfinancial corporate credit portfolio of Colombian end compared with countries in a similar position. At the same banks, about 20 percent of total assets are toward transition- time, exposures to fossil fuels (0.8 percent) are low. Individual sensitive industries, however, with large differences between banks vary substantially in their exposures to transition-sensi- individual banks.14 See figure 1.3. The highest aggregate ex- tive sectors, ranging between 1 percent and 26 percent of the posures in the credit portfolio are toward energy generation corporate credit risk portfolio. (7.9 percent) and transport (6.5 percent), which is on the high 14 Transition-sensitive industries include fossil fuels, energy generation, heavy industry, transport, and agriculture. This description is in line with the classification by Battis- ton and others (2017), except that we define heavy industries instead of energy-intensive firms. See https://www.finexus.uzh.ch/en/projects/CPRS.html. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 15 > > > F I G U R E 1 . 3 . - Sectoral breakdown of commercial lending portfolio (percentage of total) (a) Exposures to transition-sensitive industries, in Col$1 million Agriculture 2.6% Transport 6.5% 20.4% Heavy industry 2.6% Energy generation 7.9% Fossil fuels 0.8% 0 5 000 000 10 000 000 15 000 000 20 000 000 (b) Exposures to transition-sensitive sectors, per bank Bank #1 Bank #2 Bank #3 Bank #4 Bank #5 Bank #6 Bank #7 Bank #8 Bank #9 Bank #10 Bank #11 Bank #12 Bank #13 Bank #14 Bank #15 Bank #16 Bank #17 Bank #18 Bank #19 Bank #20 0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% 40.0% 45.0% Fossil fuels Energy generation Heavy industry Transport Agriculture Other Source: SFC, data for December 2020; staff calculations. Note: Based on a subset of 20 large Colombian banks covering most of the banking assets. Based on sectors 1, 2, 3, 5, 6, 9, 16, 17, 19, 20, 24, 29, 30, 35, 49, 50, 51, 52, and 53. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 16 Main climate-related risks to the banking sector Climate change and its mitigation lead to economic and finan- warming could have significant impacts on the economy. How- cial impacts on traditional financial risks such as credit and ever, policy pressure to tackle environmental pollution and market risk, through different channels. Physical risks origi- improve livelihoods can also lead to significant adjustment nate from natural disasters and climate change that can lead costs for companies and households. Disruptive technologi- to economic costs and financial losses. Physical sources of cal change, for example, in alternative and cleaner sources risk can either be chronic (gradual) in nature, such as rising of energy as well as changing consumer and market behav- temperatures and sea levels and changes in precipitation; or iors toward greener products and services can also result in acute, such as in the case of extreme weather. Transition risks structural economic shifts. In this process toward a greener are related to economic adjustment during the transition to- and carbon-neutral economy, particularly when happening ward a greener, carbon-neutral economy. These risks can be abruptly, revaluations of underlying financial assets are likely. related to climate mitigation efforts, whereby abrupt policies Climate risks and their transmission channels are summarized to reduce carbon dioxide emissions and thereby limit global in figure 1.4. > > > F I G U R E 1 . 4 . - Transmission channels for physical and transition risks Climate risks Economic transmission channels Financial risks Transition risks Micro Credit risk • Policy and regulation Affecting individual businesses and households • Defaults by businesses • Technology and households development Businesses Households • Collateral depreciation • Consumer • Property damage and • Loss of income (from preferences business disruption from weather disruption and severe weather health impacts, labor Market risk • Stranded assets and new market frictions) capital expenditure due to • Property damage (from • Repricing of equities, Financial system contagion transition severe weather) or fixed income, • Changing demand and costs restrictions (from low- commodities, etc. • Legal liability (from failure to carbon policies) increasing mitigate or adapt) costs, and affecting valuations Underwriting risk • Increased insured losses Physical risks Macro • Increased insurance gap • Chronic (e.g., Aggregate impacts on the macroeconomy temperature, • Capital depreciation and increased investment Operational risk precipitation, agricultural • Shifts in prices (from severe heat, diversion of investment to • Supply chain productivity, sea mitigation and adaptation, higher risk aversion) disruption levels) • Labor market frictions (from physical and transition risks) • Forced facility closure • Acute (e.g., • Socioeconomic changes (from changing consuption patterns, heatwaves, floods, migration, conflict) cyclones, and • Other impacts on international trade, government revenues, fiscal space, output, interest rates, and exchange rates. Liquidity risk wildfires) • Increased demand for liquidity • Refinancing risk Climate and economy feedback effects Economy and financial system feedback effects Source: NGFS 2020a. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 17 The Colombian banking sector is to some extent exposed to the Orinoquia region (IDEAM and others 2017). Changes in climate-related natural disasters through assets in vulnerable temperature may affect macroeconomic variables such as areas and through substantial exposure to the official sector. labor productivity and thereby GDP growth over time. Sev- Floods are the main climate-related disaster risk in Colombia, eral banks have indicated that they consider agriculture to be leading to business interruption, damages to real estate, and specifically vulnerable and intend to work on insurance solu- damages to other capital goods. This directly affects the credit- tions to manage this risk. We note that it is already certain that worthiness of affected businesses and households. Corporate some degree of climate change will occur over the coming loans make up an important part of the aggregated banking decades, regardless of emission scenarios. portfolio, amounting to 33 percent of total assets. Historically, floods have affected the Colombian banking negatively owing Transition risks affect a substantial portion of the corporate to increased credit losses and related provisions. Only a mi- loan portfolio of Colombian banks, as well as other asset nor fraction of economic damages after major flood events are classes. Specifically, decarbonization trends can lead to in- insured (between 2 to 4 percent), hence, the impact on under- creased credit and market risk for corporates that are in transi- writing risks is likely low.15 A low degree of insurance penetra- tion-sensitive sectors. The Colombian banking sector is mainly tion, however, leaves a burden on the government budget and exposed to nonfinancial corporates through their credit portfo- affects the creditworthiness of firms in case of a large natural lio, as most investments (bonds and equity) are in the public disaster. Besides firm defaults, a large climate-related disas- sector and toward financial entities. Nevertheless, Colombian ter could also affect the creditworthiness of the Colombian banks may have indirect exposures to transition-sensitive cor- government, leading to a downgrade of official sector debt. porates through their investment portfolio and through invest- Such a downgrade may be particularly worrisome if disaster ments in subsidiaries and associated entities that make up strikes during an economic downturn (that is, a double shock). one-third of the investment portfolio. Furthermore, Colombian Government exposures make up about 8 percent of assets. banks are vulnerable to potential macroeconomic impacts Finally, natural disasters can lead to operational and liquidity because of a broad adjustment of the economic structure risks for banks when branches and payment infrastructure are in stronger transition scenarios. Stress testing by the Dutch affected by a disaster and in case households and firms draw Central Bank (DNB) has shown that in sudden transition sce- on their deposits to finance a recovery. narios, wider macroeconomic effects may lead to losses in the financial sector (Vermeulen and others 2019).16 Besides direct Furthermore, climate change may affect the underlying eco- impacts on transition-sensitive sectors, these macroeconom- nomics of businesses in a more gradual way, thereby af- ic impacts can lead to credit losses on other corporate and fecting credit risk, market risk, and potentially affecting mac- household loans (for example, because of growth and higher roeconomic variables over time. Changes in temperature, unemployment) as well as losses caused by changing interest precipitation, and droughts can have an important impact on rates. Also, the collateral value of real estate loans may de- the market value and creditworthiness of the agricultural and crease for energy inefficient buildings. We note that transition energy sectors, as well as increase the frequency and sever- policies may not only originate in Colombia but also abroad. ity of natural disasters. Agriculture is an important sector in Impacts on the Colombian banking system may then manifest Colombia, representing 6.3 percent of GDP in 2018 and gen- themselves through cross-border exposures through foreign erating more than half of the employment (59.7 percent) of the subsidiaries and associated entities and trade channels (for rural population (Melo and others 2019). Regional changes in example, exports of GHG-intensive products such as coal and temperature, rainfall, and the amount of sun affect agricultural oil). Transition risks in Colombia are high compared with most output and may require a shift in crops and techniques to miti- other countries in the region, because of a combination of a gate the effects of a changing climate. Furthermore, the reli- high carbon intensity of the economy, fossil fuel exports, and ance of the Colombian energy sector on hydropower makes it high policy targets for GHG emission reduction. vulnerable to changes in precipitation patterns (and increasing uncertainty surrounding them). For hydropower in particular, High-level sectors that are most relevant for banks to con- an early assessment or risks is important because of the long sider for transition risks include agriculture, energy genera- lifespan of the plant. Colombian banks should closely moni- tion, and transport. All these sectors contribute substantially tor their exposures to these sectors. Relatively high impacts to GHG emissions and are substantially represented in the of changes in temperature and precipitation are expected in banks’ corporate loan portfolios. The highest aggregate tran- the Andean region as well as in the Amazon foothills and in sition-sensitive exposures in the credit portfolio are toward 15 Insurance penetration based on data from EM-DAT. 16 DNB, among others, investigates an overnight application of a US$100 carbon tax. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 18 energy generation (about 6 percent) and transport (around 4 one hand, risks I and II are mostly gradual in nature that are percent). At the same time, the direct exposures to fossil fuels expected to play out over a relative long time. These risks are (about 1 percent) are low, although that sector is the most vul- primarily relevant from the perspective of a business model, nerable to an energy transition. Some banks have indicated where banks that do not adjust to changing circumstances that they focus on specific sectors with respect to transition (for example, in pricing and loan origination practices) could risks, including agriculture (specifically cattle breeding), min- become less profitable over time, eroding their capacity to ing projects, the manufacturing industry, electricity generation, replenish financial buffers when needed. In these scenarios, the construction sector, the waste management sector, road there also could be more abrupt shocks when prices of cer- transport, plastic and rubber production, and tourism.17 Fur- tain assets rapidly change because of better market under- ther exploration of vulnerable sectors to transition risks is car- standing (for example, rapid decreasing real estate prices in ried out as part of the analysis in chapter 3. coastal areas). On the other hand, more acute risks could play out over a shorter time, including natural disasters and sud- In sum, the Colombian banking sector is vulnerable to gradual den tightening of climate policies in the coming decade (un- and more acute risks that stem from both transition and physi- til 2030). These more acute risks are the risks that have the cal risks. We summarize the main risks in table 1.2. On the highest potential for banking sector stress in the next decade. > > > T A B L E 1 . 2 . - Summary of main climate-related risks for the Colombian banking sector Potential for Risk Likelihood Channels banking sector stress • Increasing loan losses in I. Gradually increasing carbon Medium Low transition-sensitive sectors price and climate policies • Value of commercial real estate II. Gradually increasing • Increasing loan losses in temperature and changing High Low/medium vulnerable sectors (e.g., weather-patterns agriculture) • Increasing loan losses in III. A sudden tightening of Low/medium Medium/Large transition-sensitive sectors climate policies • Value of commercial real estate • Macroeconomic effects • Real estate, corporates, IV. Severe flood Medium Medium households in affected areas • Sovereign credit downgrades • Real estate, corporates, V. Severe flood plus Low/medium Large households in affected areas recession (double shock) • Sovereign credit downgrades • Macroeconomic effects Source: World Bank staff. 17 Based on the second climate risk and opportunities survey, conducted by the SFC amongst Colombian financial institutions. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 19 Cross-border risks One specific feature of the Colombian banking system is that High-level country indicators show that physical risks could climate-related risks may manifest through substantial cross- be relatively high, and transition risks more limited, in foreign border exposures, which are significant in highly climate-sen- exposures compared with domestic exposures. See table sitive economies in Central America. The banks’ investment 1.3. Compared with the main countries where the Colombian portfolio contains almost no exposure to nonfinancial corpo- banking system has exposures, Colombia scores relatively rates but may experience indirect losses owing to investments high on transition risk indicators (fossil fuel exports, carbon in other financial institutions. In total, more than 7 percent of dioxide intensity, and NDC goals for 2030), whereas it scores total assets in the Colombian sector are toward subsidiaries, relatively low on physical risk indicators (natural disasters and branches, and affiliated entities. Most of these exposures are vulnerability to climate change). Despite detailed data not be- part of the investment portfolio and toward entities that are lo- ing available, these indicators imply that foreign exposures of cated mainly in Central America, Paraguay, and Peru. In the Colombian banks are relatively vulnerable to physical risks. investment portfolio, a third of affiliated exposures are toward Some of these physical risks may be correlated to domes- entities in Colombia (29 percent), whereas more than two- tic risks, in as far as they are caused by the same weather thirds (71 percent) are toward entities abroad. This exposes phenomena (for example, El Niño and La Niña). Transi- the banking sector to indirect climate risks (both physical and tion risks may also be correlated across countries, owing to transition). This section provides a first analysis of cross-border concerted efforts to achieve NDC targets related to the 2015 risks based on high-level indicators of climate risks in countries Paris Agreement. that the Colombian banking sector has exposures to. > > > T A B L E 1 . 3 . - Climate-risk vulnerability indicators per country Fossil fuel CO2 intensity CO2 intensity NDC goals Natural Climate exports exports domestic for 2030 disasters change Fossil fuel CO2 intensity of Carbon intensity Emissions reduction Average ND GAIN export revenue manufacturing (kg CO2 per compared weather-related vulnerability (normalized) export (normalized) US$ of GDP) with BAU losses (% of GDP) index Colombia 0.26 0.37 0.17 51% 0.16 39 Costa Rica 0.14 0.28 0.12 44% 0.13 39 El Salvador 0.11 0.33 0.13 NA 0.67 45 Guatemala 0.13 0.32 0.15 NA 0.50 46 Honduras 0.09 0.33 0.24 NA 0.47 46 Nicaragua 0.10 0.29 0.19 10%* 0.65 45 Panama 0.19 0.30 0.13 11.5%* 0.01 41 Paraguay 0.22 0.25 0.12 NA 0.78 38 Peru 0.18 0.38 0.15 30% 0.14 43 Sources: World Bank; Global Carbon Project; World Resources Institute; Germanwatch; Notre Dame GAIN; staff calculations. Note: BAU = business as usual; CO2 = carbon dioxide; GDP = gross domestic product; kg = kilogram; NA = not available. ND-GAIN = Notre Dame Global Adap- tation Initiative; * = target applies to specific sectors, hence providing an upper bound. No data available for the Cayman Islands and Barbados. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 20 2. >>> Physical risk assessment In this chapter we take a more detailed look at the vulnerability of Colombian banks to severe flood scenarios. Riverine floods are the main climate-related disaster risk in Colombia, leading to business interruption, damages to real estate, and damages to other capital goods. The La Niña–related floods in 2010 and 2011 resulted in economic damages of US$8.6 billion, making it the costliest climate-related disaster in recent history in Colombia.18 Moreover, the economic damage of floods is expected to increase owing to a combination of socioeconomic development and climate change. The flood hazard (measured in flood volumes with a 100-year return period) is expected to increase in many regions globally during the coming decades because of climate change, with estimates suggesting that the flood hazard in Colombia could increase between 25 and 65 percent between 1980 and 2080 because of climate change alone.19 We explore the impact of riverine floods on Colombian banks by modeling bank stress in three steps (figure 2.1). In a first step, we estimate economic damages per municipality in different sce- narios. These scenarios vary on the basis of their return period (that is, the frequency with which they occur) and the year in which they occur (with later years experiencing a larger effect of climate change). Because probabilistic flood risk scenarios are not available in Colombia for economic damages per municipality, we downscale national-level flood risk estimates by the World Resourc- es Institute (WRI) using a relative economic flood risk indicator at a municipal level. In a second step, we estimate financial losses through two channels: (a) an increase in loan loss provisions and (b) an increase in the sovereign credit spread. In a third step we combine our estimates of financial losses per unit of exposure with data on municipal loan exposures, sovereign bond expo- sures, and balance sheets per bank to obtain a first estimate of the effect of each scenario on the capital position of banks. A detailed description of data and variables used is provided in figure 2.5. > > > F I G U R E 2 . 1 . - Main elements of the flood vulnerability assessment 1. Flood risk 2. Financial impact 3. Banking sector stress • Relative economic risk (municipal) • Loan provisions impact • Balance sheet and exposures (municipal) • Economic damage (national) • Sovereign credit spread impact • Impact on capital adequacy ratio (CAR) Source: Staff illustration. 18 We note that earthquakes also could cause substantial damages in Colombia but are out of scope of this climate-risk assessment. 19 See Winsemius and others (2013). Presented estimates represent the average of flood hazard simulations by five Global Climate Models for the representative concen- tration pathway (RCP) 2.6 and RCP 8.5 scenarios. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 21 Flood risk Floods in Colombia can inundate substantial areas throughout rainfall (IDEAM 2012). Periodical and less severe floods are the country, with the highest levels of inundation historically observed in the east and south of the country (light blue areas observed during strong La Niña episodes. According to data in figure 2.2), specifically in the Arauca and Casanare depart- by IDEAM, floods can occur throughout the country but espe- ments. The flood maps from IDEAM include the areas that cially in the northwest (dark blue areas in figure 2.2), among were flooded during La Niña of 1988–1989 (strong event), others covering the Sucre, Bolívar, Antioquia, and Magdalena 1999–2000 (moderate event), and 2010–2012 (strong event). departments. The north and west areas of the country have We aggregate this data to the municipal level to create an indi- also seen the most damaging floods during La Niña, a recur- cator of flood susceptibility to use as a starting point for further ring weather phenomenon in the Pacific Ocean causing heavy analysis. See panel a in figure 2.3. > > > F I G U R E 2 . 2 . - Areas with high, medium, and low susceptibility to floods in Colombia High Medium Low Source: IDEAM. Note: map depicts outer limits for extreme events. The areas covered do not provide an exhaustive mapping of potentially floodable areas. Some scenarios lead to additional areas being flooded, for example, when specific water defense infrastructure fails. See, for example, Cardona and others (2017). EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 22 A relative economic risk indicator shows that economic losses ity, obtained from the National Administrative Department of are highest in the north, center, and west of the country and Statistics (DANE) in panel b.20 We calculate the relative eco- are geared toward the rural areas. We create an indicator for nomic risk by multiplying the flood hazard with the economic the relative economic flood risk per municipality by combining exposure (C = A x B). This gives an indication of the fraction flood hazard with economic exposure data. Throughout our of value added that is at risk per municipality. Larger cities further analysis we aggregate or disaggregate all data to the tend to be less vulnerable to floods and, hence, pose a lower municipal level, which is needed to link our estimates to avail- risk despite the high concentration of economic activity (for able financial sector exposure data. We create an economic example, in Bogotá, Medellín, and Cali). Departments that are flood risk indicator (figure 2.3, panel c) by combining the IDE- at a relatively high economic risk include Antioquia, Cundina- AM data on the fraction of a municipality that is susceptible to marca, and Bolívar. flood (panel a) with the value added in that same municipal- > > > F I G U R E 2 . 