## 1mexea2022008 — Executive Summary and Key Findings

## Source details

**Canonical URL:** [1mexea2022008 — Executive Summary and Key Findings](https://www.imf.org/-/media/files/publications/cr/2022/english/1mexea2022008.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2022/english/1mexea2022008.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2022/english/1mexea2022008.pdf.json)

---

### Executive summary: introduction and scope
- Mexico is exposed to both transition and physical risks from climate change.
- Mexico’s total energy supply is dominated by fossil fuels and non-renewable energy sources.
- Mexico is the second largest emitter of greenhouse gases (GHG) in Latin America.
- Emissions are highly concentrated in a few economic sectors that play an important role in the Mexican economy.
- The Mexican financial sector has sizable exposures to these emission intensive sectors.
- Mexico is exposed to acute physical risks (floods and tropical cyclones), droughts, heatwaves, and chronic risks.
- FSAP analysis approach:
  - Transition risk: integrated micro-macro approach; two NGFS-motivated scenarios plus a baseline current unchanged policies scenario.
  - Physical risk: acute risks associated with tropical cyclones and floods using direct-damage modelling combined with macroeconomic and financial-sector impact modelling.
  - Transition scenarios: Global action (orderly, below 2 degrees) and Delayed-uncertain (disorderly).
  - Innovations: stochastic financial framework for delayed transition tail risks; physical-risk focus on an extreme season (sequence of events) rather than single events.
- Analysis horizon for corporate transition simulations: five years (till 2026).
- Analysis horizon for macro-financial physical scenarios: three years till 2024.

### Key findings — transition risk (aggregate and sectoral)
- Under global action scenario aggregate impact on financial sector is modest; some sectors and banks are more exposed.
- Delayed-uncertain (disorderly) transition flags potentially non-negligible risks to corporates and banks; delays increase tail risks to financial stability.
- Earlier action reduces tail risks.
- Quantitative examples and sectoral concentrations:
  - About 38 percent of the total CO2 emissions in 2018 was attributable to the utilities/power generation sector.
  - About 25 percent of the total CO2 emissions in 2018 was attributable to the manufacturing sector.
  - About 16 percent of the total CO2 emissions in 2018 was attributable to the transportation sector.
  - Mining sector (including oil, gas, coal extraction) accounts for 11 percent of overall CO2 emissions.
  - All other sectors combined constitute about 10 percent of the total CO2 emissions.
  - About 37 percent of the commercial banking system’s total corporate credit is concentrated in emission intensive sectors.
  - Largest banking exposure: manufacturing sector accounting for about 15 percent of total credit.
  - Exposures to the mining sector are 4 percent of total credit.
- Under global action (static balance-sheet assumptions):
  - System average cumulative impact on bank capital due to expected credit losses: about 0.35 percent.
  - System wide delta PD rises to above 0.7 percent by 2026; delta PDs for some banks reach almost 1 percent.
  - Diversification: about 37 percent of banking sector portfolio belongs to emission intensive sectors.
- Under delayed-uncertain pathways (time-varying distributions and tail risk amplification):
  - Example: chemicals sector output impact increases from about 10 percent (global action) to 23 percent by 2030 if global action is delayed to 2025/26.
  - Relative to acting in year 2023 (maximum possible credit risk around 3 percent), under delayed 2026 risk could increase to as high as 14 percent.
  - Bank delta PDs: under early action in 2023 maximum delta PD across banks ~0.2 percent; under delays to 2026 delta PDs could rise to as high as 1.5 percent.
  - Cumulative bank capital impact under delayed-uncertain path could reach as high as 0.8 percent with non-trivial probability, relative to a maximum impact of around 0.3 percent under global action.

### Modeling framework for transition risk (integrated micro-macro)
- CGE model: IMF-ENV (recursive-dynamic, multi-regional, multi-sectoral).
  - Covers 25 regions (including Mexico) and 37 distinct sectors.
  - Scenarios anchored to NGFS narratives and aligned with IPCC targets.
- Corporate micro-simulation:
  - Corporate data: DataStream and S&P Capital IQ, balance sheet data 2002–2020, average ~100 firms per year, sample firms represent about 42 percent of total outstanding debt of the nonfinancial corporate sector.
  - Only 2 percent of firms (large firms) account for more than 65 percent of outstanding bank credit across 44 commercial banks.
  - Key firm variables: EBIT, sales revenue, COGS, interest expenses, average/effective interest rates, total debt, total assets, current assets, current liabilities.
- Mapping sectoral output to firm financials:
  - Structural link: corporate sales revenues mapped to sectoral output and carbon price paths from IMF-ENV.
  - Recursive evolution for EBIT includes term COSCARB_t = carbon price times emissions (scope 1 only).
  - Elasticity assumptions:
    - Elasticity of sales to output/GVA per sector assumed to be 1.5.
    - Elasticity G (COGS to sales) estimated as 0.97.
  - Carbon price floors by 2030 under global action:
    - $25 tCO2e for LICs
    - $50 tCO2e for MICs
    - $75 for HICs
- Corporate vulnerability indicators constructed:
  - Interest coverage ratio (ICR) = EBIT / interest expenses.
  - Leverage ratio (LR) = total debt / total assets.
  - Current ratio (CR) = current assets / current liabilities.
- Bridge equation (panel fixed effects with logit-transformed PDs):
  - slogit(PD)_{i,t} = alpha_i + beta_1 ICR_{i,t} + beta_2 CR_{i,t} + beta_3 LR_{i,t} + epsilon_{i,t}
  - PDs sourced from Moody’s EDF database for Mexico.
  - Estimated PD sensitivity: leverage ratio most influential, followed by current ratio and interest coverage ratio.

### Bridge aggregation, bank mapping, and capital-impact assumptions
- Bridge-generated scenario-dependent firm-level PD paths aggregated to exposure-weighted sectoral PDs.
- Bank-level delta PD paths computed by weighting sectoral delta PDs by each bank’s credit exposures (assumed constant at starting point).
- Capital-impact calculation assumptions:
  - Considered only expected credit losses under static balance sheet.
  - RWA held constant.
  - Common LGD across all banks; system average implied LGD found to be 53 percent.
  - Tax and dividend payout impact ignored.
  - Resulting estimated impact on bank capital ratios driven solely by loan impairment charges.

### Financial-market channel: jump-diffusion model and delayed-uncertain construction
- Corporate spreads model: affine jump-diffusion (Cox-Ingersoll-Ross with jumps) under martingale measure ℚ:
  - d r_t = κ(μ − r_t + s_t) d t + σ sqrt(r_t) d W_t + d J_t
  - Feller condition: 2 κ μ ≥ σ^2 to keep spreads non-negative.
  - Scaling factor s_t constructed from sectoral outputs to map climate risks into corporate spread evolution.
  - Jumps capture sudden large increases in corporate spreads due to policy uncertainty; diffusion captures continuous volatility with volatility increasing when spreads are higher (sqrt term).
- Calibration and simulation:
  - Calibrated to Corporate Emerging Markets Bond Index (CEMBI) for aggregate Mexican corporate bond spreads.
  - Simulation produced 20,000 stochastic paths at daily frequency until end of 2026.
  - Mapping to firm balance sheets: time-varying spreads translated into firm interest rate shocks in micro-simulations.
- Delayed-uncertain pathway construction (binomial tree / Monte Carlo interpretation):
  - No global action in 2022; for each subsequent year world may act or delay, producing multiple paths (five possible paths to 2026 in stylized setting).
  - Percentile mapping for constructed interest-rate shock paths:
    - Delayed-2023: average across all paths.
    - Delayed-2024: 75th percentile across paths.
    - Delayed-2025: 90th percentile across paths.
    - Delayed-2026: 95th percentile across paths.
  - Appendix assumption: one-time large jump at year-end 2022 calibrated to 99th percentile of rolling 2-year changes in corporate spreads.

