## Bank Stress Testing of Physical Risks under Climate Change — Executive Summary (Content Unit)

## Source details

**Canonical URL:** [Bank Stress Testing of Physical Risks under Climate Change — Executive Summary (Content Unit)](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022163-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022163-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022163-print-pdf.pdf.json)

---

### Background and methodological contribution
- New macro-scenario stress testing method for banks to assess physical risks from climate change, applied to the Philippines.
- Four sub-modules coupled: climate scenario, disaster (CAT) scenario, macro-financial (DSGE) scenario, and bank solvency stress test.
- Key methodological advances:
  - Uses country-specific climate scenarios from a Philippines study (PAGASA 2018) downscaling GCMs with RCMs under RCP 8.5 (mid-21st century, 2036–65).
  - Establishes damage functions with micro-foundations using a catastrophe risk model (CAT) rather than arbitrary IAM functions.
  - Couples climate model outputs (GCM + RCM downscaling under RCP 8.5) with the CAT model to generate future hazard parameters for damage estimation.
  - Translates CAT-estimated capital damage into macro shocks via a DSGE model and into bank solvency outcomes using the IMF FSAP solvency stress testing method.

### Climate and disaster scenario findings
- Climate projections and frequency/intensity:
  - RCP 8.5 produces global temperature increases by 2100 of 4.0–6.1°C above pre-industrial levels; mid-21st century warming is stated as 1.4 to 2.6 °C.
  - Overall typhoon frequency likely to decline; relative number of intense typhoons could rise.
  - Out of five simulations: three suggest significant decreases in tropical cyclone frequency; two suggest little change. Four models agree on an increase in intensity, with two showing significant increases.
- CAT model setup and sample:
  - Simulation set: 50,000 simulated typhoons (10,000 per each of five RCM outputs).
  - Exercise uses the higher estimated losses (90th percentile) for conservatism.
- Damage rates and examples (private-sector assets unless noted):
  - For once-in-500-year events, the future damage rate is about 70 percent higher than the current damage rate.
  - For once-in-100-year events, the damage rate increases by 40 percent.
  - Average annual (chronic) damage rates:
    - Current annual average damage rate of capital: about a quarter percentage point.
    - Future annual average damage rate under RCP 8.5: about a third percentage point.
  - Acute tail-event examples:
    - About 3 percentage points lost for once-in-100-year events currently.
    - Up to about 8½ percent for once-in-500-year events in the mid-21st century (example ranges cited).

### Macroeconomic transmission and DSGE scenario calibrations
- Physical damage treated as a one-time capital depreciation shock applied to capital stock.
- DSGE amplifying channels:
  - Persistent decline in total factor productivity (TFP) calibrated at twice the magnitude of the increase in depreciation (TFP declines in parallel); empirical studies imply roughly one-third of GDP impact from capital destruction and two-thirds from TFP shock.
  - Time-to-reconstruct effects capture financial, regulatory, and technical constraints slowing recovery.
- Chronic (annual) increases in damage raise steady-state capital depreciation and cumulative growth effects.
- Model structure and assumptions (summary):
  - New Keynesian DSGE with macro-financial linkages; RBC specification produced broadly similar GDP responses.
  - Three disaster channels: physical capital damage, temporary depreciation-rate shock (short-lived), persistent TFP shock (long persistence).
  - Quasi-static bank balance sheet assumption in the stress test: bank assets and credit-to-GDP ratios remain constant over the three-year horizon.
  - Baseline economic paths held identical for current and future scenarios to isolate marginal effects of climate change.

### Key macroeconomic impact estimates
- Rare tail-event GDP peak impacts:
  - Under current climate conditions, rare disasters with return periods of 100 years or above could reduce GDP by 3⅔–8⅔ percentage points at the peak.
  - With climate change (RCP 8.5), these rare typhoons could reduce GDP by 5–14 percentage points — about 40–60 percent more than now.
- Cumulative and chronic impacts:
  - Chronic typhoons increase annual GDP growth reduction by -0.12 percentage points.
  - Cumulative effects of chronic typhoon damage on GDP over 43 years reach 5.2 percent — equivalent to the peak GDP impact of a once-in-100-year typhoon.
- Summary figures from Table 1 (Physical capital damage rate and GDP peak decline by return period):
  - Return period 10 years: Physical capital damage rate Current = 0.52; Future = 0.76; GDP peak decline Current = -0.86; Future = -1.25.
  - Return period 25 years: Physical capital damage rate Current = 0.92; Future = 1.10; GDP peak decline Current = -1.52; Future = -1.82.
  - Return period 50 years: Physical capital damage rate Current = 1.42; Future = 1.98; GDP peak decline Current = -2.34; Future = -3.27.
  - Return period 100 years: Physical capital damage rate Current = 2.23; Future = 3.12; GDP peak decline Current = -3.68; Future = -5.15.
  - Return period 250 years: Physical capital damage rate Current = 3.68; Future = 5.70; GDP peak decline Current = -6.07; Future = -9.41.
  - Return period 500 years: Physical capital damage rate Current = 5.03; Future = 8.47; GDP peak decline Current = -8.65; Future = -14.00.
  - Chronic: Physical capital damage rate Current = 0.23; Future = 0.31; GDP steady-state growth difference = -0.12.
- Memo: Actual peak GDP shocks in past crises:
  - GFC (global financial crisis): -2.90
  - AFC (Asian financial crisis): -5.69
  - COVID: -14.49
  - 1980s political turmoil episode: -16.50

### Macro-financial scenarios and compounding shocks
- Scenario design:
  - 10 macro-financial scenarios: baseline scenarios with and without compounding shocks (COVID-19 used as additional stressor), each with adverse cases for once-in-25- and once-in-500-year typhoons using current and future disaster risks.
  - Baseline with COVID-19 uses IMF WEO forecast as of October 2020; baseline without pandemic uses January 2020 vintage.
  - Severe typhoon hits assumed in Q3 2020 for scenario timing.
- Compounding-shock findings (marginal increases in typhoon impact when combined with pandemic):
  - Joint typhoon and pandemic increases marginal impact of typhoons by:
    - 2 percentage points for a 25-year return period typhoon (current climate).
    - 5⅓ percentage points for a 500-year return period typhoon (current climate).
    - 2.2 percentage points for a 25-year return period typhoon (future climate).
    - Nearly 8⅔ percentage points for a 500-year return period typhoon (future climate).
  - The 15-percentage point GDP growth reduction observed in 2020 (from about 6 percent in 2019 to -9½ percent in 2020) is similar to the GDP impact of a once-in-500-year typhoon under the future scenario.