3 . - Relative economic risk to severe riverine flood events (a) Flood hazard (b) Economic exposure (c) Relative economic flood risk Fraction of area susceptible to flood Value added, year-end 2018 Fraction of value added at risk Sources: Data obtained from IDEAM and DANE; staff calculations. Note: Darker shades indicate a higher value for the variable. 20 In taking this approach, we assume that economic activity is spread equally across each municipality. We have to make this assumption because we do not have data on the precise location of the assets of Colombian banks. However, municipalities are reasonably small, and our analysis uses a total of 1,122 municipalities. We use weight 1 for areas with high susceptibility to floods and weight 0.5 for areas with medium susceptibility to floods. These weights align with flood depths that occur once every 50 years as calculated by Ingeniar in UNGDR (2018). UNGDR is Colombia’s National Unit for Disaster Risk Management. See appendix B on Colombia’s flood vulnerability. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 23 Global probabilistic riverine flood models estimate total eco- culation is based on an increase in flood hazard as a result of nomic damages in Colombia of up to US$23 billion to US$25 changes in precipitation patterns, according to an averaging of billion in 2030 in severe but plausible scenarios. The WRI Aq- five global climate models (Winsemius and others 2013) and ueduct database provides probabilistic estimates on economic assuming economic damages proportionally grow with flood damages at a national level from riverine flooding in Colombia, volumes. Although estimates are subject to a relative high de- based on the average of a set of five global climate models gree of model uncertainty, all five models for Colombia point (Ward and others 2020). Economic damages are estimated as to a substantial increase in flood hazard until 2080. Estimates a function of the return period (RP), with lower return periods range from a 25 percent increase between 1980 and 2080 indicating more likely events. For example, an event with an RP in RCP 2.6 (declining emissions from 2020 and net-zero by of 500 years is expected to occur with a probability of 1 in 500 2100) to a 65 percent increase in RCP 8.5 (worst-case sce- in each single year. Looking at longer time spans, however, it is nario with a continued rise of GHG emissions throughout the increasingly likely that such events will occur at least once dur- 21st century).21 To illustrate, these estimates imply that an RP ing that period. An event with an RP of 500 years is expected 250 event in 2030 may turn into an RP 50 event in 2080. It to occur within the next 50 years with a probability close to 10 also underscores the point that events in the past decades percent. The WRI estimates that an RP 500 event in 2030 could may already have been more severe owing to climate change cause US$23 billion in economic damages, on the basis of an that has already occurred. We note that economic damages intermediate climate change scenario (representative concen- may increase more than 25–65 percent until 2080 because tration pathway [RCP] 4.5) and intermediate socioeconomic of economic development (and, hence, higher exposures). developments (shared socioeconomic pathway [SSP] 2). This We exclude this from our analysis, however, because balance amount increases to US$25 billion in an RP 1,000 scenario and sheets of banks can be expected to grow in line with economic reduces to US$13 billion in an RP 25 scenario. See figure 2.4, development, and we take the 2030 estimates by the WRI as the blue lines (2030). Using these estimates, riverine flooding our baseline.22 We also note that adaptation investments (for of the magnitude of the 2010–2011 La Niña floods would occur example, additional flood management infrastructure) and di- roughly every eight to nine years in 2030. saster-resilient planning for new economic activity can reduce overall economic damages as well as the effect on Colombian Climate change could increase economic damages of riverine banks. Estimates for 2050 and 2080 are shown in figure 2.4, floods with 25–65 percent between 1980 and 2080. This cal- under RCP 2.6 (panel a) and RCP 8.5 (panel b). > > > FIGURE 2.4. - Projected economic damages for riverine flood in Colombia in 2030 and increases over time because of climate change (a) RCP 2.6. Declining emissions from 2020 (b) RCP 8.5. Worst-case scenario with a continued and net-zero by 2100 rise of GHG emissions throughout the 21st century 40 40 Economic damage (US$billion) Economic damage (US$billion) 35 35 30 30 25 25 20 20 15 15 10 10 5 5 0 0 0 500 1000 0 500 1000 Return period (years) Return period (years) 2080 2050 2030 2080 2050 2030 Sources: World Resources Institute Aqueduct Database; Winsemius and others 2013; staff calculations. Note: RCP = representative concentration pathway; SSP = shared socioeconomic pathway. Estimates for 2030 based on RCP 4.5 and SSP 2; estimates for 2050 and 2080 show the effect of climate change only, based on RCP 2.6 (panel a) and RCP 8.5 (panel b). All estimates are in today’s currency (no inflation). 21 For our analysis we also assume that increases in economic damages owing to climate change grow constantly over time and that flood hazard scales linearly with eco- nomic damages. 22 Estimates were obtained while holding the level of socioeconomic development constant. Socioeconomic development is less relevant for our analysis as an increase in socioeconomic activity would likely also lead to an increase in banking activity (including available capital). In doing so, we assume that balance sheets of Colombian banks grow in line with socioeconomic activity but otherwise remain similar in structure (semistatic balance sheet). EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 24 Financial impact Historically, natural disasters have led to higher loan losses visions that were made over the shock period amounted for banks as well as credit rating downgrades to the debt of to 11.9 percent of total assets in flooded areas.25 national governments (Klomp 2014 and Standard & Poor’s 2015). To link disaster risks to losses on financial assets, we b. Furthermore, S&P data show that every 1 percentage point obtain empirical estimates for the elasticity of loan loss provi- of GDP lost owing to natural disasters on average leads to sions and the sovereign credit rating to economic damages a downgrade of sovereign debt by 0.28 notch. Sovereign owing to natural disasters. For loan loss provisions, we make credit rating downgrades occur regularly because of major our own estimates based on historical provisioning data avail- natural disasters. Based on S&P data on disaster events able at the SFC. For sovereign credit ratings, these national that occurred in other countries, we estimate that every data are not sufficient to make a comparable estimate. Hence, percentage point of GDP in economic damages leads to we base ourselves on data from the credit rating agency Stan- a downgrade of Colombian sovereign debt by 0.28 notch dard & Poor’s (S&P) that estimates the relationship between (Standard & Poor’s 2015). Downgrades of sovereign debt losses from natural disasters and credit ratings from a global affect the balance sheets of banks mainly through their sample of countries, including several in Latin America. holdings of sovereign bonds, through their impact on the credit spread that investors require to hold the debt. A de- a. Based on historical data available for Colombia, we esti- crease in a sovereign’s credit rating on average leads to mate that every 1 percentage point of GDP lost owing to an increase in the credit spread and hence a decrease in flood damages causes a 0.12 percentage point increase the market value of the debt. This decrease has an imme- in loan loss provisions.23 Using a difference-in-difference diate impact on sovereign bonds that are held at market (DID) panel model specification and SFC spatial credit value. Moreover, from a prudential perspective it would data, we estimate excess credit provisions in affected also be important to consider bonds that are not held at areas that cannot be attributed to other cyclical and market value to estimate the potential losses that a bank bank-specific factors (for example, interest rate changes, would incur in case of insolvency and sale of its assets. national policies, differences between bank business To that end, we provide a comparison between stress re- models, and so forth). Our model estimates that the La sults under accounting rules and stress results where we Niña floods of 2010 and 2011 led to excess provisions of value the whole sovereign debt portfolio at market value. 0.12 percentage point of the loan portfolio per 1 percent- We use an estimated 49 basis points change in credit risk age point of departmental GDP in economic damages.24 spread per rating notch, obtained from S&P (Standard & This is equivalent to total excess provisions of Col$1.5 tril- Poor’s 2019). This calculation is based on the average of lion for the entire banking sector (0.3 percent of the total the five rating categories around Colombia’s 2019 credit credit portfolio). Our estimates furthermore imply that pro- rating (BBB-). 23 A loan loss provision is an income statement expense set aside to cover for anticipated uncollected loan payments. 24 Further details on the regression specification are in appendix A. We use this parameter to estimate the total amount of excess provisions based on economic damages as modeled by the WRI. We then allocate excess provisions to municipalities using the relative economic flood risk indicator (figure 2.3, panel c). 25 This calculation is under the assumption that losses occurred in flooded areas only (that is, no spillovers) and that exposures in affected areas are a good representation of total exposure within a municipality. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 25 Banking sector stress We investigate four scenarios using a basic stress test- the relative economic risk per municipality and (b) determin- ing model, based on the strongest flood to date and plausi- ing market-value losses to sovereign exposures. We finally bly more severe scenarios. To gauge the effects of severe distribute loan losses proportionally to banks with exposures floods, we link the disaster risk and financial impact estimates in those municipalities and market-value losses on sovereign to an accounting-based stress test framework as developed bonds in line with banks’ sovereign exposures. This gives us by Čihák (2014). See figure 2.5. The Čihák model uses bal- an estimate per bank of the impact of severe flood events. The ance sheet and profit and loss data from banks to determine four scenarios that we investigate are (A) floods with the se- the effect of traditional financial sector shocks on the bank’s verity of the 2010–2011 La Niña occurring in 2020; (B) severe capital adequacy ratio (CAR) and other profitability and sol- floods, using a 500-year return period, based on estimates of vency metrics. We adapt this model to allow for the use of flood risk in 2030 obtained from the WRI Aqueduct database; granular spatial and sectoral credit exposure data and to in- (C) severe floods occurring in 2080 distinguishing between a clude a sovereign credit spread channel. We calibrate the low (RCP 2.6) and high (RCP 8.5) climate change scenario model using projected losses by WRI and estimates for the based on Winsemius and others (2013), and (D) severe floods relationship between economic damages and financial sector occurring in 2030 (similar to scenario B) at the same time as outcomes (that is, credit provisions and sovereign credit rat- a recession. The recession is calibrated to reflect loan losses ings). We then investigate severe flood events by (a) allocat- during the 1998–2000 banking crisis in Colombia. ing estimated excess provisions in the loan portfolio based on > > > F I G U R E 2 . 5 . - Overview of the flood vulnerability model 1. Disaster 2. Financial 3. Bank risk modeling impact modeling stress modeling • Flood areas (IDEAM) • Elasticity estimate • Spatial credit • Balance sheets (SFC) • Value added (DANE) (WB, CEPAL, SFC) exposures (SFC) • P&L data (SFC) Economic flood risk Economic damage Credit provisions Credit provisions Capital adequacy per monicipality per monicipality per municipality per bank ratio (CAR) Economic damage Change in sovereign Market value Profitability (national level) credit spread losses per bank (ROA, ROE) • Economic damage • Elasticity estimate • Sovereign debt (WRI) (S&P) exposure (SFC) • Climate change impact (Winsemius and others, 2013) Source: Staff illustration. Note: CEPAL = Economic Commission for Latin America and the Caribbean; DANE = National Administrative Department of Statistics; IDEAM = Institute of Hy- drology, Meteorology and Environmental Studies; P&L = profit and loss; ROA = return on assets; ROE = return on equity; SFC = Financial Superintendence of Colombia; S&P = Standard & Poor’s; WB = World Bank; WRI = World Resources Institute. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 26 Overall, our analysis shows that flood scenarios can lead to floods. Furthermore, we find that a worst-case climate change declines in capital adequacy that could moreover coincide with scenario (RCP 8.5) could add an additional 0.1 to 0.6 percent- other shocks (figure 2.6). We present high-level outcomes for age points in CAR impact—per event—compared with a sce- two groups: domestic (12 banks) and foreign (8 banks). For nario with limited climate change (RCP 2.6) depending on the domestic banks, the average decline in the CAR for the se- bank. This range reflects differences in individual bank’s expo- vere flood scenarios (scenarios A to C) ranges between 0.3 sures and hence their vulnerability to climate change. These and 1.0 percentage point. For foreign banks, the decline rang- results are under the assumption that banks do not change es between 0.3 and 1.2 percentage point. In the double shock their business model and asset allocation to reflect climate scenario (scenario D) the average decline in the CAR is 3.1 change. If banks would adjust their operations, impacts could and 3.4 percentage points for domestic and foreign banks, re- be more limited. Adjustments could, among others, include al- spectively. Across all banks, total losses in the double shock tering origination processes and requiring better insurance for scenario can be mostly attributed to the economic recession households and firms at high risk of flooding. (76 percent) with about a quarter of losses caused by severe > > > F I G U R E 2 . 6 . - Estimated impact on capital adequacy for physical risk scenarios with varying severity (a) La Niña 2011–2012 in 2020 (B) Severe flood (RP500) in 2030 20% 20% 15% 15% 10% 10% 5% 5% 0% 0% Domestic Foreign Domestic Foreign (C) Severe flood (RP500) in 2080 (D) Double shock (1998–2000 crisis) 20% 20% 15% 15% 10% 10% 5% 5% 0% 0% Domestic Foreign Domestic Foreign Pre-shock CAR (%) Post-shock CAR 9% minimum CAR Source: Staff calculations. Note: CAR = capital adequacy ratio; RP = return period. Results are obtained using an accounting-based framework, in which banks only recognize market value losses on sovereign exposures in the trading book (and not in the banking book). For future scenarios, we assume semistatic balance sheets with socioeconomic development that grows in line with the size of banking sector balance sheets and the structure of banks’ lending staying constant. The double shock scenario combines the severe flood scenario (RP 1 in 500 years) occurring in 2030 (scenario B) with a credit risk shock that is calibrated to the 1998–2000 banking crisis in Colombia, using a portfolio-wide increase in nonperforming loans of 14 percent of total loans. We calculate loan losses using a provisioning rate of 25% for nonperforming loans while accounting for a reduction in required capital related to the provisioned loans. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 27 The vulnerability of individual banks shows high heterogene- shock scenarios (scenario D) the impact is high for all banks, ity, reflecting exposures in more metropolitan areas versus with less pronounced differences between individual banks. In more rural areas and exposures to sovereign debt. Looking this case the difference between the worst and least affected at the severe flood scenario in 2080 (scenario C), aggregated banks is about 2 times (figure 2.7). We note that this analysis loan losses for Colombian banks range between 0.2 percent assumes that exposures of banks are homogeneous. In real- of total assets for the least vulnerable bank to 2.2 percent for ity, some banks will have loans that are more exposed than the most vulnerable ones. The difference between the banks others, for example, in specific sectors such as agriculture and with the lowest relative vulnerability compared with the high- mining. Such differences in sectoral exposures could lead to est is more than tenfold, pointing to focusing supervisory ef- higher or lower estimates than the ones provided in figure 2.7 forts at those banks that are most vulnerable. For the double and would require more detailed analysis of each bank. > > > F I G U R E 2 . 7 . - Accounting-based losses for flood scenarios per bank (percentage of total assets) 4.0 3.5 3.0 Percentage of total assets 2.5 2.0 1.5 1.0 0.5 0 #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12 #13 #14 #15 #16 #17 #18 (A) La Niña 2011—2012 in 2020 (B) Severe La Niña (1 in 500) in 2030 (C) Severe La Niña (1 in 500) in 2080 (D) Double shock (1998—2000) banking crisis Source: Staff calculations. Note: accounting-based losses only account for market valuation changes in part of the sovereign bond portfolio that is market-to-market. Losses would be higher if a market-value based approach is used. Losses for banks in due to floods are primarily driven by loan losses in vulnerable areas. Under an accounting based approach, credit losses account for 85 percent and 54 percent of total losses for domestic and foreign banks, respectively. However, espe- cially for foreign banks the potential impact through the sovereign bond channel also constitutes an important driver of total losses. Figure 2.8 shows the contribution of the credit risk and sovereign bond channel to the percentage point reduction in the CAR for the different groups of banks. Moreover, results using a full market-based valuation approach show that banks may be more vul- nerable than the accounting based approach would suggest, due to sovereign holdings not valued on a market basis. If we repeat the analysis by treating all sovereign debt being valued on a market-consistent basis, we find that losses increase substantially, owing to the relatively large share of sovereign exposure held as part of the banking book. For domestic banks, using a full market value-based approach increases the CAR impact with an average of 0.6 percentage point in the 2010–2011 La Niña scenario (scenario A) representing a 113 percent increase in losses through the sovereign bond channel. For foreign banks this increase is 0.5 percentage points while representing a 28 percent increase in losses on sovereign exposures. Hence, using an accounting- based approach could mask a part of the economic losses that may arise during a severe disaster. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 28 > > > F I G U R E 2 . 8 . - Breakdown of losses through different channels (percentage point decline in the CAR) Accounting based Market value based 0 0 -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 -0.4 -0.4 -0.5 -0.5 Domestic Foreign Domestic Foreign Sovereign bond channel Credit risk channel Source: Staff calculations. Note: CAR = capital adequacy ratio; data based on the 2010–2011 La Niña scenario (A). Our estimates for riverine flood risks are most likely conservative, that is, providing lower bound estimates. There are three main reasons for this. First, provisioning estimates are likely on the conservative side owing to forbearance (that is, lower initial provi- sioning than needed with some additional provisions possibly being taken over time) and supply-chain effects (which could lead to losses outside of the affected areas and, hence, falling outside of our loss estimate). Second, our focus is on the credit risk and sovereign bond channels. Additional losses may occur through other channels, including simultaneous disaster losses incurred through investments in foreign subsidiary banks. And, third, our analysis assumes that losses from floods are growing linearly with the scale of the flood hazard. However, more severe floods could cause disproportionally higher losses in some cases (for example, if a building fully collapses). In general, our estimates provide a first indication of vulnerability, but more detailed analysis should be carried out to understand the exposures of individual banks (which may, for example, depend on the type of exposure of individual banks per region). EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 29 3. > > > >>> Transition risk assessment A broad spectrum of sectors are potentially at risk when the economy decarbonizes. Firms in transition-sensitive sectors may lose a substantial share of their value when the economics of their industries change because of new policies, technological innovation, and changing con- sumer preferences. Those industries include sectors that have high direct GHG emissions as part of their business processes, such as utilities and agriculture (scope 1). However, a much larger share of the economy could be affected. In particular, this includes firms that are part of the fossil fuel supply chain, such as fossil fuel extraction and automobile producers, or that use large amounts of heat or electricity, such as heavy manufacturing (scope 2 and 3). For individual sectors, the economic impact depends on their reliance on GHG emissions but also on the ease with which companies can reduce their internal footprint (for example, owing to abatement investments) and on the ease with which their products can be substituted for lower carbon al- ternatives (for example., traditional versus electric cars). Currently, for Colombia, only estimates for direct (scope 1) emissions per sector are available, as shown in figure 3.1. We address this issue by working with a macroeconomic model that also includes scope 2 and scope 3 effects in sectoral estimates. F I G U R E 3 . 