### Physical risk: hazard exposure, scenarios, and direct-damage estimates
- Hydrometeorological burden (last two decades): floods and tropical cyclones affected more than 15 million people.
- Geographic concentration of GDP (2020) — six states ~50 percent of national GDP:
  - Mexico City: 17.51 percent
  - Mexico State: 9.15 percent
  - Nuevo León: 7.74 percent
  - Jalisco: 6.92 percent
  - Veracruz: 4.53 percent
  - Guanajuato: 4 percent
- Credit exposure concentration (end of 2021):
  - Total credit: 3,434 billion MXN
  - Mexico City: 1,513 billion MXN (more than 44 percent of total credit)
  - Nuevo León: 13 percent of total credit
  - Jalisco: 6.4 percent
  - State of Mexico: 4.9 percent
  - Sinaloa: 3.2 percent
  - Remaining 27 federal states: 28 percent of total credit exposure
- Scenario set (national-scale return period > 50 years):
  - TC1: tropical cyclone with two landfalls (Yucatan Peninsula Category 4; northeast Mexico tropical storm / Category 1).
  - TC2: two tropical cyclones impacting Veracruz ~10 days apart.
  - TC3: tropical cyclone landfall in Colima impacting Guadalajara (Category 1-2).
  - F1: river flood in Tabasco and surrounding states.
  - ES1: extreme season combining TC1, TC2, TC3, and F1 within a single season.
- Historical-frequency baseline (EMDAT 2000–2021 means):
  - Flash flood mean: 0.55
  - Riverine flood mean: 0.91
  - Tropical cyclone mean: 3.09
  - Total mean: 4.55
- Direct-damage estimates (percent of total national capital stock):
  - ES1-hist: 0.9 percent
  - ES1-cc end-of-century (RCP8.5): 1.9 percent
- Monetary direct-damage range for individual scenarios under historical conditions:
  - From 30 billion MXN (approximately <0.1 percent of capital stock) to 600 billion MXN (approximately 0.7 percent of capital stock).
- Scenario-specific direct damage rates and climate change impacts (Table 3 reproduced values):
  - TC1
    - historical conditions: 0.14 percent
    - climate change conditions mid-century: 0.21 percent — 43 percent
    - climate change conditions end-century: 0.23 percent — 59 percent
  - TC2
    - historical conditions: 0.03 percent
    - climate change conditions mid-century: 0.05 percent — 54 percent
    - climate change conditions end-century: 0.08 percent — 124 percent
  - TC3
    - historical conditions: 0.67 percent
    - climate change conditions mid-century: 1.01 percent — 51 percent
    - climate change conditions end-century: 1.45 percent — 118 percent
  - F1
    - historical conditions: 0.05 percent
    - climate change conditions mid-century: 0.08 percent — 54 percent
    - climate change conditions end-century: 0.13 percent — 145 percent
  - ES1
    - historical conditions: 0.90 percent
    - climate change conditions mid-century: 1.35 percent — 50 percent
    - climate change conditions end-century: 1.89 percent — 110 percent
- Climate-change damage scaling statements:
  - Damages may increase by 40-50 percent by mid-century under RCP8.5.
  - Damages may increase by more than 100 percent by end-of-century under RCP8.5.

### Physical-to-macro and financial transmission (macro-financial modeling)
- Direct damage inputs to macro model (GFM) for ES1:
  - ES1-hist direct damage used in macro model: 0.9 percent
  - ES1-cc direct damage used in macro model: 1.9 percent
- Macro channels and calibration:
  - Direct destruction of physical capital treated as immediate depreciation shock.
  - TFP shocks calibrated at twice the damage rate and assumed highly persistent.
  - Concurrent unemployment shock informed by elasticity of unemployment to changes in capital stock in Mexico.
  - Negative shocks to equity markets informed by elasticity of equity returns to changes in capital stock.
- GDP-level impacts under extreme season scenarios (deviation from baseline, growth terms in 2022):
  - ES1-hist: about 1.4 percent in 2022.
  - ES1-cc end-of-century: as much as 2 percent in 2022.
- Banking-system capital impacts (aggregate CAR deviation from baseline):
  - ES1-hist: aggregate CAR decline about 1 percentage point versus baseline.
  - ES1-cc end-of-century: capital ratio depletion approximately 1.2 percentage points versus baseline.
  - Heterogeneity: at some banks CAR could decline by almost 4 percent under historical conditions and over 4 percent under climate change conditions, relative to baseline.

### Policy recommendations and priorities
- Strengthen data collection and conduct further analysis to understand potential impact of climate risks.
- Enhance disclosure to collect granular data for understanding carbon footprint across the full spectrum of firms, not just listed firms.
- Partner with other government entities, the private sector, and research institutions to leverage expertise in climate and catastrophe modelling for granular, data-driven assessment of physical risks and incorporation into banking-sector resilience analysis.
- Improve and harmonize data collection and modelling to enable more comprehensive and granular analysis of transition and physical risks to the financial system.
  - Responsible Authorities: Banxico, CNBV
  - Time: MT
  - Priority: H
- Strengthen, expand, and accelerate development of climate-related risk analysis and its integration with financial system analysis.
  - Responsible Authorities: Banxico, CNBV
  - Time: MT
  - Priority: M
- Time definitions:
  - “I (immediate)” is within one year;
  - “NT (near-term)” is one–two years;
  - “MT (medium-term)” is three–five years.
- Priority definitions:
  - H: high; M: medium; L: low.
- Additional recommendations:
  - Further strengthen Mexico’s climate and disaster risk finance strategy and explore opportunities to expand insurance for climate-related risks.
  - Develop catastrophe risk insurance markets given current low levels of insurance penetration.
  - Prioritize integrated financial and operational preparedness at local, state, and national levels; leverage private risk capital (e.g., insurance and catastrophe bonds).

### Uncertainties, caveats, and suggested next steps
- Analysis is exploratory and subject to multiple layers of scenario and model uncertainties; not a standard stress test and not focused on quantifying capital needs.
- Key limitations highlighted:
  - Lack of granular exposure data and access to full catastrophe models used in Mexico.
  - Exposure disaggregation assumes capital stock distribution mirrors population distribution.
  - Hazard module: global products not locally calibrated; translating precipitation to flood depth involves simplifying assumptions.
  - Vulnerability curves: regionally calibrated curves applied without local validation; inability to differentiate asset characteristics.
  - Macro model: aggregated national-scale linearized dynamics limit capture of regional non-linear amplifications and indirect channels (e.g., supply-chain disruptions).
  - Transition analysis: simulations assume constant business models and technology over projection horizon.
- Suggested next steps:
  - Strengthen data collection and harmonization; acquire granular exposure and catastrophe-model outputs.
  - Expand scenarios to include droughts, heatwaves, chronic risks, and compound events.
  - Incorporate market-risk channels, consumer portfolios, and additional financial-sector channels into future analyses.
  - Explore sensitivity of results to key modelling assumptions.

*Source: 1mexea2022008 — EXECUTIVE SUMMARY and selected sections of the IMF staff report.*

### EXECUTIVE SUMMARY __________________________________________________________________________ 5

### EXECUTIVE SUMMARY

### Introduction
- Mexico is exposed to both transition and physical risks from climate change.
- Mexico’s total energy supply is dominated by fossil fuels and non-renewable energy sources.
- Mexico is the second largest emitter of greenhouse gases (GHG) in Latin America.
- Emissions are highly concentrated in a few economic sectors that play an important role in the Mexican economy.
- The Mexican financial sector has sizable exposures to these emission intensive sectors.
- Mexico is exposed to acute physical risks (floods and tropical cyclones), droughts, heatwaves, and chronic risks.

### Scope and Methodology
- The FSAP analyzed potential risks to financial stability posed by:
  - Transition risk: using an integrated micro-macro approach and two NGFS-motivated scenarios plus a baseline current unchanged policies scenario.
  - Physical risk: focused on acute risks associated with tropical cyclones and floods using direct-damage modelling combined with macroeconomic and financial-sector impact modelling.
- Transition scenarios analyzed:
  - Global action: orderly transition broadly in line with limiting temperature increase to below 2 degrees centigrade by end of century.
  - Delayed uncertain: disorderly transition as in the NGFS delayed transition scenario.
- Innovation: a novel stochastic financial framework to assess the impact of delayed transition paths on climate tail risks.
- Physical-risk innovation: focus on a severe tropical cyclone/flood season (sequence of extreme events within a single season) rather than single extreme events.