### Bank stress testing design
- Coverage: 46 universal and commercial banks, ~92 percent of banking system assets.
- Test horizon: three years after a disaster; two starting points: one in 2020 and one in mid-21st century.
- Risk channels included: credit and market risks, shocks on net interest income and pre-impairment income.
- Balance sheet and capital assumptions:
  - Bank balance sheet assumed to grow at same rate as nominal GDP; asset/liability structure unchanged during horizon.
  - Banks strengthen capital only through retained earnings (post-tax and dividend payments).
  - Capital ratios calculated following Basel III standardized approach adopted by national regulator.
  - RWAs evolve with credit growth net of provisions and adjusted by new NPLs unprovisioned to reach 150 percent risk weight.
- Satellite models estimated using Philippines quarterly data for 2005–19.

### Bank stress test results and financial-system implications
- Baseline and tail-event bank capital impacts (starting conditions at end-2019):
  - End-2019 system metrics used as starting point:
    - Total capital ratio ≈ 15 percent.
    - Non-performing ratio = 2 percent.
    - Liquid-assets-to-total-assets ratio = 32 percent.
    - GDP growth rate = 6 percent.
    - Consumer price inflation = 2.5 percent.
    - Gross public debt ratio = 37 percent.
    - External debt = 22 percent.
    - Current account deficit = 0.9 percent.
  - Climate change alone in future reduces bank capital ratio visibly only in tail events (once in 500 years); decline small at one percentage point.
  - Difference of capital ratios between baseline and 500-year typhoon:
    - Current scenario maximum difference = 0.2 percentage points.
    - Future scenario maximum difference = 0.9 percentage points.
  - About 25 percent of bank assets are securities (primarily domestic sovereign bonds), creating short-run valuation gains when monetary policy eases.
- Compound-event vulnerability:
  - The compound risk of an extreme typhoon and a COVID-like pandemic significantly reduces bank capital.
  - In future scenarios, a joint shock increases the impact on bank capital by 2.2 and nearly 8⅔ percentage points for 25- and 500-year return period events, respectively, compared with the baseline.
  - Highlights non-linear amplification when climate shocks compound with other systemic shocks.
- Interpretation:
  - Effects on bank capital appear largely manageable given end-2019 starting-point health, but compound shocks could be systemic.
  - Chronic damages accumulate to be more damaging over decades than single extreme tail events, though annual impact is small and likely absorbable by bank profit buffers.

### Comparisons with NGFS and broader model insights
- Comparison outcomes:
  - Both approaches show future damage rates for once-in-a-100-year typhoons increase about 30–40 percent under RCP 8.5, with corresponding GDP impacts of about 5 percent where projections overlap.
  - Authors’ CAT + country-specific approach provides richer tail-event information (multiple return periods, model uncertainties, geographic detail) but requires substantial country-specific data and climate-model analyses.
  - NGFS scenarios offer estimates across various emissions scenarios but typically only for once-in-a-100-year disasters.
- Table 2 summary (authors versus NGFS, mid-21st century RCP 8.5 and related entries):
  - NGFS 2021: 17.6 median (projected increase in damage, summary entry).
  - NGFS RCP 8.5: 29.5 median; 36.5 upper bound.
  - Authors’ RCP 8.5: 20-68 90th pct. (range reported).
  - Return-period illustrative damage-rate and peak GDP-decline entries (authors’ column shown among NGFS comparators):
    - 100-year: damage-rate entries include 2.23 (authors current) to 3.12 (authors future); peak GDP decline entries include -3.7 (authors current) to -5.1 (authors future).
    - 500-year: damage-rate entries include 5.03 (authors current) to 8.47 (authors future); peak GDP decline entries include -8.7 (authors current) to -14.0 (authors future).

### Limitations, uncertainties, and cautions
- Scope limitations:
  - Focused on typhoon wind destruction alone; excludes related hazards such as floods, storm surge, and sea-level rise.
  - Excluded damages from physical risks other than infrastructure damages from typhoon wind.
  - Focused on macro-level transmission channels; did not account for micro-level concentrated effects such as lower property collateral values or credit-risk concentration due to lack of bank loan data by both industry and location conjoint.
- Uncertainty layers:
  - (i) Uncertainties linking socioeconomic activities and greenhouse gas emissions;
  - (ii) Uncertainties in climate science measuring emissions’ effect on global warming and translation to local phenomena (typhoons, sea-levels);
  - (iii) Uncertainties about effects of local climate phenomena on environmental, social, and economic systems and implications for financial stability.
  - Additional complicating and not yet well understood feedback relationships among these dynamics.
- Data and modeling caveats:
  - CAT damage functions rely on exposure and vulnerability data; the exercise did not incorporate potential future improvements in infrastructure resilience from mitigation and adaptation policies.
  - Using damage to buildings and infrastructure as a proxy for all productive capital is an approximation.
  - Short-run valuation gains from securities and monetary easing can offset initial capital declines; alternative scenarios with large country risk premia and capital outflows could reverse this and cause bond valuation losses.

### Policy implications and recommended analytical priorities
- For financial stability analysis and stress testing:
  - Coupling country-specific climate science and CAT models with macro-financial frameworks yields detailed tail-event assessments and is useful for EMDEs exposed to physical hazards.
  - Stress tests should consider compound shocks (for example, pandemic plus extreme weather) because nonlinear amplification can materially increase financial-system vulnerability.
  - Chronic increases in average damage rates warrant attention due to cumulative long-run impacts even when single-year effects are small.
- Data and analytical needs:
  - Micro-level data linking bank exposures by industry and location (conjoined) to assess collateral-value declines and credit concentration risks.
  - Incorporation of flood, sea-level rise, and other climate-related hazards for more comprehensive risk assessments.
  - Consideration of adaptation/mitigation policy effects and potential improvements in infrastructure resilience could change future damage projections.

*Source: wpiea2022163-print-pdf - Executive Summary*

### Executive Summary ......................................................................................................