1 . - Direct (scope 1) GHG emissions per sector in Colombia (Gt CO2e per Col$1 billion value added) Cattle raising Utilities Fossil fuel extraction (coal, oil, natural gas) Land transport Nonmetallic minerals, including cement Air transport Petroleum refining Pulp, paper, and printing Chemical products Iron, steel, and metals Textiles and leather Food, beverage, and tobacoo 0 2 4 6 8 10 12 Gt CO2e per Col$1 billion value added Sources: DANE; IDEAM 2014; staff calculations. Note: Col$ = Colombian peso; GHG = greenhouse gas; Gt CO2e = gigatonnes of equivalent carbon dioxide. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 30 The objective of this chapter is to investigate the potential for banking sector stress in delayed transition scenarios, following three steps. First, we define transition risk scenarios that are in line with Colombian NDC targets. We do this based on an already avail- able emission trading system (ETS) scenario at the DNP while increasing the severity of the scenarios under investigation for our stress test exercise. Second, we model what the consequences in these scenarios can be for different sectors in the economy, looking at the value-added impact per sector. For this we include dependencies between sectors in our analyses by employing a macroeconomic modeling approach.26 In this way, we look at scope 1, 2, and 3 emissions. As a third step, we perform a simplified stress test to size the potential impact on the Colombian banking sector. See figure 3.2. > > > F I G U R E 3 . 2 . - Main elements of the transition vulnerability assessment 1. 2. 3. Transition Risk Scenarios Macroeconomic Impact Modeling Banking Sector Stress Source: Staff illustration. Transition risk scenarios We define four scenarios that are relevant for Colombia and scenarios with a high target (51 percent GHG emission reduc- that vary based on the target GHG emission reduction in 2030 tion in 2030). We use a combination of the higher target and and the timing of climate policies. We start with an existing delayed scenarios as adverse scenarios to size the potential scenario used by the DNP to investigate the impact of an ETS impact on the banking system, including a scenario that in- on the Colombian economy, which assumes a 10-year phase corporates both dimensions (scenario D). See table 3.1 for a in and an achievement of a 20 percent emissions reduction summary of the scenarios. We note that the high target (51 until 2030 (scenario A). We then extend this scenario along percent) is in line with the current NDC in Colombia. We note two dimensions. With respect to timing, we define “smooth” that GHG emission reduction targets in the Colombian NDC scenarios in which climate policies are introduced from 2021 are reductions with respect to a business-as-usual scenario and “delayed” scenarios in which policies are introduced only which is part of the DNP model. Compared with 2020 emis- from 2026. With respect to targets, we define scenarios with a sions, the reduction in the high target scenario in 2030 equals low target (20 percent GHG emission reduction in 2030) and 41 percent. > > > T A B L E 3 . 1 . - Construction of delayed transition scenarios based on the existing DNP ETS policy scenario Smooth Delayed (from 2021) (from 2026) Old NDC target (A) (B) (20 percent in 2030) Smooth 20% (DNP ETS policy scenario) Delayed 20% New NDC target (C) (D) (51 percent in 2030) Smooth 51% Delayed 51% Source: Staff illustration. Note: DNP = National Planning Department; ETS = emission trading system; NDC = nationally determined contribution. 26 In the computable general equilibrium model, dependencies between sectors are modeled using input-output tables. This method provides a much better, but not per- fect, estimate of broad economic impacts owing to decarbonization policies. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 31 The implicit carbon price in Colombia in these four scenarios struments are considered. In general, the implied carbon price is expected to increase between US$64–216 in 2030 depend- estimates are valid as long as emissions are reduced in line ing on the reduction target. This implicit carbon price is de- with the NDC targets, with a similar scope of sectors (in this rived from a macroeconomic model used by DNP, which is de- case, the whole economy) and at its lowest economic cost. scribed in the next section. This model determines the carbon We note that climate policies may exclude some sectors or dif- price based on a reduced supply of emission rights over time ferentiate between sectors because of political choices. In this to achieve the 2030 GHG emission reduction target. A simi- case implied carbon prices between sectors could vary. Figure lar pricing dynamic would be observed if, instead of emission 3.3 shows the implicit carbon price as part of the four sce- rights, the government would tax GHG emissions at a rate narios and represents the carbon price that would be needed equal to the price of an emission right. Hence, whereas the to achieve the stipulated GHG emission reduction goals in the DNP modeling is based on ETS scenarios, outcomes in sec- Colombian NDCs. toral impacts are roughly similar when other climate policy in- > > > F I G U R E 3 . 3 . - Implicit carbon price development in the four scenarios 250 200 US$/TnCO2eq 150 100 50 0 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 (A) Smooth 20% (B) Delayed 20% (C) Smooth 51% (D) Delayed 51% Source: Data obtained from DNP CGE model. Note: ETS = emission trading system; TnCO2eq = ton of carbon dioxide equivalent. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 32 Macroeconomic impact We explore the potential effect of climate policies on the Co- Results from the DNP macroeconomic modeling show that lombian economy and banking sector using a CGE model of decarbonization trends may affect a broad range of sectors the DNP on the basis of four transition scenarios. This model and can potentially be sizeable. In the scenarios with a low provides estimates for GDP and employment impact per eco- GHG emission reduction target (20 percent) the impact on nomic sector in Colombia, on the basis of scenarios in which GDP growth is not leading the economy into negative growth an ETS is implemented. Such an analysis provides a good territory. However, the GDP impact increases strongly with indication of vulnerable sectors, as the implicit carbon price higher GHG emission reduction targets. In the scenarios with used in this modeling specifically targets those industries that a high GHG emission reduction target (51 percent), relatively contribute the most to global warming and, hence, are the most high negative growth rates are obtained in the second half likely target of any type of climate policies. The DNP model of the decade., driven by strong reductions in output in GHG achieves a cut in GHG emissions through final consumption, intensive sectors and in sectors that are dependent on those intermediate consumption, and process emissions by increas- sectors.27 Whereas the impact on sectors is mainly limited to ing the implicit price of carbon using an ETS. The ETS covers the fossil fuel sector in the low target scenarios, we find that all sectors of the economy and also shows broader effects on sectors throughout the economy are affected in high target sectors that are outside the set of transition-sensitive sectors. scenarios. With a 51 percent GHG emission reduction target Revenue from the ETS is assumed to be recycled with a lump and a delayed start of climate policies in 2026, the model finds sum transfer to the representative agent, and firms are per- that total value added declines up to 56 percent in the worst af- fectly competitive. Furthermore, the model takes into account fected industries between 2028 and 2030 (figure 3.4). A sum- an existing US$5 per ton emission tax on coke, gasoline, and mary table of the effect on change in value added between other refined products. Baseline GDP growth in the model is 2028 and 2030 for all sectors in each of the four scenarios is set at 3.5 percent per year. provided in appendix D. > > > FIGURE 3.4. - Decline in value added in the worst affected sectors between 2028 and 2030 in the delayed scenario with 51 percent GHG emission reduction (scenario D) Crude petroleum and natural gas 55.7% Gasoline and other fuels 39.8% Coal and lignite 38.5% Other agriculture 35.4% Live animals and animal products 35.4% Sewerage and waste 34.2% Meat and fish 32.7% Household gas 31.6% Diary products 31.1% Electricity 30.4% Metal ores 55.7% 0% 10% 20% 30% 40% 50% 60% Source: Data obtained from DNP CGE model. Note: ETS = emission trading system; GDP = gross domestic product; GHG = greenhouse gas. Results show outcomes under adverse conditions, assuming little or no technological change in the economy. 27 We note already here that these strong reductions in output in GHG intensive sectors may be partially mitigated by adaptation in the economy, for example due to techno- logical change. Since only little technological change is imposed in the CGE modeling, estimates on GDP impact are likely on the high end (i.e., assuming that not much technological change can or will take place). We discuss this in more detail at the end of the chapter. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 33 One important caveat with our approach is that our CGE mod- aggregation for commercial loans.28 Results are reported as el does not impose any technological change in the simula- differences with respect to the baseline smooth 20% scenario tions and can thereby overestimate economic impacts when sector. The SYSMO model estimates nonperforming loans in the economy has time to adjust, making them most useful to four loan categories, comprising commercial loans, mortgage analyze impacts over a short time. The CGE model by DNP is loans, consumer loans, and microcredits. For the analysis we based on input-output tables for the current economy in Co- assume constant balance sheets, implying that banks do not lombia, which means that the model does not fully capture change their sectoral exposures significantly until the start of technological change such as firm level adaptation and elec- the shock window in 2028. If banks do anticipate transition trification of end uses. Hence, using this model implies that risks and adjust their portfolio in the meantime, this may lead our estimates for sector level impact are most accurate when to lower exposures and a lower banking sector impact. decarbonization occurs over a very short time span, such as 1 year, but it may in-creasingly overestimate any effects We find that especially the transition scenarios with higher tar- when looking at longer time spans. Thus, this approach es- gets have the potential to negatively affect banks their capital pecial-ly overestimates outcomes in our smooth scenarios in position. Losses to the banking sector are very limited in a which the Colombian economy has 10 years to adjust. There- more delayed scenario that has a low GHG emission reduc- fore, our estimates represent an upper bound of losses under tion target of 20 percent, which leads to no tangible impact on adverse economic conditions in which Colombia’s economy nonperforming loans and capital ratios of Colombian banks has little time to adapt. Announcing and implementing climate (figure 3.5, scenario B). The main reason is that in this sce- policies early could substantially reduce any impacts shown in nario, the only severely affected sectors are the fossil fuel and our modeling. waste management sectors, to which the Colombian banking sector has almost no loan exposures. For the scenarios with a GHG reduction target of 51 percent, however, we find siz- Banking sector stress able impacts on bank’s nonperforming loans and capital ratios (figure 3.5, scenarios C and D). Higher impacts during the fi- nal years before 2030 can be explained by the observation To investigate the impact of transition scenarios on the fi- that many more sectors start to be affected when reductions nancial sector we obtain estimates from the Banco de la have been achieved in the fossil fuel sector and reductions República SYSMO model (Gamba and others 2017). This have to be obtained elsewhere, including in manufacturing model, amongst others, links changes in value added and un- and agriculture. In the most severe scenario (D) we find that employment to changes in nonperforming loans in the bank- nonperforming loans can increase between 0.5 percent and ing sector. We estimate the impact of our three more severe 1.9 percent across loan categories, with CARs for Colombian scenarios on the Colombian banking system during a 2-year banks declining with more than one percentage point during shock window, covering 2028 and 2029, using a sectoral dis- the two year shock period due to the credit shock alone. 28 We align our analysis with the timeframe inherent in the SYSMO model, which is two years. We use 2028 and 2029 as they represent the two years with the highest GDP growth impact in all four scenarios. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 34 > > > FIGURE 3.5. - Estimated impact on nonperforming loans and capital adequacy for transition risk scenarios with varying severity, compared to the smooth scenario with 20 percent GHG emission reduction (scenario A) (B) Delayed 20% NPL increase CAR 2.0% 16% 1.5% 12% 1.0% 8% 0.5% 4% 0.0% 0% Domestic Foreign Commercial Mortgage Consumer Micro Pre-shock CAR Post-shock CAR 9% minimum CAR (C) Smooth 51% NPL increase CAR 2.0% 16% 1.5% 12% 1.0% 8% 0.5% 4% 0.0% 0% Domestic Foreign Commercial Mortgage Consumer Micro Pre-shock CAR Post-shock CAR 9% minimum CAR (D) Delayed 51% NPL increase CAR 2.0% 16% 1.5% 12% 1.0% 8% 0.5% 4% 0.0% 0% Domestic Foreign Commercial Mortgage Consumer Micro Pre-shock CAR Post-shock CAR 9% minimum CAR Source: Banco de la República SYSMO model; staff calculations. Note: NPL = nonperforming loan; CAR = capital adequacy ratio. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 35 Results show large differences in decarbonization vulnerabili- vision regarding climate risks and specific attention to banks ties in credit portfolios between banks. We calculate the esti- that are at relatively high risk. We do note that the banks with mated loan losses as a fraction of total assets, based on the a high vulnerability may also have a high potential to contrib- total loan exposures of each bank to the respective sector at ute to greening the Colombian economy, given their focus on a 2-digit level (83 sectors). The three most vulnerable banks relevant transition industries. They could do so, for example, have estimated loan losses ranging between 1.8 and 2.7 per- by engaging with their clients, by adjusting their origination cent of total assets in the most severe scenario (D) while the practices, or by offering green products (such as products tai- three least vulnerable banks have estimated loan losses rang- lored to finance energy savings technologies or other clean ing between only 0.1 and 0.6 percent of total assets (figure technologies). 3.6). This result points to the importance of risk-based super- > > > FIGURE 3.6. - Cumulative loan losses for transition scenarios per bank (percentage of total assets) 3.0 2.5 Percentage of total assets 2.0 1.5 1.0 0.5 0 #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12 #13 #14 #15 #16 #17 #18 #19 #20 Delayed 20% Smoorth 51% Delayed 51% Source: Staff calculations. Results should be interpreted with some caution. Both our economic and financial modeling rest on important assumptions. In particular, the CGE model implicitly assumes limited adaptive capacity of the Colombian economy. Because it does not impose technological change, it does for example not account fully for the emergence of new (renewable) sectors. If the adaptive capac- ity of the Colombian economy is higher than modeled, this could lead to a lower impact on GDP and also a lower impact on the banking sector. This mitigating effect of economic adaptation is likely the most pronounced for the smooth scenarios, in which poli- cies are stretched over a longer time period, giving the economy and the financial sector more time to adjust. Also, our modeling assumes that deforestation emissions are cut in line with the NDC targets. If this would not be feasible, the burden of achieving those targets would be but on other activities. Furthermore, in our financial modeling we model outcomes under adverse condi- tions that include no profitability over the stress testing horizon. Also, due to data limitations we do not include indirect exposures to transition-sensitive sectors abroad. This could lead to an underestimation of transition risks to Colombian banks with substantial indirect exposures. Finally, there may be diversification benefits for banks that have loans toward sectors that benefit from a transi- tion to a low carbon economy, such as renewables and real estate that is energy efficient. Decreased losses on such exposures may mitigate some of the adverse impacts in transition-sensitive exposures. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 36 4. > > > Risk >>> Conclusions and policy recommendations Main risks and vulnerabilities The Colombian banking sector is vulnerable to gradual and more acute risks—stemming from both transition and physical risks. On the one hand, some risks (I and II) are mostly gradual in nature that are expected to play out over a relative long-time horizon. These risks are primarily relevant from a business model perspective, where banks that do not adjust to changing circumstances (for example, in pricing and loan origination practices) could become less profitable over time, eroding their capacity to replenish financial buffers when needed. In these scenarios, more abrupt shocks are possible when prices of certain assets rapidly change because of better market understanding (for example, rapid decreasing real-estate prices in coastal areas). On the other hand, more acute risks (III to V) could play out over a shorter time, including natural disasters and sudden tightening of climate policies in the coming decade (until 2030). These more acute risks are the risks that have the highest potential for banking sector stress in the next decade (table 4.1). T A B L E 4 . 1 . - Summary of main climate related risks for the Colombian banking sector Likelihood Potential for banking sector stress Channels I. Gradually increasing carbon • Increasing loan losses in transition-sensitive sectors Medium Low price and climate policies • Value of commercial real estate II. Gradually increasing • Increasing loan losses in vulnerable sectors temperature and changing High Low/medium (e.g., agriculture) weather-patterns • Increasing loan losses in transition-sensitive sectors III. A sudden tightening of Low/medium Medium/Large • Value of commercial real estate climate policies • Macroeconomic effects • Real estate, corporates, house-holds in affected areas IV. Severe flood Medium Medium • Sovereign credit downgrades • Real estate, corporates, households in affected areas V. Severe flood plus recession Low/medium Large • Sovereign credit downgrades (double shock) • Macroeconomic effects Source: World Bank staff. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 37 Our quantitative analysis finds that there are climate-related Results per bank show large differences in climate-related risks scenarios that can lead to substantial stress for the Co- vulnerabilities in credit portfolios between banks, pointing to lombian banking sector, however, only in adverse (unlikely) the importance of risk-based supervision. A few banks are scenarios. We have investigated flood risks and transition sce- substantially (about two to three times) more vulnerable to narios that are in line with the Colombian NDC targets for 2030. flood hazards than most others because of high exposures in more rural areas. Also, significant heterogeneity is found in a. For flood risk we find that that flood scenarios can lead to the exposure of individual banks to transition-sensitive indus- declines in capital adequacy that could moreover coincide tries, ranging between 1 percent and 26 percent of their credit with other shocks. Although our estimates suggest that portfolio. These observations support a risk-based approach the CAR of Colombian banks can decline substantially, to addressing climate-related risks into SFC supervision. the scenarios with the historic La Niña and the severe floods by themselves do not cause many banks to drop below the required minimum CAR of 9 percent. For those Financial impact scenarios we find an average decline in the capital ad- equacy ratio (CAR) for Colombian banks between 0.3 and 1.1 percentage points. In the double shock scenario, the Based on these findings, the World Bank has identified sev- average decline in the CAR increases to 3.2 percentage eral short-term and medium-term actions to improve climate points. This scenario, however, represents the combina- risk identification and mitigation in the banking sector. In gen- tion of a flood shock with a generic economic recession eral, the SFC could adopt a risk-based supervision approach and, hence, both the most severe and most unlikely sce- for climate-related risks and continuously improve disclosures nario (but not impossible either). Results differ strongly (both by nonfinancial corporates and by financial institutions) per bank, with loan losses for individual banks ranging and data availability. The latter is also important to assess ex- between 0.2 percent of total assets for the least vulner- posures through affiliated entities abroad, for which the SFC able bank to 2.2 percent for the most vulnerable one in the could collaborate with host supervisors to conduct a cross- most severe flood scenario. country climate vulnerability assessment. For physical risks, the SFC could perform deep dives at the most vulnerable b. For transition risks we find that decarbonization scenar- institutions, promote