### Key Findings — Transition Risk
- Under the global action scenario where all countries act early, the aggregate impact on the financial sector is modest, though some sectors and banks appear more exposed.
- The delayed uncertain (disorderly) transition scenario flags potentially non-negligible risks to corporates and banks.
- Delays in transition—and the resulting need to catch up with stronger policy action to contain the faster rise in temperatures—increase tail risks to financial stability.
- The impacts get increasingly severe the longer the delay relative to early action.
- Although quantitative estimates are uncertain, the qualitative effect (direction of risk materialization) is clear: earlier action reduces tail risks.

### Key Findings — Physical Risk
- Extreme tropical cyclones and floods could have a material impact, though less likely at a scale to generate systemic financial stress.
- A sequence of extreme tropical cyclone and flood events occurring within a single season based on historical climate conditions could lower GDP by more than one percentage point within the year.
- Preliminary modelling that allows for increases in the severity of these events because of climate change—linked to high greenhouse gases emission scenario—would have larger effects.
- Scenarios suggest specific financial institutions could face material risks.
- The analysis is limited to tropical cyclones and floods; total physical risk could be much larger when including droughts, heatwaves, and chronic risks.

### Uncertainty and Caveats
- The analysis is exploratory and subject to multiple layers of scenario and model uncertainties.
- Many uncertainties are difficult to quantify; results depend on scenario and modelling assumptions and simplifying data limitations.
- The exercise is not a standard stress test and is not focused on quantifying capital needs of the financial sector.

### Policy Recommendations and Priorities
- General recommendation: strengthen data collection and conduct further analysis to understand potential impact of climate risks.
- Enhance disclosure to collect granular data for understanding carbon footprint across the full spectrum of firms, not just listed firms.
- Partner with other government entities, the private sector, and research institutions to leverage expertise in climate and catastrophe modelling for granular, data-driven assessment of physical risks and incorporation into banking-sector resilience analysis.
- Table 1. Mexico: Recommendations on Climate Risk Analysis (as presented)
  - Recommendation: Improve and harmonize data collection and modelling to enable more comprehensive and granular analysis of transition and physical risks to the financial system.
    - Responsible Authorities: Banxico, CNBV
    - Time: MT
    - Priority: H
  - Recommendation: Strengthen, expand, and accelerate the ongoing development of climate related risk analysis and its integration with financial system analysis, in collaboration with other government entities, the private sector, and research institutions.
    - Responsible Authorities: Banxico, CNBV
    - Time: MT
    - Priority: M
  - Time definitions provided:
    - “I (immediate)” is within one year;
    - “NT (near-term)” is one–two years;
    - “MT (medium-term)” is three–five years.
  - Priority definitions provided:
    - H: high; M: medium; L: low.

### Contextual and International Notes
- COP26 outcomes and commitments cited:
  - More than 120 countries, representing about 70 percent of global emissions, pledged bringing emission to net zero around 2050.
  - According to IEA (2021), if the new pledges together with previous ones are followed through, it would help limit the rise in global temperatures to below 2 degrees centigrade by the end of the century.
  - Mexico joined the Global Methane Pledge and Declaration on Forests and Land Use during COP26.
  - Global Methane Pledge aims to reduce global methane emissions by at least 30 percent by 2030 relative to 2020 levels.
  - The global warming potential of methane is about 28 times higher than that of CO2.
- Banxico actions:
  - Financial System Stability Council in 2020 established the Committee on Sustainable Finance (CFS).
  - Banxico participates in various NGFS workstreams and other standard-setting bodies.

*Source: 1mexea2022008 - EXECUTIVE SUMMARY*

### 11.      A large fraction of total CO2 emissions in Mexico is concentrated in the power

### 11.      A large fraction of total CO2 emissions in Mexico is concentrated in the power generation industry followed by manufacturing and transportation

### Emissions and financial exposure: key findings
- About 38 percent of the total CO2 emissions in 2018 was attributable to the utilities/power generation sector.
- About 25 percent of the total CO2 emissions in 2018 was attributable to the manufacturing sector.
- About 16 percent of the total CO2 emissions in 2018 was attributable to the transportation sector.
- The mining sector, which includes oil, gas, and coal extraction, accounts for 11 percent of overall CO2 emissions.
- All other sectors combined constitute about 10 percent of the total CO2 emissions.
- About 37 percent of the commercial banking system’s total corporate credit is concentrated in emission intensive sectors.
- The largest banking exposure is to the manufacturing sector, accounting for about 15 percent of total credit.
- Exposures to the mining sector (including oil, gas, coal extraction) are relatively small at 4 percent of total credit.

### Implications and channels of risk to the financial sector
- Climate-related transition policies could have implications for financial stability; the impact depends on timing of policy implementation.
- Uncertainty about future transition implementation and lack of immediate global coordination compound risks.
- Risks can materialize via financial market channels even where strong transition policies are not planned, due to global investor awareness of climate issues.
- Potential policy delays could require significantly stronger future policy steps, which could prompt reassessment of risks for emission-intensive corporates in Mexico.
- Possible financial market effects include disruptions in equity and debt capital markets, sharp rise in corporate spreads, higher risk premia, increased market risk and volatility.
- The note focuses on transition risk transmitted to the banking sector through the corporate credit risk channel.

### Modeling framework (integrated micro-macro approach)
- Overall architecture: transition risk scenarios by sector feed firm-level corporate vulnerability indicators; these link via an estimated bridge equation to firm-level PDs, which are aggregated to bank-level impacts based on credit exposures.
- Flexibility: framework can accept inputs from other models and shocks from financial markets.
- The FSAP team used a computable general equilibrium (CGE) model (the IMF-ENV model) to generate sectoral transition pathways.
  - The IMF-ENV model is recursive-dynamic, multi-regional, multi-sectoral.
  - The model covers 25 regions (including Mexico).
  - The model has 37 distinct sectors.
  - Scenarios are anchored to NGFS scenario narratives (orderly and disorderly) and aligned with IPCC temperature/emission targets.
  - IMF-ENV was used to obtain sectoral output pathways tailored to Mexico.

### Corporate micro-simulation inputs and assumptions
- Corporate data sample:
  - Sourced from DataStream and S&P Capital IQ.
  - Balance sheet data covering 2002 to 2020.
  - On average there are about 100 total firms in the sample each year.
  - Sample firms belong to the most emission intensive sectors.
  - Data limitations: limited usable/reliable data on other Mexican firms, especially SMEs.
- Large firms and banking credit:
  - Only 2 percent of the universe of firms with outstanding loan facilities across 44 commercial banks are considered large firms (based on historical maximum loan amount criterion used by Banxico).
  - These large firms account for more than 65 percent of outstanding bank credit.
  - The sample of firms in the analysis represent about 42 percent of the total outstanding debt of the nonfinancial corporate sector.
- Key corporate variables available: earnings before interest and taxes (EBIT), sales revenue, cost of goods sold, interest expenses, average/effective interest rates, total debt, total assets, current assets, current liabilities.

### Mapping sectoral scenarios into firm-level financials
- Main structural link: corporate sales revenues are mapped to sectoral output and carbon price paths from the CGE model.
- Recursive evolution used to map to EBIT (notations simplified in source):
  - EBIT_t = EBIT_{t−1} + F_t * [Sales_{s,t−1} − G * COGS_{s,t−1}] − COSCARB_t
  - F_t is related to sensitivity of sales to sectoral gross output/GVA paths per-industry.
  - G is elasticity of cost of goods sold (COGS) to sales revenues.
  - COSCARB_t is direct additional operating cost due to firm-level emission projections and scenario-dependent carbon prices (carbon price times emissions in tCO2eq).
- Assumptions and parameter estimates:
  - Elasticity of sales to output/GVA per sector assumed to be 1.5 (informed by Gross and others (2022)).
  - Elasticity G (COGS to sales) estimated by panel fixed effects regression as 0.97.
  - Carbon price effects are modeled as additional operating cost (carbon tax as operating cost in EBIT; only scope 1 emissions used).
  - Urgentem provided projections of firm-level GHG emissions aligned with the three NGFS scenario categories (hot house world/business as usual, orderly, and disorderly).
- Simulations assume constant business models and technology over the projection horizon.
- The horizon of the analysis is five years (till 2026).