### Bank Stress Testing of Physical Risks under Climate Change — Executive Summary (Content Unit)

### Major sections and structure
- Executive Summary (page 4)
- Introduction (page 6)
- Climate Scenario (page 10)
- Disaster Scenarios (page 13)
- Macro Scenarios (page 16)
- Bank Stress Test (page 22)
- Alternative Scenarios and Estimates (page 24)
- Conclusion (page 27)

### Annexes and supporting material
- Annex I. Macro Scenario Model (page 30)
- Annex II. Bank Stress Test Model (page 37)
- Annex III. Financial Soundness Indicator of the Philippines (page 40)
- Annex IV. Philippines Selected Economic Indicators, 2016-21 (page 41)
- References (page 42)

### Figures listed in the content unit
- Figure 1. Approach—Bank Stress Testing of Physical Risks (page 8)
- Figure 2. Climate Change’s Impact on Typhoon Characteristics in the Philippines under RCP 8.5 (page 12)
- Figure 3. Estimated Physical Capital Losses (page 15)
- Figure 4. GDP Assumptions for Climate Change Stress Test (page 21)
- Figure 5. Climate Change Stress Test—Impact on Total Bank Capital Ratio (page 23)
- Figure 6. Damage from typhoons in Philippines, change since 2005 (page 26)

### Tables listed in the content unit
- Table 1. Disasters’ Impact on Capital and GDP for the Philippines (page 20)
- Table 2. Difference between NGFS and Authors’ estimates: Damage rate and GDP (page 27)

### Annexed model and data components
- Macro Scenario Model (Annex I)
- Bank Stress Test Model (Annex II)
- Financial Soundness Indicator of the Philippines (Annex III)
- Philippines Selected Economic Indicators, 2016-21 (Annex IV)

### Glossary (acronyms and definitions as listed)
- AE Advanced economy
- AFC Asian financial crisis
- AR Assessment report
- BIS Bank for International Settlements
- BoE Bank of England
- CAT Catastrophe (risk model)
- CGM Computable general equilibrium model
- DNB De Nederlandsche Bank
- DSGE Dynamic stochastic general equilibrium
- ECB European Central Bank
- EMDE Emerging market and developing economy
- FSAP Financial Sector Assessment Program
- FSI Financial Stability Institute
- GCM General circulation model
- GFC Global financial crisis
- IAM Integrated assessment model
- IPCC Intergovernmental Panel on Climate Change
- LGD Loss given default
- NGFS Network of Greening the Financial System
- PAGASA Philippines Atmospheric, Geophysical, and Astronomical Services Administration
- RCM Regional climate model
- RCP Representative Concentration Pathway
- TC Tropical cyclone
- TCFD Task Force on Climate-Related Financial Disclosures
- TFP Total factor productivity
- UNEP United Nations Environmental Programme
- VaR Value at Risk
- WEO World Economic Outlook

*Source: wpiea2022163-print-pdf - Executive Summary*

### Executive Summary

### Executive Summary

### Background and methodological contribution
- The paper develops a new macro-scenario stress testing method for banks to assess physical risks from climate change, applied to the Philippines.
- Four sub-modules are coupled: climate scenario, disaster (CAT) scenario, macro-financial (DSGE) scenario, and bank solvency stress test.
- Key methodological advances:
  - Uses country-specific climate scenarios from an existing Philippines study.
  - Establishes damage functions with micro-foundations using a catastrophe risk model (CAT) rather than the typical arbitrary functions used in integrated assessment models (IAMs).
  - Couples climate model outputs (GCM + RCM downscaling under RCP 8.5) with the CAT model to generate future hazard parameters for damage estimation.
  - Translates CAT-estimated capital damage into macro shocks via a DSGE model and then into bank solvency outcomes using the IMF FSAP solvency stress testing method.

### Climate and disaster scenario findings
- Climate scenario basis:
  - Uses the Philippines Atmospheric, Geophysical, and Astronomical Services Administration (PAGASA) 2018 study that downscaled GCMs with RCMs under RCP 8.5 (mid-21st century, 2036–65).
  - RCP 8.5 produces global temperature increases by 2100 of 4.0–6.1°C above pre-industrial levels; mid-21st century warming is stated as 1.4 to 2.6 °C.
- Climate model projections:
  - Overall typhoon frequency is likely to decline; relative number of intense typhoons could rise.
  - Out of five simulations: three suggest significant decreases in tropical cyclone frequency; two suggest little change. Four models agree on an increase in intensity, with two showing significant increases.
- CAT model outputs and damage rates (private-sector assets unless otherwise noted):
  - Simulation set: 50,000 simulated typhoons (10,000 per each of five RCM outputs).
  - Future median losses are higher than current median losses for each return period.
  - For once-in-500-year events, the future damage rate is about 70 percent higher than the current damage rate.
  - For once-in-100-year events, the damage rate increases by 40 percent.
  - The exercise uses the higher estimated losses (90th percentile) for conservatism.
  - Average annual (chronic) damage rates:
    - Current annual average damage rate of capital: about a quarter percentage point.
    - Future annual average damage rate under RCP 8.5: about a third percentage point.
  - Acute tail event example: about 3 percentage points lost for once-in-100-year events currently; up to about 8½ percent for once-in-500-year events in the mid-21st century (example ranges cited).

### Macroeconomic transmission and DSGE scenario calibrations
- Translation of physical damage into macro shocks:
  - Typhoon destruction treated as a one-time capital depreciation shock (applied to capital stock as approximation).
  - DSGE model includes amplifying channels:
    - Long-lasting decline in total factor productivity (TFP) calibrated to be twice the magnitude of the increase in depreciation (i.e., TFP declines in parallel, with roughly one-third of GDP impact from capital destruction and two-thirds from TFP shock in empirical studies).
    - Time for reconstruction capturing financial, regulatory, and technical constraints that slow recovery (examples: budgetary transfer delays, procurement/regulatory delays for “build back better,” sectoral capacity constraints causing demand surge).
  - Chronic (annual) increases in damage raise steady-state capital depreciation and cumulative growth effects.

### Key macroeconomic impact estimates
- Rare tail event GDP impacts (peak impacts):
  - Under current climate conditions, rare disasters with return periods of 100 years or above could reduce GDP by 3⅔–8⅔ percentage points at the peak.
  - With climate change (RCP 8.5), these rare typhoons could reduce GDP by 5–14 percentage points—about 40–60 percent more than now.
  - The cumulative impact of chronic typhoons over 43 years reaches 5.2 percent—equivalent to the peak GDP impact of a once-in-100-year typhoon.
- Chronic annual impact:
  - Chronic typhoons increase annual GDP growth reduction by 0.12 percentage points (reported as small per year, but cumulative effects significant).