capacity building throughout the sector, ios can materialize before 2030 that lead to substantial and encourage further development of insurance markets. losses in the banking sector. We estimate that a delayed Regarding transition risks, the SFC can work with the bank- scenario with a high GHG reduction target could lead to ing sector and other relevant authorities to address risks in aggregated loan losses for Colombian banks that range a timely and forward-looking manner. This requires dialogue between 0.1 percent of total assets for the least vulner- between public and private stakeholders on how climate miti- able bank and 2.7 percent for the most vulnerable one. gation and adaptation measures will take shape, so that banks This scenario could materialize when climate policies are can address them early on (for example, in engagement with introduced late (from 2025), and these policies aim for a customers, origination practices, and pricing). It also requires high GHG emissions reduction of 51 percent in 2030. As capacity building in the banking sector, potentially including part of this modeling, we assume that the adaptive capac- new tools such as scenario analysis and bottom-up stress ity in the Colombian economy is limited over a short (2 testing. This is particularly urgent with respect to domestic year) time window. Our transition impact estimates should transition risks, as substantial policy efforts to decarbonize the be viewed as estimates based on adverse outcomes in Colombian economy can already be expected this decade. the tail for the probability distribution and not as fore- See table 4.2 for an overview of recommendations. casts of expected outcomes. They are also conservative since they cover a two-year time frame, with more losses that could accumulate before and after the shock period. Moreover, we focus on the credit portfolio only. Losses on other assets could aggravate outcomes. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 38 > > > T A B L E 4 . 2 . - World Bank recommendations to the SFC and other stakeholders. Recommendation Timing Agency General Issue guidelines on governance, risk management, and climate risk disclosure to the banking sector. ST SFC Promote the development of forward-looking climate risk tools in the banking sector, including scenario analysis and stress ST/MT SFC testing. Incorporate climate risks as part of the risk-based supervision of banks.a MT SFC Continue to encourage climate risk disclosure by nonfinancial firms in Colombia, and support the improvement and use of MT SFC climate-risk data at the firm level. Conduct a cross-country climate vulnerability assessment with host supervisors, including improving data collection on SFC, host MT spatial and sectoral exposures through related entities. supervisors Physical risks Identify the information and data needed to carry out physical risk assessments tailored to individual institutions’ exposure. ST/MT SFC Promote technical capacity building in the banking sector to understand and manage physical risks, covering both disaster ST/MT SFC risks and gradual changes in climatic conditions (for example, through a platform). Promote the development and implementation of mechanisms to mitigate physical risks and their impacts, such as disaster SFC, MT risk insurance.b Government Transition risks Promote capacity building in the banking sector to understand and manage transition risks, specifically focusing on ST and ST SFC MT transition risks (that is, 2030 NDC targets). Promote further dialogue between climate policy makers and the financial sector including banks, other investors, and SFC, BR, ST/MT financial authorities. Government Provide more detailed guidance to the financial sector on how de-carbonization policies will be implemented and on what is Government, ST/MT the time line for implementation (for example, a road map until 2030 to allow banks to better align their portfolios). SFC Encourage better data collection to perform firm-level stress tests by authorities and banks. Specifically, this would require ST/MT SFC firm-level GHG emission data for nonfinancial firms in Colombia. Note: BR = Central Bank of Colombia; NDC = nationally determined contribution; SFC = Financial Superintendence of Colombia; ST = short term (within one year) and MT = medium term (within one to three years). a. This could include requiring increased supervisory reporting, addressing the risk in the banks’ Internal Capital Adequacy and Assessment Process, and ad- dressing the risk in on-site supervision (including in board-level conversations). b. We note that sometimes the most cost-effective adaptation lies in preventing the creation of new risks by ensuring that land-use planning and infrastructure regulations take disaster risks into account. Banks could play a role here by integrating disaster and flood risks into their loan origination process (and requir- ing alignment with land-use plans and infrastructure regulations). Development of insurance markets could include (a) strengthening the legal framework including on the use of parametric insurance and (b) improving the availability of disaster insurance data. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 39 >>> References and Suggested Readings Bank of England. 2019. “The 2021 Biennial Exploratory Scenario on the Financial Risks from Climate Change.” Discussion Paper. Bank of England, London. BCBS (Basel Committee on Banking Supervision). 2021. “Climate-Related Financial Risks— Measurement Methodologies.” Report. Bank for International Settlements, Basel. Battiston, Stefano, Antoine Mandel, Irene Monasterolo, Franziska Schütze, and Gabriele Visen- tin. 2017. “A Climate Stress-Test of the Financial System.” Nature Climate Change 7 (March): 283–288. Bernal-Ramírez, Joaquin, and Jose Antonio Ocampo. 2020. “Climate Change: Policies to Man- age Its Macroeconomic and Financial Effects.” Working Paper on Economics. Banco de la República, Bogotá. Bolton, Patrick, Morgan Despres, Luiz Awazu Pereiera da Silva, Frédéric Samama, and Romain Svartzman. 2020. The Green Swan: Central Banking and Financial Stability in the Age of Cli- mate Change. Basel: Bank for International Settlements. Burns, Andrew, Calvin Djiofack Zebaze, Dinar Prihardini. 2018. “Modeling Macroeconomic Im- pacts and Global Externalities.” Good Practice Note 7, Energy Subsidy Reform Assessment Framework (ESRAF). ESMAP and World Bank Group, Washington, DC. Cai, Wenju, Guojian Wang, Agus Santoso, Michael J. McPhaden, Lixin Wu, Fei-Fei Jin, and others. 2015. “Increased Frequency of Extreme La Niña Events Under Greenhouse Warming.” Nature Climate Change 5 (2):132–137. Campiglio, Emanuele, Yannis Dafermos, Pierre Monnin, Josh Ryan-Collins, Guido Schotten, and Misa Tanaka. 2018. “Climate Change Challenges for Central Banks and Financial Regula- tors.” Nature Climate Change 8 (6): 1–13. Cardona, Omar Dario, Gabriel Andres Bernal, Daniela Zuloaga Romero, Maria Alejandra Es- covar Bernal. 2017. “Modelación probabilista de inundaciones en La Mojana.” Technical report. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 40 CEPAL (Comisión Económica para América Latina y el Caribe). 2012. Valoración de daños y pérdidas. Ola invernal en Colombia, 2010–2011. Bogotá: Misión BID—CEPAL. Čihák, Martin. 2014. “Stress Tester: A Toolkit for Bank-by-Bank Analysis with Accounting Data.” In A Guide to IMF Stress Testing, edited by Li Lian Ong, 17–44. Washington, DC: International Monetary Fund. Delgado Ricardo, Thomas B. Wild, Ricardo Arguello, Leon Clarke, and German Romero. 2020. “Options for Colombia’s mid-century deep decarbonization strategy.” Energy Strategy Reviews 32 (November). DNB (Dutch Central Bank). 2017. “Waterproof? An Exploration of Climate-Related Risks for the Dutch Financial Sector.” Report. Dutch Central Bank, Amsterdam. DNP (National Planning Department). 2018. “Índice Municipal de Riesgo de Desastres Ajustado por Capacidades.” Report. DNP, Bogotá. Feyen, Erik, Robert Utz, Igor Zuccardi Huertas, Olena Bogdan, and Jisung Moon. 2020. “Macro- Financial Aspects of Climate Change.” Policy Research Working Paper No. 9109. World Bank, Washington, DC. FSB (Financial Stability Board). 2020. “The Implications of Climate Change for Financial Stabil- ity.” FSB, Basel. Gamba, Santiago, Oscar Jaulín, Angélica Lizarazo, Juan Carlos Mendoza, Paola Morales, Dan- iel Osorio, and Eduardo Yanquen. 2017. “SYSMO I: A Systemic Stress Model for the Colombian Financial System.” Working Paper on Economics. Banco de la República, Bogotá. Good, Peter, Jason Lowe, Jeff Ridley, Jonathan L. Bamber, Tony Payne, Brandon Keen, J.C. Stroeve, and others. 2014. “An Updated View of Tipping Points and the Relevance for Long- Term Climate Goals.” Technical Report, NERC Senior Knowledge Exchange Fellowship. Grippa, Pierpaolo, Jochen Schmittmann, and Felix Suntheim. 2019. “Climate Change and Fi- nancial Risk.” Finance & Development 56 (4): 26–29. IDEAM (Institute of Hydrology, Meterology and Environment Studies). 2012. “Evaluación, análi- sis y seguimiento a las Afectaciones por inundaciones asociadas al Fenómeno de la Niña 2010– 2011.” Technical report. Bogotá. IDEAM (Institute of Hydrology, Meterology and Environment Studies). 2015. “Nuevos Escenari- os de Cambio Climático para Colombia 2011–2100.” IDEAM, Bogotá. IDEAM (Institute of Hydrology, Meterology and Environment Studies), PNUD (United Nations Development Programme), MADS (Ministry of Environment and Sustainable Development), DNP (National Planning Department), Cancillería (Ministry of Foreign Affairs). 2017. Análisis de vulnerabilidad y riesgo por cambio climático en Colombia. Tercera Comunicación Nacional de Cambio Climático. Bogotá: IDEAM. IMF (International Monetary Fund). 2017. “The Effects of Weather Shocks on Economic Activ- ity: How Can Low-Income Countries Cope?” In World Economic Outlook: Seeking Sustainable Growth—Short-Term Recovery, Long-Term Challenges. Washington, DC: IMF. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 41 Klomp, J. 2014. “Financial Fragility and Natural Disasters: An Empirical Analysis.” Journal of Financial Stability 13 (2014): 180-192. Melo, S. F., L. Riveros, G. Romero, J. C. Farfán, A. Alvarez-Espinoza, and C. Díaz. 2019. “Esti- mación de impactos del cambio climático en el sector agricultura y seguridad alimentaria.” DNP, Archivos de Economía, documento 504. Bogotá. NDC Colombia. 2020. “Actualización de la Contribución Determinada a Nivel Nacional de Co- lombia (NDC).” Government of Colombia, Bogotá. NGFS (Network for Greening the Financial System). 2020a. “Overview of Environmental Risk Analysis by Financial Institutions.” Technical document. NGFS, Paris. NGFS (Network for Greening the Financial System). 2020b. “NGFS Climate Scenarios for Cen- tral Banks and Supervisors.” Report. NGFS, Paris. OECD (Organisation for Economic Co-operation and Development). 2016. “Colombia: Review of the Financial System.” Report. OECD, Paris. Rabatel, Antoine, Jorge Luis Ceballos, Natan Micheletti, Ekkehard Jordan, Michael Braitmeier, Javier González, Nico Mölg, and others. 2018. “Toward an Imminent Extinction of Colombian Glaciers?” Geografiska Annaler: Series A, Physical Geography 100 (1): 75–95. Standard & Poor’s. 2015. “The Heat Is On: How Climate Change Can Impact Sovereign Rat- ings.” November 25, S&P Global Ratings. Standard & Poor’s. 2019, “Credit Trends: The Cost of a Notch.” March 16, S&P Global Ratings. SFC (Superintendencia Financiera de Colombia). 