### Corporate vulnerability indicators and dynamics
- Main vulnerability indicators constructed in simulations:
  - Interest coverage ratio (ICR) = EBIT / interest expenses.
  - Leverage ratio (LR) = total debt / total assets.
  - Current ratio (CR) = current assets / current liabilities.
- Initial heterogeneity:
  - The construction sector has significantly low solvency and liquidity metrics and may be sensitive to small shocks.
  - Chemicals and non-metallic segment and utilities have relatively higher leverage ratios.
- Interest rate channel:
  - Initial median interest costs across sectors highlight risk propagation via cost of debt capital.
  - Market-driven sudden increases in credit risk premia can drive higher corporate spreads; motivates exploration of a jump-diffusion model of corporate spreads in the analysis.

### Bridge equation: linking vulnerability indicators to PDs
- Structural relationship (panel fixed effects with logit-transformed PDs):
  - slogit(PD)_{i,t} = alpha_i + beta_1 ICR_{i,t} + beta_2 CR_{i,t} + beta_3 LR_{i,t} + epsilon_{i,t}
- PDs sourced from Moody’s EDF database for Mexico (forward-looking default risk measure).
- Estimation results summary:
  - PDs are most sensitive to leverage ratio, followed by current ratio and interest coverage ratio.
- Interpretation caveats:
  - Right-hand side variables of the bridge equation exclude explicit macro-financial variables because historical estimation already embeds state variables of the economy.
  - Projected indicators are driven by CGE sectoral outputs, carbon prices, and other shocks (e.g., to interest rates).

### Analysis scope and next steps indicated in source
- The exercise mainly assesses transition risk impact on the banking sector through the corporate credit risk channel.
- The framework is capable of incorporating other shocks (financial markets) and additional channels (market risks, consumer portfolios) in future analyses.
- The note will further quantify transition-related risks and vulnerabilities in subsequent sections.

*Source: IMF staff, “11.      A large fraction of total CO2 emissions in Mexico is concentrated in the power generation industry followed by manufacturing and transportation,” 1mexea2022008.*

### 32.      The bridge equation helps generate scenario dependent PD paths for each firm which

### 1mexea2022008 - 32.      The bridge equation helps generate scenario dependent PD paths for each firm which

### Methodology: bridge equation and exposure-weighted PDs
- The bridge equation generates scenario dependent firm-level PD paths which are then aggregated to exposure-weighted sectoral PD paths.
- Bank-level exposures could not be mapped to individual firms; only aggregated credit exposures by broad economic sectors were available, necessitating an exposure-weighted projection of sectoral PDs for each scenario.
- For each scenario:
  - Delta PDs for each sector were computed as the difference between projected PDs in each year from 2022 to 2026 and the starting PDs in 2021.
  - For each bank, sectoral delta PDs over the projection horizon were weighted by each bank’s credit exposures to that sector to generate bank-level delta PD paths.
  - This computation assumed that sectoral credit exposures remain constant as at the starting point.

### Bank capital impact: simplifying assumptions
- Capital impact calculation considered only expected credit losses under a static balance sheet assumption.
- Sectoral breakdowns of non-performing loans and performing loans were not available.
- Risk Weighted Assets (RWA) were held constant at the starting value.
- In absence of sector and bank specific Loss Given Default (LGD) rates, a common LGD was used across all banks.
- Tax and dividend payout impact was ignored.
- Result: estimated impact on bank capital ratios is driven solely by loan impairment charges.
- System average corporate portfolio implied LGD from the coverage ratios of non-performing corporate exposures was found to be 53 percent.

### Transition scenarios and macro-sectoral pathways
- Two climate policy scenarios explored, motivated by NGFS scenarios:
  - global action (orderly transition, NGFS “below 2 degrees”)
  - delayed-uncertain (disorderly transition, NGFS “delayed transition”)
- A CGE model (IMF-ENV) was used to obtain sectoral output pathways tailored to Mexico to capture sectoral heterogeneity.
- An additional jump-diffusion stochastic financial modelling layer of corporate spreads is added to the delayed-uncertain scenario.
- Global action scenario carbon price floors by 2030:
  - $25 tCO2e for low-income countries (LICs)
  - $50 tCO2e for medium income countries (MICs)
  - $75 for high income countries (HICs)
- Under global action, relatively carbon intensive sectors face larger declines by 2030:
  - chemicals: decline in level of output of about 10 percent
  - fossil fuel sector: decline of 12 percent
  - non-metallic, transportation services, utilities see notable impact
  - transportation equipment is positively impacted by 2030

### Delayed-uncertain scenario: binomial tree and Monte Carlo interpretation
- The delayed-uncertain scenario models uncertainty about the timing of global action using a binomial tree (path branching) structure:
  - No global action in 2022; in each subsequent year (2023, 2024, 2025, 2026) the world may either act or continue delaying, creating multiple possible states.
  - In the stylized setting described, five possible paths through 2026 were constructed.
- The approach mimics a stochastic Monte Carlo simulation by running the CGE model multiple times under different delayed scenarios and stitching outputs into a binomial tree, allowing analysis of distributions and tail risks.
- Impact amplification from delay example:
  - Chemicals sector output impact more than doubles, from 10 percent in global action to 23 percent by 2030 if global action is delayed to 2025/26.
  - Similar amplification for fossil fuel and transportation services sectors.

### Jump-diffusion model of corporate spreads (financial market channel)
- To capture market-channel risks under delayed-uncertain paths, corporate spreads are modeled with a jump-diffusion stochastic process (Cox-Ingersoll-Ross with jumps / affine jump-diffusion):
  - Dynamics under martingale measure ℚ:
    - d r_t = κ(μ − r_t + s_t) d t + σ sqrt(r_t) d W_t + d J_t
  - Parameters and features:
    - κ > 0 is rate of mean-reversion
    - μ > 0 is long-run mean
    - σ > 0 is volatility parameter
    - W_t is a standard Brownian motion
    - J_t is a pure jump process with intensity and jump-size distribution
    - Feller condition 2 κ μ ≥ σ^2 ensures spreads stay non-negative
    - sqrt(r_t) term generates higher effective volatility when spreads are higher
  - A scaling factor s_t is constructed from sectoral outputs to map climate risks into corporate spread evolution (implementation details in Appendix I).
- Rationale:
  - Jumps capture sudden large increases in corporate spreads due to rising investor risk aversion from policy uncertainty and potential need for drastic future measures.
  - Diffusion term captures continuous smooth randomness; jump term captures discontinuous shocks (e.g., GFC, Covid, Russia-Ukraine conflict analogues).

### Simulation and mapping to firm balance sheets
- Corporate spreads model calibrated to Corporate Emerging Markets Bond Index (CEMBI) to represent aggregate Mexican corporate bond spreads.
- The evolution of corporate spreads was mapped into micro-simulation of firms’ balance sheet and P&L items as time-varying interest rates to capture debt capital market effects.
- The simulation produced 20,000 stochastic paths at daily frequency until end of 2026, enabling distributional analysis of spreads and exploration of tail outcomes.
- Calibrating spreads to individual firms was not feasible due to lack of liquid corporate bond or CDS data for most firms and absence of coherent corporate segment indexes for Mexico.

### Heterogeneity and key drivers of results
- Important heterogeneous drivers:
  - Initial risk characteristics and financial health of firms across sectors (e.g., interest coverage ratio, current ratio).
  - Heterogeneous sectoral impacts from the CGE model map into firms’ sales revenues and vulnerability indicators differently.
  - Sensitivity coefficients (betas) in the bridge equation differ across firms and sectors.
  - Banking system credit exposure concentration across sectors determines materiality of bank capital impacts.
- Under global action scenario sectoral PD effects (examples):
  - Maximum PDs in global action versus baseline average over horizon:
    - Chemicals and non-metallic sector: difference reaches almost 10 percent
    - Construction sector: deviation in PDs around 0.65 percent
    - Mining and transportation: notable deviations
  - High PD sensitivity in construction arises from initial weak solvency and liquidity metrics despite lower emission profile compared to mining and transportation.

*Source: IMF staff (extracted from the provided content unit).*

### 50.      The above asymmetric result highlights an important point regarding unintended

### 1mexea2022008 - 50.      The above asymmetric result highlights an important point regarding unintended

### Transition risk: asymmetric exposures and vulnerability
- Despite low emissions and hence low direct exposures to transition risks, some firms and sectors might still face heightened risks if there are weakness in their existing financial standing summarized for key vulnerability indicators.
- Limited scope of the analysis: does not capture additional constraints and indirect channels and spillover effects (e.g., reduced demand of their products due to production, supply-chain linkages) that could further amplify effects.
- Conclusion: the relatively benign PD impact in many sectors needs to be interpreted with caution.