### Bank stress testing results and financial-system implications
- Baseline and tail-event bank capital impacts (Philippines context, starting conditions at end-2019):
  - Without compounding events, climate change would reduce bank capital ratios visibly only in tail events of once in 500 years.
  - The decline in bank capital ratio in such tail events is small (reported as one percentage point in one statement; elsewhere decline example figures: maximum of 0.2 percentage points in the current scenario and 0.9 percentage points in the future scenario for the difference between baseline and a 500-year event).
  - The difference of capital ratios between baseline and a 500-year return period typhoon remains minimal due in part to Philippines’ healthy macro-financial conditions at the start of stress (end-2019).
- Compound-event vulnerability:
  - The compound risk of an extreme typhoon and a COVID-like pandemic significantly reduces bank capital.
  - In future scenarios, a joint shock increases the impact on bank capital by 2.2 and nearly 8⅔ percentage points for 25- and 500-year return period events, respectively, compared with the baseline.
  - The exercise highlights non-linear amplification when climate shocks compound with other systemic shocks.

### Comparisons with NGFS and broader model insights
- Comparison with NGFS 2021 scenarios:
  - Both approaches show future damage rates for once-in-a-100-year typhoons increase about 30–40 percent under RCP 8.5, with corresponding GDP impacts of about 5 percent where projections overlap.
  - The CAT + country-specific approach provides richer information for tail events (multiple return periods, model uncertainties, geographic detail) but requires substantial country-specific data and climate model analyses.
  - NGFS scenarios offer estimates across various emissions scenarios but typically only for once-in-a-100-year disasters.

### Limitations, uncertainties, and cautions in interpretation
- Scope limitations:
  - Focused on typhoon wind destruction alone; excludes related climate change risks that could amplify impacts, such as floods and sea-level rise.
  - Excluded damages from physical risks other than infrastructure damages from typhoon wind.
  - Focused on macroeconomic-level transmission channels; did not account for micro-level concentrated effects such as lower property collateral values or credit risk concentration due to data limitations (bank loan data not available by both industry and location conjoint).
- Uncertainty layers highlighted:
  - (i) Uncertainties linking socioeconomic activities and greenhouse gas emissions;
  - (ii) Uncertainties in climate science measuring emissions’ effect on global warming and translation to local phenomena (typhoons, sea-levels);
  - (iii) Uncertainties about effects of local climate phenomena on environmental, social, and economic systems and implications for financial stability.
  - Additional complicating and not yet well understood feedback relationships among these dynamics.
- Data and modeling caveats:
  - CAT damage functions rely on exposure and vulnerability data; the exercise did not incorporate potential future improvements in infrastructure resilience from mitigation and adaptation policies.
  - Using damage to buildings and infrastructure as a proxy for all productive capital is an approximation.

### Policy implications and recommended analytical priorities
- For financial stability analysis and stress testing:
  - Coupling country-specific climate science and CAT models with macro-financial frameworks yields detailed tail-event assessments and is useful for EMDEs exposed to physical hazards.
  - Stress tests should consider compound shocks (for example, pandemic plus extreme weather) because nonlinear amplification can materially increase financial-system vulnerability.
  - Chronic increases in average damage rates warrant attention due to cumulative long-run impacts even when single-year effects are small.
- Data and analytical needs:
  - Micro-level data linking bank exposures by industry and location (conjoined) would enable assessments of collateral-value declines and credit concentration risks.
  - Incorporation of flood, sea-level rise, and other climate-related hazards would produce more comprehensive risk assessments.
  - Consideration of adaptation/mitigation policy effects and potential improvements in infrastructure resilience could change future damage projections.

*wpiea2022163-print-pdf - Executive Summary*

### 2.5 to 6.5 percent—e.g., by four percentage points— 푧푧

### wpiea2022163-print-pdf - 2.5 to 6.5 percent—e.g., by four percentage points— 푧푧

### Model structure and assumptions
- Uses a New Keynesian DSGE with macro-financial linkages estimated with Philippines data; an RBC specification produced broadly similar GDP responses.
- Three disaster channels incorporated: physical capital damage, temporary depreciation-rate shock (short-lived), and persistent TFP shock (long persistence) assuming capital and TFP return to pre-disaster levels.
- Includes habits in consumption and investment adjustment costs, implying time-to-reconstruct effects.
- Standard monetary policy follows a Taylor rule; tax revenue used for fiscal spending.
- Excludes climate- or disaster-specific policies, insurance market development effects, property-collateral-LGD channels, and remittance/international-aid amplifiers.
- Quasi-static bank balance sheet assumption in the stress test: bank assets and credit-to-GDP ratios remain constant over the three-year horizon.
- Baseline economic paths are held identical for current and future scenarios to isolate marginal effects of climate change.

### Disaster impacts on GDP (summary of key quantitative findings)
- Extremely rare typhoons (return periods ≥ 100 years) under current climate could reduce GDP by 3⅔–8⅔ percentage points at the peak.
- With climate change, rare typhoons (return periods ≥ 100 years) could reduce GDP by 5–14 percentage points — about 40–60 percent more than under current climate conditions.
- For one-in-a-500-year typhoons, GDP impact is comparable to that observed during the COVID crisis.
- For events with return period ≤ 25 years, economic impact is relatively small; climate change increases future GDP impact by 20 percent for once-in-25-year events compared with current scenario.
- Increase in GDP impact is larger for rarer events, reaching 60 percent for once-in-500-year events.
- Across return periods: a one percentage point increase in damage rate deepens peak GDP declines by 1⅔ percentage points (calculated as columns G over C in Table 1).
- Chronic (steady-state) impact: reduction of annual average GDP growth rate is -0.12 percentage points per year.
- Relationship for chronic vs extreme: a one percentage point increase in chronic damage rate implies a 0.4 percentage point increase in GDP impact (columns F over B in Table 1, chronic disaster row), versus 1⅔ percentage points for extreme events.
- Under RCP 8.5, cumulative effects of chronic typhoon damage on GDP over 43 years reach 5.2 percent — equivalent to peak GDP impact of a once-in-100-year typhoon.