2019. “Riesgos y oportunidades del cambio climático.” Superintendencia Financiera de Colombia, Bogotá. UNGRD (Unidad Nacional para la Gestión del Riesgo de Desastres). 2018. Atlas de riesgo de Colombia: revelando los desastres latentes. UNGRD, Bogotá. Ward, P. J., H. C. Winsemius, S. Kuzma, M. F. P. Bierkens, A. Bouwman, H. de Moel, and A. Díaz Loaiza. 2020. “Aqueduct Floods Methodology.” Technical Note. World Resources Institute, Washington, DC. https://www.wri.org/publication/aqueduct-floods-methodology. Winsemius H.C., L. P. H. Van Beek, G. Jongman, P. J. Ward, and A. Bouwman. 2013. “A Frame- work for Global River Flood Risk Assessment.” Hydrolology and Earth System Sciences 17 (5): 1871–92. World Bank. 2019. “Beyond Stranded Assets—Climate Strategies of Fossil-Fuel Dependent Countries.” Report, World Bank Sovereign ESG Resources. World Bank, Washington, DC. Vermeulen, Robert, Edo Schets, Melanie Lohuis, Barbara Kölbl, David-Jan Jansen, and Willem Heeringa. 2019. “The Heat Is On: A Framework Measuring Financial Stress Under Disruptive Energy Transition Scenarios.” DNB Working Paper No. 625. De Nederlandsche Bank NV, Am- sterdam.  EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 42 >>> Appendixes Appendix A. Flood regression model We investigate the impact of large-scale floods to the Colombian banking sector by performing an event study on past large-scale flooding. To calibrate the credit risk shock, we estimate a regres- sion model to estimate the impact of historical floods on provisioning. Specially, we focus on the 2010–2011 floods that are connected to La Niña and are among the most economically damaging disasters that occurred in Colombia over the past decades. This event falls well within the sample period for which there is credit risk data available, where we cover 15 years from 2005 to 2019. We identify the shock size by employing a difference-in-difference (DID) modeling strategy, which compares the provisions growth rate between affected regions and nonaffected regions. In this way, we control for macroeconomic effects that are common for the entire country. We estimate four variants of the model. Models (1) and (2) use a traditional DID approach also employed in previous exercises. Models (3) and (4) instead weigh impacts based on the economic losses as a fraction of GDP. All models provide significant and consistent shock estimates, indicating that exposures in more severely affected areas have experienced higher provisioning than areas that were less severely affected (or not at all). The specifications are provided in table A.1, while the re- gression results are provided in table A.2. Most exposure and credit risk data are obtained through the SFC, while data on economic impact are obtained from CEPAL (2012). EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 43 > > > T A B L E A . 1 . - Regression specifications (1) DID, time and bank fixed effects ∆PROVISIONSi,t,d = SHOCKd,t + Di + Dt (2) DID, with bank*time fixed effects ∆PROVISIONSi,t,d = SHOCKd,t + Di + Di,t (3) DID, with flood vulnerability factor (FVF) ∆PROVISIONSi,t,d = FVFd * SHOCKt + Di + Dt (4) DID, with bank*time fixed effects and flood vulnerability factor (FVF) ∆PROVISIONSi,t,d = FVFd * SHOCKt + Di,t Source: World Bank staff. Note: in these specifications, the delta provisions are the percentage point increase in provisions as a fraction of total loans in the department compared with the previous quarter. The flood vulnerability factor (FVF) is defined as the economic damage as a fraction of GDP per department as obtained from CEPAL (2012). The shock variable is 1 when La Niña–related floods affected the specific departments over a period of six quarters following the initial floods. Fur- thermore, D denotes a dummy variable, while t stands for time (quarter), i for bank, and d for department. > > > T A B L E A . 2 . - Regression results Binary shock variable Weighted shock variable (1) (2) (3) (4) Shock estimate 0.0020* 0.0026** 0.0011* 0.0012** Time FE Yes No Yes No Bank FE Yes No Yes No Bank*time FE No Yes No Yes Source: World Bank staff, based on SFC data. Note: FE = fixed effects; results obtained using robust standard errors. p < 10% (*); p < 5% (**). EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 44 Appendix B. Flood vulnerability > > > T A B L E B . 1 . - Riverine flood risk in Colombia with a 50-year return Source: Ingeniar in UNGRD 2018. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 45 Appendix C. Stress test model To estimate the size of climate-related risks to the Colombian banking sector, we employ a basic stress testing model and extend it to provide results using granular sectoral and regional data. The model builds on stress tester 3.0 by Martin Čihák (2014) and uses balance sheet and profit and loss data covering 21 banks for which detailed sectoral and regional breakdowns are available. Results are estimated per bank, per group of banks (state owned, domestic, foreign), and aggregated for the entire sector. Five smaller banks are excluded from the dataset owing to unavailability of a sectoral breakdown. The excluded banks represent 1.6 percent of total assets in the dataset. All data are for year-end 2019. An overview of the model’s characteristics is presented in table C.1. > > > T A B L E C . 1 . - Characteristics of climate stress testing model • Banking sector • Top-down solvency risk assessment Main characteristics • Individual bank and aggregated balance sheets • Credit exposures at granular spatial and sectoral levels • The main output is the decline in total bank capital and losses as a fraction of total assets / total loans • Credit losses from banks’ loan portfolios, excluding off-balance-sheet credit commitments Risks assessed • Accounting and market-value losses from credit spread adjustments of banks’ holding of sovereign debt securities • Balance sheet • Risk-weighted assets • Credit risk data (nonperforming loans, provisions, collateral) Data inputs for model • Sectoral structure of lending • Regional structure of lending • Share of sovereign exposures in banking book • Average provisioning rate for nonperforming loans (25%) Key parameters • Average duration of bank’s sovereign bond portfolios (4.2 years) • Current sovereign credit rating (S&P BBB-) • Regional and sectoral shocks (% of performing loans becoming nonperforming loans or provisions taken) Key calibration inputs • Change in sovereign credit spread Source: World Bank staff. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 46 Appendix D. DNP CGE model results > > > T A B L E D . 1 . - Change in value added per sector between 2028 and 2030 Smooth Delayed Smooth Delayed (NDC: (NDC: (NDC: (NDC: 20%) 20%) 51%) 51%) Productos de café 31.3% 48.9% 15.9% 41.7% Otros productos agricolas -0.4% -4.2% -22.1% -35.4% Animales vivos y productos animales y productos de la caza -0.9% -5.1% -21.7% -35.4% Productos de silvicultura y extraccion de madera y actividades conexas 4.1% 2.6% -9.1% -16.5% Productos de la pesca y la acuicultura y servicios relacionados 3.6% 1.8% -8.9% -16.8% Carbon mineral -5.1% -11.9% -22.7% -38.5% Petroleo crudo y gas natural y minerales de uranio y torio -6.7% -12.7% -36.3% -55.7% Minerales metalicos 2.0% -0.6% -15.8% -26.7% Minerales no metalicos 3.6% 1.8% -10.5% -18.6% Carnes y pescados 0.4% -3.1% -20.0% -32.7% Aceites y grasas animales y vegetales 3.0% 0.9% -13.2% -22.8% Productos lacteos 1.2% -1.8% -19.2% -31.1% Productos de molineria y almidones y sus productos 2.2% -0.4% -15.3% -25.9% Productos de cafe y trilla 31.6% 49.7% 15.6% 40.2% Azucar y panela 3.1% 1.1% -13.0% -22.4% Cacao y chocolate y productos de confiteria 3.2% 1.3% -14.1% -23.6% Productos alimenticios ncp 3.8% 2.2% -10.7% -19.0% Bebidas 4.0% 2.4% -8.4% -16.1% Productos de tabaco 4.5% 3.2% -6.5% -13.1% Fibras textiles naturales e hilazas e hilos y tejidos de fibras textiles e incluso afelpados 5.5% 4.8% -4.2% -9.1% Articulos textiles excepto prendas de vestir 6.2% 5.7% -3.4% -7.5% Tejidos de punto y ganchillo y prendas de vestir 5.2% 4.2% -4.5% -9.9% Curtido y preparado de cueros y productos de cuero y calzado 4.8% 3.6% -6.6% -12.7% Productos de madera y corcho y paja y materiales trenzables 4.1% 2.6% -8.9% -16.4% Productos de papel y carton y sus productos 4.4% 3.1% -7.4% -14.5% Edicion e impresion y articulos analogos 4.4% 3.1% -7.0% -13.6% Coque y semicoqu de hulla y de lignito o de turba y carbon de retorta y alquitranes 21.7% 14.4% 11.3% 7.6% Gasolinas y otros combustibles -2.1% -6.9% -24.7% -39.8% Diesel -5.7% -12.1% -29.4% -47.1% Otros refinados -0.6% -4.5% -27.7% -42.4% Sustancias y productos quimicos 4.8% 3.7% -7.8% -14.6% Productos de caucho y de plastico 4.0% 2.5% -8.4% -15.9% Productos minerales no metalicos 5.1% 4.2% -6.7% -13.1% Productos metalurgicos basicos excepto maquinaria y equipo 2.8% 0.6% -13.6% -23.5% Maquinaria y equipo 4.1% 2.6% -10.2% -17.9% Otra maquinaria y suministro electrico 4.8% 3.7% -10.0% -15.5% Equipo de transporte 5.1% 4.1% -5.9% -12.6% Muebles 6.0% 5.6% 0.2% -8.5% Otros bienes manufacturados ncp 6.7% 6.7% 7.9% -3.3% EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 47 Energia electrica 1.8% -0.8% -30.2% -30.4% Gas domiciliario 0.1% -3.8% -32.8% -31.6% Agua 5.0% 4.0% -1.2% -9.2% Trabajos de construccion y construccion y reparacion de edificaciones y servicios de arrendamiento de equipo con operario 3.9% 2.3% -12.5% -17.1% Trabajos de construccion y construccion de obras civiles y servicios de arrendamiento de equipo con operario 3.6% 1.7% -13.3% -17.8% Comercio 3.4% 1.5% -15.3% -19.0% Servicios de reparacion de automotores y de articulos personales y domesticos 4.0% 2.4% -9.9% -15.0% Servicios de alojamiento y suministro de comidas y bebidas 3.9% 2.3% -14.8% -18.9% Servicios de transporte terrestre 5.8% 5.3% 5.1% -4.8% Servicios de transporte por via acuatica 9.1% 10.2% 38.2% 16.8% Servicios de transporte por via aerea 6.5% 6.3% 12.9% 0.5% Servicios complementarios y auxiliares al transporte 4.4% 3.0% -6.5% -12.7% Servicios de correos y telecomunicaciones 5.1% 4.1% -0.9% -9.0% Servicios de intermediacion financiera y de seguros y servicios conexos 4.2% 2.8% -6.2% -12.5% Servicios inmobiliarios y de alquiler de vivienda 5.2% 4.2% 0.3% -8.1% Servicios a las empresas excepto servicios financieros e inmobiliarios 4.4% 3.0% -8.2% -13.9% Administracion publica y defensa y direccion y administracion y control del sistema de seguridad social 5.7% 5.1% 2.2% -6.8% Servicios de ensenanza de mercado 5.7% 5.0% 2.9% -6.3% Servicios de ensenanza de no mercado 5.7% 5.1% 3.1% -6.1% Servicios sociales y de salud de mercado 5.2% 4.3% -1.2% -9.1% Servicios de alcantarillado y eliminacion de desperdicios y saneamiento y otros servicios de proteccion del medio ambiente -2.7% -8.2% -37.3% -34.2% Servicios de asociaciones y esparcimiento y culturales y deportivos y otros servicios de mercado 5.0% 4.0% -2.3% -9.9% Servicios de asociaciones y esparcimiento y culturales y deportivos y otros servicios de no mercado 5.3% 4.5% -1.3% -9.2% Servicio domestico 5.9% 5.4% 5.9% -4.2% Source: Data obtained from DNP CGE model. Note: NDC = nationally determined contribution. EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT <<< 48