### Sectoral and bank-level credit risk under Global Action
- Delta PDs across banks, weighted by exposures to vulnerable sectors, show a significant rise in corporate credit portfolio risks for some banks under the global action scenario.
- Observations:
  - Delta PDs for some banks reach almost 1 percent.
  - System wide delta PD rises to above 0.7 percent by 2026.
- Interpretation: credit-risk-based evaluation (not exposure-based) indicates non-trivial impact, with pockets of vulnerabilities in parts of the system.

### Bank capital impact under Global Action (static balance sheet assumption)
- Cumulative impact on bank capital due to expected credit losses from exposures to these sectors is about 0.35 percent.
- Diversification effects:
  - About 37 percent of banking sector portfolio belongs to these sectors.
  - Diversified exposures across banks contain the impact on bank capital.

### Delayed-uncertain pathways: distributions, time variation, and tail risks
- Risk propagation channels are largely similar to Global Action, but risk metrics (sectoral PDs and capital impacts) are better characterized by time-varying distributions.
- Delayed-uncertain pathways consist of increasingly stringent variants of early global action effects; uncertainty implies multiple possible states at each point in time.
- Key quantitative findings:
  - Sectoral PD distributions shift right with longer delays; tail mass increases sharply.
  - Example: relative to acting in year 2023 (maximum possible credit risk around 3 percent), under delayed 2026 the risk could increase to as high as 14 percent. The support of the distribution increases by more than 4 times under action in 2026 relative to acting in 2023.
  - Bank delta PDs: under early action in 2023 the maximum delta PD across banks was around 0.2 percent; under delays to 2026 delta PDs could rise to as high as 1.5 percent (more than 7-fold rise in maximum tail risk).
  - Bank capital ratios: absolute impacts appear modest but relative increases are large. Maximum range of the distribution increases by almost 5 times under delaying actions till 2026 versus 2023.
  - Cumulative bank capital impact under delayed-uncertain path could reach as high as 0.8 percent with non-trivial probability, relative to a maximum impact of around 0.3 percent under global action.
- Mechanisms increasing tail risk: jump component of financial model increases frequency of large shocks (tails); diffusion term (square root) increases effective volatility after jumps.

### Physical risk — overview and focus
- FSAP focuses on acute physical risks (hydrometeorological disasters: tropical cyclones, storms, floods, droughts, heatwaves), while noting chronic risks (sea level rise, gradual temperature increases) may also be substantial and relevant for future analysis.
- Acute risks analyzed primarily: floods and tropical cyclones.
- Note: Mexico is also subject to substantial non-climate natural disasters, including earthquakes.

### Hydrometeorological exposure and geographic concentration
- Floods and tropical cyclones have been the largest hydrometeorological hazard burden in the last two decades, in frequency and people impacted: more than 15 million people affected by these hazards over the last two decades.
- Geographic concentration of GDP (2020) — six states contributed collectively around 50 percent of national GDP:
  - Mexico City: 17.51 percent
  - Mexico State: 9.15 percent
  - Nuevo León: 7.74 percent
  - Jalisco: 6.92 percent
  - Veracruz: 4.53 percent
  - Guanajuato: 4 percent
- Credit exposure concentration (end of 2021):
  - Total credit: 3,434 billion MXN
  - Mexico City: 1,513 billion MXN (more than 44 percent of total credit)
  - Other state shares of total credit: Nuevo León 13 percent, Jalisco 6.4 percent, State of Mexico 4.9 percent, Sinaloa 3.2 percent
  - Remaining 27 federal states account for 28 percent of total credit exposure
- Caveat: credit recording at firm headquarters (often Mexico City) may understate geographic distribution of financed operations.

### Flood and tropical cyclone characteristics and recent impactful events
- Mexico experiences diverse flood phenomena: river floods, pluvial floods, coastal floods (storm surge).
- Main rainfall drivers: tropical cyclones and easterly waves in summer; cold fronts in winter; most rainfall concentrated in rainy season (generally June to November).
- Notable historical impacts:
  - October 2007 floods in Tabasco and surrounding states: economic losses of more than 35 billion MXN and affected more than 2 million people.
  - Veracruz 2010: Hurricane Karl and Tropical Storm Matthew caused 25 billion MXN losses and affected more than 500,000 people.
  - Hurricane Alex 2010: 25 billion MXN economic losses and affected 650,000 people.
- Eight of the ten topmost impactful tropical cyclone events since 2020 impacted states overlooking the Gulf of Mexico.

### Climate change projections for extreme precipitation and tropical cyclones
- Extreme precipitation projections under SSP5-8.5:
  - Historical (1985-2014) 1-in-100-year maximum 5-day cumulative precipitation expected to occur with a return period of less than 50 years by the end of the century under SSP5-8.5.
  - Historical 1-in-100-year maximum 1-day cumulative precipitation expected to occur with a return period of less than 40 years by the end of the century under SSP5-8.5.
- Tropical cyclone projections:
  - Frequency estimated to decrease in the North Atlantic and increase in the northeast Pacific under RCP4.5, but both basins expected to experience an increase in frequency of Saffir-Simpson Category 4 events.
  - For landfalling tropical cyclones in the Yucatan Peninsula, intense hurricanes estimated to be more frequent under RCP8.5.
  - Projected precipitation associated with tropical cyclones ranges from ~5 percent to 40 percent increases.
- Uncertainty: substantial uncertainty exists in flood and tropical cyclone projections; some studies find risk may decrease in some regions (e.g., Tabasco and other southeastern states).

### Physical risk scenarios analyzed
- Five extreme (but plausible) scenarios with national-scale return period greater than 50 years:
  - TC1: tropical cyclone with two landfalls (Yucatan Peninsula as Category 4 hurricane and northeast Mexico with heavy precipitation)
  - TC2: two tropical cyclones impacting Veracruz within approximately 10 days
  - TC3: tropical cyclone making landfall in Colima impacting Guadalajara with Category 1-2 winds
  - F1: river flood in Tabasco and surrounding states
  - ES1: extreme season combining TC1, TC2, TC3, and F1 occurring within a single season
- For climate-change analysis, two versions per scenario:
  - Historical conditions: TC1-hist, TC2-hist, TC3-hist, F1-hist, ES1-hist (based on historical/current climate distributions)
  - Climate change conditions (RCP8.5): TC1-cc, TC2-cc, TC3-cc, F1-cc, ES1-cc (hazard severity increased per climate projections; frequency/return period kept fixed; vulnerability and exposure assumed constant)
- Note: a sixth scenario (urban flood in Mexico City) was explored but not included due to very high uncertainty in direct damage estimates.

*Sources: Moody’s; Banxico; CENAPRED; and IMF staff calculations.*

### 70.      Scenario TC1 is defined as a North Atlantic tropical cyclone event with two landfalls in

### 1mexea2022008 - 70.      Scenario TC1 is defined as a North Atlantic tropical cyclone event with two landfalls in

### Scenario definitions
- TC1: North Atlantic tropical cyclone with two landfalls in Mexico.
  - First landfall: Yucatan Peninsula as a Category 4 hurricane (comparable to Janet 1955, Gilbert 1988, Wilma 2005, Dean 2007).
  - Second landfall: Tamaulipas (northeast Mexico) at Tropical Storm / Category 1 hurricane intensity with heavy rainfall impacting Nuevo León, including Monterrey (comparable to 1909 Monterrey Hurricane, Beulah 1967, Gilbert 1988, Alex 2010).
- TC2: Two tropical cyclones impacting Veracruz 10 days apart with heavy precipitation (based on 2010 sequence: Hurricane Karl and Tropical Storm Matthew).
- TC3: Eastern Pacific tropical cyclone making landfall at Manzanillo, Colima as a Category 4 hurricane and impacting Guadalajara, Jalisco as a Category 1-2 hurricane (comparable events include Hurricane Patricia 2015; estimated losses higher than Patricia).
- F1: River flood event in Tabasco and surrounding states (based on historic floods including 2007 Tabasco floods; 2007 event estimated to have a 1-in-100-year return period).
- ES1: Extreme season comprised of TC1, TC2, TC3 and F1 occurring within a few months in one season (historical analog: 2010 which included Hurricane Alex, Hurricane Karl, Tropical Storm Matthew, and Tabasco flooding).