### Table 1 — Selected figures (Damage rates and GDP peak declines by return period)
- Return period 10 years: Physical capital damage rate Current = 0.52; Future = 0.76; C = 0.24; D = 46; GDP peak decline Current = -0.86; Future = -1.25; G = -0.4; H = 46
- Return period 25 years: Physical capital damage rate Current = 0.92; Future = 1.10; C = 0.18; D = 20; GDP peak decline Current = -1.52; Future = -1.82; G = -0.3; H = 20
- Return period 50 years: Physical capital damage rate Current = 1.42; Future = 1.98; C = 0.56; D = 39; GDP peak decline Current = -2.34; Future = -3.27; G = -0.9; H = 39
- Return period 100 years: Physical capital damage rate Current = 2.23; Future = 3.12; C = 0.89; D = 40; GDP peak decline Current = -3.68; Future = -5.15; G = -1.5; H = 40
- Return period 250 years: Physical capital damage rate Current = 3.68; Future = 5.70; C = 2.2; D = 55; GDP peak decline Current = -6.07; Future = -9.41; G = -3.3; H = 55
- Return period 500 years: Physical capital damage rate Current = 5.03; Future = 8.47; C = 3.44; D = 68; GDP peak decline Current = -8.65; Future = -14.00; G = -5.4; H = 62
- Chronic: Physical capital damage rate Current = 0.23; Future = 0.31; GDP steady-state growth difference = -0.12

- Memo: Actual peak GDP shocks in past crises:
  - GFC (global financial crisis): -2.90
  - AFC (Asian financial crisis): -5.69
  - COVID: -14.49
  - 1980s political turmoil episode: -16.50

### Macro-financial scenarios and compounding shocks
- 10 macro-financial scenarios established: baseline scenarios with and without compounding shocks (COVID-19 used as additional stressor), each with four adverse cases: once-in-25- and once-in-500-year typhoons using current and future disaster risks.
- Baseline with COVID-19 uses IMF WEO forecast as of October 2020; baseline without pandemic uses January 2020 vintage.
- Assumption: severe typhoon hits in Q3 2020 in scenario timing.
- Compounding shock findings:
  - The 15-percentage point GDP growth reduction observed in 2020 (from about 6 percent in 2019 to -9½ percent in 2020) is similar to the GDP impact of a once-in-500-year typhoon under the future scenario.
  - Joint typhoon and pandemic increases marginal impact of typhoons by:
    - 2 percentage points for a 25-year return period typhoon (current climate).
    - 5⅓ percentage points for a 500-year return period typhoon (current climate).
    - 2.2 percentage points for a 25-year return period typhoon (future climate).
    - Nearly 8⅔ percentage points for a 500-year return period typhoon (future climate).

### Bank stress test design
- Coverage: 46 universal and commercial banks, ~92 percent of banking system assets.
- Test horizon: three years after a disaster; two starting points: one in 2020 and one in mid-21st century.
- Includes credit and market risks, shocks on net interest income and pre-impairment income.
- Bank balance sheet assumed to grow at same rate as nominal GDP; asset/liability structure unchanged during horizon.
- Banks strengthen capital only through retained earnings (post-tax and dividend payments).
- Capital ratios calculated following Basel III standardized approach adopted by national regulator.
- RWAs evolve with credit growth net of provisions and adjusted by new NPLs unprovisioned to reach 150 percent risk weight.
- Satellite models estimated using Philippines quarterly data for 2005–19.

### Bank stress test results (key quantitative outcomes)
- Climate change alone in future reduces bank capital ratio visibly only in tail events (once in 500 years); decline small at one percentage point.
- Difference of capital ratios between baseline and 500-year typhoon:
  - Current scenario maximum difference = 0.2 percentage points.
  - Future scenario maximum difference = 0.9 percentage points.
- About 25 percent of bank assets are securities (primarily domestic sovereign bonds), creating short-run valuation gains when monetary policy eases.
- End-2019 system metrics used as starting point:
  - Total capital ratio ≈ 15 percent.
  - Non-performing ratio = 2 percent.
  - Liquid-assets-to-total-assets ratio = 32 percent.
  - GDP growth rate = 6 percent.
  - Consumer price inflation = 2.5 percent.
  - Gross public debt ratio = 37 percent.
  - External debt = 22 percent.
  - Current account deficit = 0.9 percent.

### Interpretation, caveats, and robustness
- Climate change makes extremely rare typhoons potentially systemic for the Philippines; impacts worsen substantially with climate change.
- Chronic damages accumulate to be more damaging over decades than single extreme tail events, though annual impact is small and likely absorbable by bank profit buffers.
- The exercise excludes flood and sea-level rise damages, storm-surge, infrastructure spillovers, LGD increases from lower collateral values, loan concentration by location, and evolving insurance/reinsurance market development.
- Short-run valuation gains from securities and monetary easing can offset initial capital declines; alternative scenarios with large country risk premia and capital outflows could reverse this and cause bond valuation losses.
- Robustness check: Comparison with NGFS scenarios noted differences in climate scenarios, return periods, time horizons, and exposure data; NGFS provides damage pathways for once-in-100-year typhoons from 2020 to 2100 under multiple climate scenarios.

*Source: wpiea2022163-print-pdf - 2.5 to 6.5 percent—e.g., by four percentage points— 푧푧*

### 8.5 by linearly interpolating the results under RCP 4.5.

### wpiea2022163-print-pdf - 8.5 by linearly interpolating the results under RCP 4.5.

### Climate scenario and methodology
- Climate scenario: Philippines-specific study under the RCP 8.5 (Gallo and others 2018) for the mid-21st century; a specific point in time was picked rather than a whole pathway.
- Return periods examined: once in 10, 25, 50, 100, 250, and 500 years.
- Climate model uncertainty: used typhoon simulation results from all five climate models preferred by Gallo and others (2018).
- Exposure data: value and vulnerability of buildings and infrastructure by location estimated using gridded GDP data from ISIMIP.
- Local exposure dataset: built by the World Bank and the government of the Philippines, drawing on various sources of local information on buildings and infrastructure.
- Intercomparison with NGFS:
  - NGFS scenarios are provided by Climate Analytics; NGFS damage rates are cut off at a maximum of 40 percent.
  - NGFS created RCP 4.5 scenarios by scaling historical tropical cyclone data with Knutson and others (2015); other RCPs created by linear interpolation with respect to global temperatures following Aznar-Siguan and others (2021).

### Damage estimates and comparisons
- Mid-21st century change in physical capital damage from typhoons (authors’ climate scenario, 90th percentile):
  - For once-in-100-year events: rises by about 40 percent from 2020.
- Median estimates (authors’ Figure 3, panel 1): about one-third.
- NGFS RCP 8.5 comparable estimates: about 30 percent for median and above 40 percent for more tail estimates in 2060.
- NGFS pathways under current (as of 2021) policies:
  - Damage rate could double from 2060 (below 20 percent) to 2100 (about 40 percent).
- Tail sensitivity in authors’ estimates (Table 1 summary):
  - Damage rate increases by about 40 percent for once-in-100-year typhoons by mid-21st century.
  - Damage rate increases by about 55 percent for once-in-250-year typhoons by mid-21st century.
  - Damage rate increases by about 68 percent for once-in-500-year typhoons by mid-21st century.
- Relative differences versus NGFS (authors’ 90th percentile vs NGFS current policy):
  - Comparable for 25-year return period events.
  - Difference grows to about 40 percent for 500-year return period events.