### Historical event frequency (EMDAT 2000–2021)
- Flash flood mean: 0.55 (Number of events meeting EMDAT event criteria 2000-2021)
- Riverine flood mean: 0.91
- Tropical cyclone mean: 3.09
- Total mean: 4.55

### Methodology for direct damages
- Overall approach:
  - Combine historical loss data analysis with a model-based approach using exposure, hazard, and vulnerability modules.
- Exposure module:
  - 2013 total physical capital stock data from Instituto Nacional de Estadística y Geografía (INEGI, 2013) scaled to 2020 exposure, spatially disaggregated using Global Human Settlement Layer.
- Hazard module:
  - Perils modeled: wind and/or flood (storm surge not modelled).
  - Wind: CLIMADA’s implementation of Holland (2008) method for wind fields.
  - Flood: Fathom riverine (defended) hazard maps; selection of return periods based on historical footprints and precipitation/streamflow analyses.
- Vulnerability module:
  - Flood vulnerability curves: European Commission Joint Research Centre curves (Huizinga and others, 2017).
  - Tropical cyclone vulnerability: Eberenz and others (2021) regional tropical cyclone impact functions.
- Climate-change adjustments:
  - Tropical cyclone windspeed and precipitation increases based on CLIMADA methodology (Aznar-Siguan and Bresch, 2019) applied to Knutson and others (2015) estimates.
  - Precipitation changes translated to flood depth assuming precipitation-streamflow elasticity for extreme events approaches unity and applying published streamflow return period relationships.
  - Flood scenario changes from literature (Raffaele and Di Sante, 2020) converted to flood return period and depth.
- Historical loss scaling:
  - Historical CENAPRED loss data for comparable events scaled to 2020 based on exposure growth and inflation assumptions.

### Direct damage estimates and key statistics
- Extreme season (ES1) direct damages (percent of total national capital stock):
  - ES1-hist (historical conditions): 0.9 percent
  - ES1-cc end-of-century (climate change conditions): 1.9 percent
- Direct damage monetary range for individual scenarios under historical conditions:
  - From 30 billion MXN (approximately <0.1 percent of capital stock) to 600 billion MXN (approximately 0.7 percent of capital stock).
- National-scale return periods (direct damages):
  - TC1-hist: 50 years
  - TC2-hist: 50-100 years
  - TC3-hist: 500 years
  - F1-hist: 50-100 years
- Climate change impacts (comparison to historical conditions):
  - Damages may increase by 40-50 percent by mid-century under RCP8.5.
  - Damages may increase by more than 100 percent by end-of-century under RCP8.5.

- Table 3 — Direct damage rate (percent of capital stock) and climate change impact (relative to “historical conditions” scenario)
  - TC1
    - historical conditions: 0.14 percent
    - climate change conditions mid-century: 0.21 percent — 43 percent
    - climate change conditions end-century: 0.23 percent — 59 percent
  - TC2
    - historical conditions: 0.03 percent
    - climate change conditions mid-century: 0.05 percent — 54 percent
    - climate change conditions end-century: 0.08 percent — 124 percent
  - TC3
    - historical conditions: 0.67 percent
    - climate change conditions mid-century: 1.01 percent — 51 percent
    - climate change conditions end-century: 1.45 percent — 118 percent
  - F1
    - historical conditions: 0.05 percent
    - climate change conditions mid-century: 0.08 percent — 54 percent
    - climate change conditions end-century: 0.13 percent — 145 percent
  - ES1
    - historical conditions: 0.90 percent
    - climate change conditions mid-century: 1.35 percent — 50 percent
    - climate change conditions end-century: 1.89 percent — 110 percent
  - Note: “Climate change conditions” are based on RCP8.5.

### Assumptions, uncertainties, and limitations
- Lack of access to full catastrophe models used in Mexico (e.g., Universidad Nacional Autónoma de México, Evaluación de Riesgos Naturales, Verisk, RMS) limits precision.
- Exposure disaggregation assumes capital stock distribution mirrors population distribution; this may misrepresent inequalities and non-homogenous wealth distribution.
- Hazard module uncertainties:
  - Projected changes in windspeed and precipitation.
  - Simplifying assumptions in translating precipitation to flood depths that may lack local validation.
- Vulnerability module uncertainties:
  - Reliance on regionally calibrated curves not locally validated for Mexico.
- Model sensitivity:
  - Results sensitive to exposure, hazard, and vulnerability components.
- Recommendation noted: Further analysis with more granular exposure information and leveraging local catastrophe modeling and climate science would refine damage estimates.

### Macro-financial impact modeling approach
- Damage estimates from ES1 used as key input to generate scenario-dependent macro-financial paths.
  - Direct damage used in macro model: 0.9 percent (ES1-hist) and 1.9 percent (ES1-climate change conditions).
  - Analysis horizon: three years till 2024.
- Rationale:
  - Analysis focuses on extreme but plausible tail-risk scenarios to quantify channels and propagation mechanisms.
  - Uses the same underlying macro model framework as the FSAP stress testing exercise (2022-2024) to isolate climate-related impacts.
- Channels and amplifiers incorporated into macro model (GFM):
  - Direct destruction of physical capital:
    - Direct damages estimated regionally and aggregated to national capital level (Table 3).
    - Treated as an immediate depreciation shock to capital stock.
  - Impact on total factor productivity (TFP):
    - TFP shocks calibrated at twice the damage rate and assumed highly persistent.
    - Rationale: productivity reduction due to slow rebuilding and complementarity of infrastructure and non-infrastructure capital.
  - Impact on unemployment:
    - Concurrent shock to unemployment informed by elasticity of unemployment to changes in capital stock in Mexico (displacement of labor from damaged productive infrastructure).
  - Effects from financial markets:
    - Negative shocks to equity markets informed by elasticity of overall equity market returns to changes in capital stock in Mexico.
    - Equity chosen given importance in solvency stress-testing satellite model calibrations.

*Source: IMF staff and World Bank analysis as presented in the supplied content.*

### 83.      The impact of physical hazards from floods and tropical cyclones on the overall

### 1mexea2022008 - 83.      The impact of physical hazards from floods and tropical cyclones on the overall

### Impact on GDP and macroeconomy
- Scenarios considered: ES1-hist (historical conditions) and ES1-cc end-of-century (climate change conditions based on RCP8.5).
- Under the extreme season scenario, impact on GDP in growth terms (in deviation from the baseline):
  - ES1-hist: about 1.4 percent in 2022.
  - ES1-cc end-of-century: as much as 2 percent in 2022.
- Key interpretation:
  - Increased climate risks could have significant impacts going forward and the economic costs of climate driven risks are non-trivial.
  - Initial impacts can persist well into the future via additional indirect channels not fully captured by the analysis.

### Impact on the financial system (capital adequacy and banking sector)
- Methodology:
  - Approach follows standard FSAP stress testing methodology projecting the capital adequacy ratio (CAR) of the entire financial system relative to a baseline (based on June 15, 2022 World Economic Outlook projections).
  - Climate scenarios treated as additional risks on top of the baseline; impacts are deviations from baseline CAR projections.
- Projected CAR impacts under extreme season scenarios:
  - ES1-hist: aggregate CAR of the Mexican banking system would decline by about 1 percentage point versus the baseline.
  - ES1-cc end-of-century: capital ratio depletion is approximately 1.2 percentage points versus the baseline.
- Heterogeneity and tail effects:
  - Significant heterogeneity across banks: at some banks the CAR could decline by almost 4 percent under historical conditions and over 4 percent under climate change conditions, relative to the baseline.
  - System-wide aggregate impact is relatively modest, but pockets of significant vulnerabilities exist among some individual banks.
- Model and scenario limitations affecting interpretation:
  - Lack of access to granular exposure data and a full catastrophe model.
  - Analysis limited to a set of tropical cyclone and flood scenarios; does not account for simultaneous other climate-related disasters or non-linear compounding/amplification.
  - Geospatially explicit modelling captured direct damages to physical capital stock but did not model potential indirect effects including business interruption and large-scale supply chain disruptions.
  - Macroeconomic model is an aggregated national-scale model with linearized dynamics, limiting capture of regional variations and non-linear amplifications.
  - The analysis also does not account for potential adaptation policies that could mitigate impacts.

### Conclusions from transition and physical risk analyses
- Transition risk findings:
  - Heterogeneous impacts across corporate sectors.
  - Global action scenario produced relatively modest effects overall; delayed-uncertain pathways revealed potential for significant risks and heavier right tails in PD distributions for 2023 to 2026.
  - Chemicals and non-metallic segments of manufacturing appeared most vulnerable despite sound initial distress metrics; sectors less emission intensive (e.g., construction) may still be negatively affected depending on initial financial conditions.
  - No evidence of imminent systemic risk at this early stage, but pockets of vulnerabilities exist; delayed action increases tail risks.
  - Supports case for an early transition to a low carbon economy to mitigate tail risk from future large policy actions.
- Physical risk findings:
  - Acute hazards from tropical cyclones and floods can produce non-trivial impacts on the financial system under both historical/current and climate change conditions, with greater severity under climate change.
  - Materiality could be much larger if other risks (droughts, heatwaves, chronic risks) are considered.
  - Need to ensure mechanisms to improve financial resilience to such disasters.

### Policy recommendations and priorities
- Data and modelling:
  - Improve data collection and modelling for more comprehensive and granular analysis of transition and physical risks.
  - Enhance disclosure to collect granular data on carbon footprint across the spectrum of firms, not just larger and listed ones.
- Collaboration:
  - Partner with government entities, private sector, and research institutions to leverage climate and catastrophe modelling expertise for highly granular, data-driven physical risk assessment.
  - Incorporate a broad range of physical risks (acute and chronic) into banking sector resilience analysis.
- Regulatory and market development:
  - Strengthen, expand, and accelerate efforts to incorporate climate-related risk analysis into financial system analysis (noting enhanced analysis in the latest Financial Stability Report).
  - Further strengthen Mexico’s climate and disaster risk finance strategy and explore opportunities to expand insurance for climate-related risks.
  - Develop catastrophe risk insurance markets to improve financial resilience given current low levels of insurance penetration in Mexico.
- Disaster risk finance considerations:
  - Prioritize integrated financial and operational preparedness at local, state, and national levels and continue to leverage private risk capital (e.g., insurance and catastrophe bonds).
  - Note: FONDEN was dismantled in 2021; it had been a natural disaster fund established in 1996 that leveraged market-based risk transfer instruments, including catastrophe insurance and catastrophe bonds.

### Uncertainties, caveats, and modelling notes
- The analysis is a preliminary “first estimate” subject to uncertainties in:
  - Definition of climate and disaster scenarios.
  - Model used to estimate direct damages.
  - Transmission channels in the macro-financial model.
  - Financial sector impact model.
- Quantitative uncertainty is substantial, but qualitative effects (sign of risk materialization) are increasingly well understood.
- Suggested next steps:
  - Further analysis of potential impacts of climate change on financial stability should be a priority.
  - Explore sensitivity of results to key modelling assumptions.
- Appendix modeling assumption (delayed-uncertain analysis):
  - Key assumption: no action by end of 2022 implies corporate bond markets sharply react; analysis simplifies to one jump at year end 2022.
  - Jump size magnitude calibrated to a large shock seen during past crises: jump size set equal to the 99th percentile of rolling 2-year changes in the corporate spread.
  - Historical example cited: spreads increased from around 2 percent to almost 12 percent during the height of the 2007–2009 GFC.

*Source: MEXICO INTERNATIONAL MONETARY FUND.*

### 3. This is a conservative assumption however, it is the heightened and persistent

### 1mexea2022008 - 3. This is a conservative assumption however, it is the heightened and persistent

### Financial-market uncertainty and delayed-uncertain pathways
- Core narrative: Heightened and persistent uncertainty after large shocks (rather than initial shock magnitude) is of utmost importance; stochastic-process evidence and historical time-varying volatility (rolling 6 month) illustrate higher effective volatility.
- Treatment of jumps:
  - Random jumps before or after 2022 could be allowed and would significantly increase risks, but sustained jumps in fixed income markets are relatively rare compared to equity.
  - A one-time large jump term is used to capture sudden rises in macro policy uncertainty.
- Construction of time-varying spreads and mapping into firm-level micro-simulations:
  - Delayed-2023 path: average across all possible paths between time slices is computed to give dynamic path of interest rate shocks.
  - Delayed-2024 path: 75th percentile across paths is used.
  - Delayed-2025 path: 90th percentile is used.
  - Delayed-2026 path: 95th percentile is used.
  - From this sequence the initial value of the corporate spread is subtracted; the constructed time-varying spreads are mapped into uncertain pathways as shocks to interest rates in the micro-simulation of firm-level variables and vulnerability indicators.
- Rationale and flexibility:
  - Approach balances practical data constraints and the narrative of delayed-uncertain pathways by reconciling fully stochastic modelling into a simplified manual construction.
  - Financial model delivers a full set of distributions at each day into a five-year horizon, enabling experimentation with different parts of the distribution (average, 75th, 90th, 95th percentiles chosen to highlight increasing draws from the tail).

### Scaling factor linking sectoral outputs to corporate spreads
- Construction:
  - Scaling factor is a random variable: a uniform random variable between zero and the weighted average of sectoral outputs as deviations from baseline by 2030 is constructed.
  - The range of this uniform distribution is multiplied by the elasticity of corporate spreads to changes in aggregate output in Mexico, which was about 0.30 percent.
- Interpretation:
  - Random variables representing the scaling factor map overall macro randomness (induced by uncertain policy pathways) into corresponding effects on corporate spreads.
  - The jump term captures one-time large shocks due to sudden rises in macro policy uncertainty.
  - The scaling factor thereby links climate policy scenarios to evolution of corporate spreads in a simple, internally consistent manner.
- Modeling choices:
  - Uniform distribution chosen because the CGE model is not run in stochastic mode; other distributions could be fitted in principle.
  - Simplicity prioritized while preserving conceptual consistency with multiple delayed-uncertain pathways.

*Source: 1mexea2022008 - 3. This is a conservative assumption however, it is the heightened and persistent (IMF PDF content unit).*

### Appendix II. Physical Risk Direct Damages Estimation — methodology overview
- Combined approach: Final direct damage estimates per scenario based on a combination of (i) analysis of historical loss data for comparable events and (ii) a model-based approach using exposure, hazard, and vulnerability data.
- Extreme season scenario (Scenario ES1): direct damages for constituent events (Scenarios TC1, TC2, TC3 and F1) are summed because spatial extents of direct damages do not overlap.
- Historical loss data processing:
  - Historical losses for comparable events scaled to a 2020 loss estimate based on exposure growth and inflation assumptions.
  - Data extracted from CENAPRED reports, which provide economic loss data divided into direct damages (“daños directos”/“daños”) and indirect losses (“daños indirectos”/“pérdidas”) by state.
  - Where split unavailable, estimated using average ratio for other events of same peril type (flood or tropical cyclone).
  - Historical data compared with other databases, including EMDAT.

### Exposure module (model-based approach)
- Two-step process:
  - Scale total physical capital stock data by state to an estimate of 2020 exposure.
  - Spatially disaggregate the 2020 exposure to ~100m resolution and split into agricultural and built-up classes.
- Data sources and adjustments:
  - INEGI capital stock dataset for 2013 used as baseline; World Development Indicators (WB, 2022) used to adjust to 2020 via inflation, nominal GDP (2000–2020), and gross fixed capital formation.
  - Ratio between total capital stock and annual GDP extracted from INEGI dataset.
- Spatial disaggregation method:
  - Uses GHS_POP and GHS_SMOD to estimate capital stock density per pixel (maximum of two normalized densities).
  - Capital stock density map classified into agriculture and built-up using Copernicus Global Land Service Land Cover dataset.
  - Capital stock data distributed by its density.
- Limitations noted:
  - Spatial population distribution (e.g., Mexico City) may not reflect capital stock distribution due to inequalities and non-homogenous spatial distribution of wealth.