### Key numeric results from Table 2 (Damage rate and GDP peak decline)
- Projected increase in damage (authors / NGFS 2021 / RCP 8.5, mid-21st century):
  - NGFS 2021: 17.6 median
  - NGFS RCP 8.5: 29.5 median; 36.5 upper bound
  - Authors’ RCP 8.5: 20-68 90th pct.
- Return-period specific damage rates (authors / NGFS 2021 / RCP 8.5 / NGFS RCP 8.5) and associated peak GDP decline (macroeconomic scenario columns):
  - 10-year: damage-rate entries: 0.52 0.6 0.7 0.7 0.8; peak GDP decline entries: -0.9 -1.0 -1.1 -1.2 -1.3
  - 25-year: damage-rate entries: 0.92 1.1 1.2 1.3 1.1; peak GDP decline entries: -1.5 -1.8 -2.0 -2.1 -1.8
  - 50-year: damage-rate entries: 1.42 1.7 1.8 1.9 2.0; peak GDP decline entries: -2.3 -2.8 -3.0 -3.2 -3.3
  - 100-year: damage-rate entries: 2.23 2.6 2.9 3.0 3.1; peak GDP decline entries: -3.7 -4.3 -4.8 -5.0 -5.1
  - 250-year: damage-rate entries: 3.68 4.3 4.8 5.0 5.7; peak GDP decline entries: -6.1 -7.1 -7.9 -8.0 -9.4
  - 500-year: damage-rate entries: 5.03 5.9 6.5 6.9 8.5; peak GDP decline entries: -8.7 -9.8 -10.7 -11.3 -14.0
- Notes on Table 2:
  - Both NGFS and authors’ estimates have ranges reflecting model and simulation uncertainties; the table uses median and upper bound estimates, while authors’ scenarios were built on the 90th percentile point estimates.
  - NGFS provides estimates only for once-in-100-year events; the same increase in damage rates is applied to other return periods for NGFS.
  - NGFS scenarios provide only the increase of damage, not the damage rates; authors used their current damage rates and applied NGFS percentage increases to calculate NGFS damage rates.

### Tail risks, time horizon, and model uncertainty
- Damage-rate differences across global climate scenarios tend to grow over time; model uncertainties (the band around the median) matter more at mid-21st century.
- Stress tests focused on mid-21st century projections should be interpreted carefully and used cautiously for long-time horizon discussions.
- Authors’ approach emphasizes probabilistic tail-risk assessment; impact of climate change on damage rates is generally higher for rarer (more tail) events.

### Macroeconomic and banking impacts
- Macro impact:
  - Extreme-tail (once in 100 or more years) impact up to mid-21st century on GDP could be systemic in the Philippines already and would worsen with climate change.
  - Less extreme disasters (such as 25-year return period events used by IMF-WB FSAP systemic stress tests) are not systemic in the authors’ results.
  - Chronic annual impact on GDP is small; cumulative chronic impact over the long run is substantially larger than rare extreme typhoons.
- Banking sector impact:
  - Effects on bank capital appear largely manageable given end-2019 starting-point data when banks were fairly healthy and economic vulnerabilities were limited.
  - Compound risk (e.g., an extreme disaster combined with a pandemic) could systemically distress the banking sector.
  - Banks are likely able to absorb annual chronic impacts with annual profits alone without using their capital buffer; long-run cumulative buffers (including cumulative profits) are higher than immediate buffers for short-term acute shocks.

### Limitations, uncertainties, and gaps
- Scope limitations:
  - Focused only on macroeconomic channels; could not analyze differentiated impact by location and industry, or micro-level channels such as collateral valuation.
  - Focused narrowly on wind-related destruction from typhoons; storm surge and flood impacts were not modeled due to data/model limitations.
  - A different CAT model is necessary to consider floods.
  - Chronic impacts such as effects on crop yields, sea-level rise, and higher-temperature effects on human productivity require different sets of models.
- Uncertainty types emphasised: risk, ambiguity (deep uncertainty), and model misspecification.
- More research needed before forming a comprehensive understanding of climate change impacts on financial stability and appropriate policy measures.

### Contributions and methodological advantages
- Tailoring: country-specific approach using detailed asset databases and insurance-industry standard vulnerability functions to produce tailored probabilistic assessments of tail-risks.
- Coupling: first known application coupling localized climate-model-based disaster projections with a CAT model for bank stress tests.
- Tail analysis: detailed analysis across multiple tail points (once in 10- to 500-year events) showing non-linear impacts for rarer events.
- Focus on emerging market: expands research to an emerging market economy where physical damage from climate change may be larger.

*Source: IMF Working Paper excerpt as provided in the content unit.*

### Annex I.  Macro Scenario Model

### Annex I. Macro Scenario Model

### Model overview and structure
- The model is a simplified version of Lipinsky and Miescu (2020) adapted for the economic and financial structure of the Philippines.  
- Agents: Households, Firms (capital-good producers and final-good firms), Banks, Monetary Authority, Government.  
- Key interactions: households supply labor and deposits; firms rent capital and borrow commercial loans from banks; banks intermediate between households and firms subject to a Basel II capital constraint; monetary authority follows a Taylor rule.

### Households: preferences, budget, and first-order conditions
- Preferences: Expected lifetime utility E0 ∑β^t Ut with Ut = z_{c,t} (c_t − h c_{t−1})^{1−σ}/(1−σ) − τ_n n_t^{1+φ}/(1+φ).  
  - z_{c,t} is a demand shock that increases consumption demand and induces more labor supply.  
  - Habit in consumption parameter: h > 0.  
- Budget constraint: c_t + d_t = w_t n_t + R_{d,t−1}/π_{t} d_{t−1} + Π_{i,t} + Π_{y,t}.  
  - Households receive wage income w_t n_t, nominal return on savings R_{d,t−1}/π_t d_{t−1}, profits from capital-goods producers Π_{i,t}, and profits from final-good firms Π_{y,t}.  
- First-order conditions (summary):  
  - Consumption Euler with habit and demand shock linking marginal utility today and expected marginal utility tomorrow.  
  - Labor supply: −τ_n n_t^{φ} + λ_t w_t = 0.  
  - Savings: −1 + β E_t[ λ_{t+1}/λ_t R_{d,t}/π_{t+1}] = 0.  
  - Investment optimality condition includes price q_t, adjustment cost S(·), and expectations of marginal costs and S′ terms.  
- Investment adjustment cost functional form: S( z_{i,t} i_t i_{t−1} ) = φ_i/2 ( z_{i,t} i_t i_{t−1} − 1 )^2.

### Firms: financing, returns, and cash flows
- Firms invest q_t k_t in productive assets k_t. Financing: commercial loans l_t and equity n_{F,t} = q_t k_t − l_t.  
- Next-period gross return on assets: R_{k,t+1} = r_{k,t+1} + (1−δ) q_{t+1}. Firms pay share Γ_{t+1} of earnings to banks for loans.  
- Firm cashflow (conceptual): −n_{F,t} + E_t[ M_{t+1} ( R_{k,t+1} k_t (1−Γ_{t+1}) ) ] (1−τ).  
  - Firm income is taxed at rate τ > 0.

### Banks: intermediation, default, and constraints
- Banks intermediate household deposits d_t (promise return R_{d,t}) into commercial loans l_t. Bank equity: n_{B,t} = l_t − d_t.  
- Next period, banks receive share Γ_{t+1} of firms’ earnings R_{k,t+1} k_t but incur default costs R_{k,t+1} k_t μ Δ_{t+1} for firm defaults.  
- Bank cashflow and objective (conceptual): −n_{B,t} + E_t[ M_{t+1} R_{k,t+1} k_t ( Γ_{t+1} − μ Δ_{t+1} ) − R_{d,t}/π_{t+1} d_t ] (1−τ).  
- Basel II capital constraint (solvency requirement): R_{k,t+1} k_t ( Γ_{t+1} − μ Δ_{t+1} ) − R_{d,t}/π_{t+1} d_t ≥ 0.  
- Value-at-risk shock: θ_t ≡ E_t[R_{k,t+1}] scaled by downside; θ_t quantifies downside risk for the bank.  
- Bank optimization: maximize expected cashflows subject to the Basel constraint and firms’ participation constraint; first-order conditions determine optimal l_t, n_{B,t}, and lending rate R_{l,t}.  
- Key tradeoffs in lending decision: benefit of financial intermediation τ ( R_{k,t+1} k_t /· ) (Γ_{t+1} − μ Δ_{t+1} ) versus cost of default −R_{k,t+1} k_t μ Δ_{t+1} (1−τ).  
- Γ(ε^*_t+1) and Δ(ε^*_t+1) are functions of the default threshold ε^*_t+1; Γ′(ε^*) = 1 − F(ε^*), Δ′(ε^*) = f(ε^*)/ε^*.  
- Lending-rate choice equates marginal cost of more defaults to marginal benefit of intermediation; first-order condition includes terms with (1−F(ε^*)) and μ f(ε^*) ε^*.

### Stock-price gap, firm net worth, and equilibria
- Firms accumulate net worth n_{F,t} from retained earnings, distributing share γ_D of earnings and receiving fixed equity injection ω:  
  n_{F,t} = R_{k,t} k_{t−1} (1−Γ(ε^*_t)) (1−τ) (1−γ_D) + ω.  
- Target net worth determined by first-order condition ∂/∂n_{F,t}: λ_{F,t} = 1.  
- Stock price gap: sss = (q_t − q_t^*) / q_t^*.  
- Equilibrium structure (optimal solution): firms’ participation constraint determines n_{F,t}; loan optimality determines k_t and l_t = q_t k_t − n_{F,t}; Basel II constraint determines d_t and n_{B,t} = l_t − d_t; household and bank first-order conditions imply λ_t V > 0.  
- Realized (suboptimal) solution: n_{F,t} exogenous law of motion; participation constraint determines k_t and l_t; loan optimality determines λ_{F,t} (time-varying).

### Financial sector aggregator and market clearing
- Representative financial-sector agent smooths dividends c_{FIN,t} by maximizing E0 ∑ β^t U^FIN_t with U^FIN_t = z_{c,t} ( c_{FIN,t} − h c_{FIN,t−1} )^{1−σ}/(1−σ).  
- Financial sector budget constraint: c_{FIN,t} + q_t k_t − d_t = [ R_{k,t} k_{t−1} (1−μ Δ_t) − R_{d,t−1}/π_t d_{t−1} ] (1−τ).  
- Stochastic discount factor: M_{t+1} = β λ_{FIN,t+1} / λ_{FIN,t}, where λ_{FIN,t} is the Lagrange multiplier on the financial-sector budget constraint.

### Final-good firms, inflation, and monetary policy
- Final-good firms rent capital k_{i,t−1}, hire workers n_{i,t}, and set prices s_{i,t} subject to quadratic price adjustment cost φ/2 ( P_{i,t}/P_{i,t−1} − 1 )^2 y_t.  
- Price and output relations: π_t ≡ P_t / P_{t−1}; s_{i,t} ≡ P_{i,t} / P_t; η_t = η_ss z_{η,t}.  
- Profit for firm i: Π_{y,i,t} = s_{i,t} y_{i,t} − r_t k_{i,t−1} − w_t n_{i,t} − φ/2 ( π_t s_{i,t}/s_{i,t−1} − 1 )^2 y_t.  
- Production function: y_{i,t} = k_{i,t−1}^{α} ( z_{y,t} n_{i,t} )^{1−α}. Aggregate y_{i,t} = s_{i,t}^{−η_t} y_t.  
- Symmetric equilibrium first-order conditions (representative firm):  
  - v_t = 1 − s_{m,t}.  
  - r_t = s_{m,t} α y_t / k_{t−1}.  
  - w_t = s_{m,t} (1−α) y_t / n_t.  
  - s_{m,t} = 1 − (1/η_t) [ 1 − φ π_t (π_t − 1) + φ β E_t[ λ_{t+1}/λ_t π_{t+1} (π_{t+1} − 1) ] y_t / y_{t−1} ].
- Monetary authority: Taylor rule for policy rate R_{d,t} relative to steady state R_{d,ss}:  
  R_{d,t}/R_{d,ss} = ( R_{d,t−1}/R_{d,ss} )^{γ_R} ( π_t / π_{ss} )^{γ_π (1−γ_R)} ( y_t / y_{ss} )^{γ_y (1−γ_R)} e^{σ_m ε_{m,t}}.

### Real-RBC limit
- In the absence of inflation and nominal frictions, the model reduces to a real-business-cycle structure with:  
  - y_t = k_{t−1}^{α} ( z_{y,t} n_t )^{1−α}.  
  - First-order conditions: r_t = α y_t / k_{t−1}; w_t = (1−α) y_t / n_t.