### Hazard module
- Perils modelled: wind and/or flood depending on scenario; storm surge is not modelled.
- Tropical cyclone wind modelling for “historical conditions”:
  - TC1-hist: wind modelled from historical IBTrACS track data; Hurricane Wilma 2005 used as reference for first landfall wind-driven component.
  - TC2-hist: wind component negligible; flooding from precipitation is main driver and wind not modelled in detail.
  - TC3-hist: no similar historical event; a similar track selected from STORM stochastic track dataset and wind fields simulated.
- Climate-change intensity adjustments (wind):
  - CLIMADA applies adjustments based on Knutson and others (2015): 4.5 percent increase in intensity for North Atlantic basin is the reference.
  - CLIMADA scales this adjustment for RCP8.5 and other time horizons by interpolating according to relative radiative forcing.
- Flood hazard:
  - Based on Fathom hazard maps (riverine defended maps used for Scenarios TC1, TC2 and F1).
  - Return periods for spatial units defined as river catchments within a state; selection of appropriate return period uses historical footprints, event descriptions, precipitation maps, precipitation and streamflow return period analyses, and calibration via comparative analysis.
- Climate-change adjustments for rainfall and streamflow:
  - For tropical cyclone-driven flooding, Knutson and others (2015) estimate a 17.3 percent increase in rainfall rate (available for RCP4.5 late 21st Century); CLIMADA scales to other RCPs/time horizons by relative radiative forcing.
  - Rainfall adjustments translated to change in streamflow assuming precipitation-streamflow elasticity for extreme events approaches unity.
  - Change in streamflow converted to change in return period based on published streamflow return period estimates for Mexican rivers.
  - For non-tropical cyclone-driven flooding, change in streamflow for a 1-in-100-year event estimated using Di Sante and others (2020) and converted to change in return period using Mora and others (2008).
- Caveats:
  - Hazard maps are global products and may not be locally calibrated for Mexico.
  - Use of different flood hazard maps can result in large differences in risk estimates.
  - Several studies indicate flood risk may decrease for some rivers in Mexico, so climate-change damage estimates may represent a “bad case” but not necessarily the “worst case.”

### Vulnerability module
- Damage estimation method:
  - Published vulnerability curves used: European Commission JRC vulnerability curves (Huizinga and others, 2017) for flood; Eberenz and others (2021) regional tropical cyclone impact functions for wind.
  - Flood curves relate flood depth to damage; tropical cyclone curves relate wind speed to damage.
  - Separate curves used where available for agriculture and built-up exposures.
- Uncertainty and limitations:
  - Substantial uncertainty in modelling vulnerability arising from aleatory and epistemic sources.
  - Vulnerability modelled deterministically as a function of wind speed or flood depth in this analysis.
  - Same vulnerability curves applied to all exposures within a class due to inability to differentiate exposure data (no differentiation by construction type, age, stories, occupancy).
  - Alternative methods (e.g., statistical sampling of damage ratios, modifiers for asset characteristics) could substantially change results.

### Estimation of return periods
- Return periods for “historical conditions” scenarios estimated based on magnitude of estimated direct damages.
- Without a full catastrophe model for Mexico, approximate return periods estimated by comparison:
  - Wind-driven scenarios (TC1, TC3): compared with probable maximum loss curve for wind events from Global Assessment Report 2015 (UNISDR, 2015).
  - Flood-driven scenarios (TC2, F1): compared with statistical analysis of historical flood event data from EMDAT using the World Bank Financial Risk Assessment Tool (WB, 2021b).
- Limitations:
  - Substantial uncertainty when fitting distributions to historical flood data and defining return periods for very infrequent extreme events.
  - Estimated national-scale return periods may differ substantially from state-scale return periods or return periods of specific physical hazard levels (e.g., a 1-in-100-year flood for a given river).

*Source: 1mexea2022008 - 3. This is a conservative assumption however, it is the heightened and persistent (IMF PDF content unit).*

### References

### 1mexea2022008 - References

### Major thematic coverage
- Climate hazards and projections
  - Global and regional flood risk projections: Alfieri et al., 2017; Hirabayashi et al., 2013; Hirabayashi et al., 2021; Raffaele and Di Sante, 2020; Breinl et al., 2021.
  - Tropical cyclone and hurricane modeling and projections: AIR, 2015; AIR, 2018; Bloemendaal et al., 2020; Eberenz, Lüthi, and Bresch, 2021; Holland, 2008; Knutson et al., 2020; Knutson et al., 2015; Appendini et al., 2019.
  - Climate change assessments and synthesis: IPCC, 2022; UNEP, 2021; UNISDR, 2015.

- Hazard, exposure, and impact data sources
  - Global hazard and exposure datasets and products: Copernicus Global Land Service (Buchhorn et al., 2020); GHS settlement grid and GHS population grid (Pesaresi et al., 2019; Schiavina et al., 2019); Copernicus/related datasets DOI and PID entries.
  - EM-DAT disaster database: Centre for Research on the Epidemiology of Disasters (CRED), 2022.
  - National and regional Mexican data and studies: CENAPRED, 2006; INEGI, 2013; Isela and Zarco, 2014; Mora et al., 2008; Ramos, Marrufo, Gonzalez, 2009; Haer et al., 2017; Haer et al., 2018; Neri-Flores et al., 2019; Magaña et al., 2003.

- Risk modeling frameworks, methods, and uncertainty
  - Catastrophe and synthetic hazard model generation: Bloemendaal et al., 2020; Trendafiloski et al., 2017; Kaczmarska, Jewson & Bellone, 2018.
  - Global flood and tropical cyclone risk model platforms and functions: Sampson et al., 2015; Aznar-Siguan and Bresch, 2019 (CLIMADA v1); Eberenz et al., 2021 (regional tropical cyclone impact functions); Huizinga et al., 2017 (Global flood depth-damage functions).
  - Uncertainty and damage modeling in catastrophe models: Trendafiloski et al., 2017; Kaczmarska et al., 2018; Breinl et al., 2021.

- Economic, financial, and policy analysis related to climate risk
  - Macroeconomic and financial stability implications: Chateau, Jaumotte, and Schwerhoff, 2022; Hallegatte, Jooste, McIsaac, 2022; Vitek, 2018; Financial Stability Board (FSB), 2020; IMF and World Bank, 2022.
  - Firm-level and banking implications: Aguilar-Gomez et al., 2022 (Thermal Stress and Financial Distress); Dieppe, Kilic Celik, and Okou, 2020 (Implications of major adverse events on productivity).
  - Climate-related stress testing and model frameworks: Grippa and Mann, 2020; Gross et al., 2022 (ENVISAGE-FIBA Model Framework).
  - Financial risk assessment tools and insurance market sources: World Bank (WB), 2021b (Financial Risk Assessment Tool); Axco, 2022 (Non-Life Insurance Market Reports – Mexico).

- Financial and quantitative methods referenced
  - Credit and asset pricing modeling: Cont and Tankov, 2004; Duffie, Pan, and Singleton, 2000; Jarrow, 2018; Lando, 2004.
  - Global macrofinancial modeling: Vitek, 2018.

### Key referenced datasets, models, and portals
- EM-DAT Database (CRED), 2022.
- Copernicus Global Land Service: Land Cover 100m: collection 3: epoch 2019: Globe (Buchhorn et al., 2020).
- GHS settlement grid and GHS population grid multitemporal R2019A (Pesaresi et al., 2019; Schiavina et al., 2019).
- CLIMADA v1 (Aznar-Siguan and Bresch, 2019).
- AIR Tropical Cyclone Model for Mexico (AIR, 2018).
- World Bank Climate Change Knowledge Portal (WB, 2021a).
- World Bank Disaster Risk Financing and Insurance Program – Financial Risk Assessment Tool (WB, 2021b).

### Methodological and sectoral focus areas reflected in references
- Flood hazard modeling, flood depth-damage functions, and hydrological analyses (Huizinga et al., 2017; Sampson et al., 2015; Breinl et al., 2021; Hirabayashi et al., 2013; Hirabayashi et al., 2021).
- Tropical cyclone hazard generation, intensity, and impact functions (Holland, 2008; Knutson et al., 2015; Knutson et al., 2020; Bloemendaal et al., 2020; Eberenz et al., 2021).
- Economic cost–benefit analysis and optimal adaptation policies for flood protection (Haer et al., 2017; Haer et al., 2018).
- Uncertainty quantification in catastrophe and natural catastrophe models (Kaczmarska et al., 2018; Trendafiloski et al., 2017).

*References list as provided in 1mexea2022008 - References*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2022/english/1mexea2022008.pdf_