### Calibrated parameters (as reported)
- α — power on capital in production function: 0.300  
- β — discount factor: 0.990  
- δ — depreciation rate of capital: 0.025  
- h — habit in consumption: 0.750  
- φ — Frisch elasticity of labor supply: 1.000  
- τ_n — disutility of supplying labor: 4.000  
- φ_i — investment adjustment cost: 3.000  
- φ (price adjustment cost): 0.750  
- η_ss — steady-state price elasticity: 1.500  
- μ — cost of default / screening cost: 0.400  
- τ — additional discount associable to taxes: 0.030  
- γ_D — dividend payout: 0.030  
- Remaining parameters and shock magnitudes were estimated as described in Lipinsky and Miescu (2020).

*Source: Annex I. Macro Scenario Model, IMF Working Paper (excerpt).*

### Annex III. Financial Soundness Indicator of the

### Annex III. Financial Soundness Indicator of the Philippines

### Capital adequacy
- Regulatory capital to risk-weighted assets
  - 2015: 15.3
  - 2016: 14.5
  - 2017: 14.4
  - 2018: 14.9
  - 2019: 15.2
  - 2020*: 15.0
- Regulatory tier 1 capital to risk-weighted assets
  - 2015: 12.8
  - 2016: 12.6
  - 2017: 12.7
  - 2018: 13.3
  - 2019: 14.0
  - 2020*: 13.9
- Capital to total assets
  - 2015: 10.5
  - 2016: 10.4
  - 2017: 10.6
  - 2018: 11.3
  - 2019: 11.5
  - 2020*: 11.0
- Non-performing loans net of provisions to capital
  - 2015: 3.1
  - 2016: 3.0
  - 2017: 3.1
  - 2018: 3.5
  - 2019: 4.6
  - 2020*: 5.1
- Net open position in foreign exchange to capital
  - 2015: 2.4
  - 2016: 2.0
  - 2017: 7.9
  - 2018: 4.7
  - 2019: 5.8
  - 2020*: 3.5
- Gross asset position in financial derivatives to capital
  - 2015: 1.7
  - 2016: 1.8
  - 2017: 1.6
  - 2018: 1.8
  - 2019: 1.2
  - 2020*: 1.6
- Gross liability position in financial derivatives to capital
  - 2015: 0.0
  - 2016: 0.0
  - 2017: 0.0
  - 2018: 0.1
  - 2019: 0.4
  - 2020*: 0.6

### Asset quality
- Nonperforming loan to gross loans
  - 2015: 1.9
  - 2016: 1.7
  - 2017: 1.6
  - 2018: 1.7
  - 2019: 2.0
  - 2020*: 2.2
- Specific provisions to nonperforming loans
  - 2015: 70.1
  - 2016: 69.7
  - 2017: 66.9
  - 2018: 63.2
  - 2019: 58.0
  - 2020*: 57.6

### Earnings and profitability
- Return on assets
  - 2015: 1.4
  - 2016: 1.4
  - 2017: 1.3
  - 2018: 1.3
  - 2019: 1.5
  - 2020*: 1.4
- Return on equity
  - 2015: 13.8
  - 2016: 13.7
  - 2017: 13.6
  - 2018: 12.7
  - 2019: 13.9
  - 2020*: 13.0
- Interest margin to gross income
  - 2015: 70.7
  - 2016: 69.2
  - 2017: 73.9
  - 2018: 75.2
  - 2019: 74.0
  - 2020*: 76.3
- Trading income to total income
  - 2015: 5.7
  - 2016: 8.3
  - 2017: 4.3
  - 2018: 3.2
  - 2019: 7.8
  - 2020*: 9.4
- Noninterest expenses to gross income
  - 2015: 61.3
  - 2016: 60.8
  - 2017: 60.9
  - 2018: 62.2
  - 2019: 58.7
  - 2020*: 53.9
- Personnel expenses to non-interest expenses
  - 2015: 37.6
  - 2016: 36.7
  - 2017: 36.6
  - 2018: 35.4
  - 2019: 34.5
  - 2020*: 33.6

### Liquidity and funding
- Liquid assets to total assets
  - 2015: 38.8
  - 2016: 35.6
  - 2017: 32.9
  - 2018: 32.6
  - 2019: 32.1
  - 2020*: 30.6
- Liquidity assets to short-term liabilities
  - 2015: 60.6
  - 2016: 54.6
  - 2017: 51.8
  - 2018: 50.7
  - 2019: 48.8
  - 2020*: 46.9
- Non-interbank loans to customer deposits
  - 2015: 76.9
  - 2016: 76.3
  - 2017: 79.6
  - 2018: 82.7
  - 2019: 85.2
  - 2020*: 83.6

### Sensitivity
- Foreign currency denominated loans to total loans
  - 2015: 11.9
  - 2016: 11.9
  - 2017: 11.1
  - 2018: 10.9
  - 2019: 10.7
  - 2020*: 11.1
- Foreign currency denominated liabilities to total liabilities
  - 2015: 20.3
  - 2016: 20.7
  - 2017: 20.2
  - 2018: 20.1
  - 2019: 19.6
  - 2020*: 19.2

### Real estate markets
- Residential real estate loans to total loans
  - 2015: 7.2
  - 2016: 7.3
  - 2017: 7.2
  - 2018: 7.1
  - 2019: 7.3
  - 2020*: 7.4
- Commercial real estate loans to total loans
  - 2015: 13.9
  - 2016: 14.3
  - 2017: 14.1
  - 2018: 12.3
  - 2019: 13.2
  - 2020*: 13.7

### Household Indebtedness
- Loans to households to total loans
  - 2015: 17.4
  - 2016: 17.8
  - 2017: 17.9
  - 2018: 17.6
  - 2019: 18.3
  - 2020*: 19.3
- Consumer loans to total loans
  - 2015: 9.5
  - 2016: 9.9
  - 2017: 10.0
  - 2018: 9.8
  - 2019: 10.4
  - 2020*: 10.9
- Mortgage loans to total loans
  - 2015: 6.8
  - 2016: 6.8
  - 2017: 6.8
  - 2018: 6.7
  - 2019: 6.9
  - 2020*: 7.6
- Loans to households as employers to total loans
  - 2015: 1.1
  - 2016: 1.1
  - 2017: 1.1
  - 2018: 1.1
  - 2019: 0.9
  - 2020*: 0.8

*Source: Philippines authorities; IMF, Financial Soundness Indicators; and IMF staff estimates.  
*As of September 2020.*

*Annex III. Financial Soundness Indicator of the Philippines — IMF (2021).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022163-print-pdf.pdf_