## EXECUTIVE SUMMARY (content unit 1jpnea2024003)

## Source details

**Canonical URL:** [EXECUTIVE SUMMARY (content unit 1jpnea2024003)](https://www.imf.org/-/media/files/publications/cr/2024/english/1jpnea2024003.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2024/english/1jpnea2024003.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2024/english/1jpnea2024003.pdf.json)

---

### Financial system resilience and evolving risks
- Japan’s financial system is large, globally integrated, and entered the COVID-19 pandemic with strong capital and liquidity buffers; credit provision to the private sector remained robust since the pandemic.
- Key evolving risks:
  - Market risk and FX liquidity risk following BOJ’s end of negative interest rates and yield curve control (YCC).
  - Intensification of regional conflicts and geoeconomic fragmentation, global supply chain disruptions, and commodity price volatility that could trigger abrupt global slowdown and higher interest rates.
  - Structural transformations from climate change, digitalization, and an aging population.

### FSAP approach and scenario-based systemic risk analysis
- Comprehensive scenario-based analysis including:
  - Solvency and liquidity stress tests for a 23-bank sample and insurance-company stress tests (top-down and bottom-up).
  - Investment funds liquidity stress tests for 183 funds.
  - Corporate and household sector risk modeling feeding bank solvency tests.
  - Climate risk analysis using IMF-ENV and an ENV-FIBA micro-macro layer assessing transition and physical risks to 271,000 firms and 22 banks.

### Main stress-test findings — banking sector
- Solvency and valuation impacts:
  - Aggregate CET1 ratio falls by 510 basis points in the first year of the adverse scenario but remains above regulatory minimums.
  - Valuation losses account for a decline in the banking system capital ratio by 6.3 percentage points in the first year of the adverse scenario.
  - Loan losses grow three-fold under the adverse scenario relative to the baseline.
- Credit and NPLs:
  - Aggregate NPL ratio: initial 0.8 percent → third year 2.5 percent.
  - Domestic banks and regional banks NPLs: initial 1.5 percent → third year 3.6-3.8 percent.
  - System aggregate loan loss rate: initial year 0.1 percent → first year of adverse scenario 0.6 percent.
  - Domestic banks and regional banks first-year loan loss rates: 1.0-1.1 percent.
- Loss composition:
  - Bond and equity valuation losses + loan losses contribute 90 percent of combined losses; loan book contributes 10 percent.
  - Bond and equity investment loss contributions: domestic ≈50 percent; foreign ≈50 percent.
  - Loan-loss contributions: domestic exposures 79 percent; foreign exposures 21 percent.
- Capital shortfalls and buffers:
  - Three banks breach CET1 hurdle of 4.5 percent; four banks breach total capital hurdle of 8 percent (these banks amount to 0.04 percent and 0.05 percent of 2022 nominal GDP, respectively).
  - The four banks represent 5 percent of total assets of the 23 banks.
  - CET1: eight banks would consume their CCoB; total capital ratios: twelve banks would consume their CCoB.
- Sensitivities:
  - Interest-rate sensitivity variant: aggregate capital ratio drops an additional 1.1 percentage points to reach 8.9 percent; system-wide total capital shortfall rises from 0.05 percent of GDP to 0.3 percent of GDP.
  - Removing hedging: aggregate capital ratio falls an additional 130 basis points in year one; overall shift = 640 basis points; ten banks fall below hurdle rates.
  - Activating the AFS filter for domestic banks: starting CET1 moves from 15.1 percent to 14.7 percent; capital ratio drop under adverse scenario with AFS filter on = 340 basis points (vs 510 basis points initial).

### Main stress-test findings — bank liquidity and FX funding
- Liquidity stress-test design:
  - Cash flow-based, 6-month horizon, weekly steps (24 weeks); securities sales hierarchy with mark-to-market securities sold first.
  - Run-off calibration (first 30 days): unsecured wholesale funding ≈50 percent of liability-driven outflows; liability-side derivatives ≈30 percent; retail unsecured funding ≈10 percent; secured wholesale funding + drawdowns ≈10 percent.
- Liquidity outcomes:
  - Five banks face liquidity shortfalls, accumulating to 1.7 percent of initial liquid asset buffers (they represent 7 percent of total assets of the 23 banks).
  - Under a doubled run-off-rate sensitivity, banks with shortfall rise to twelve and shortfalls accumulate to about 6 percent of initial liquid asset stocks.
  - Selling mark-to-market securities to meet outflows causes capital declines >2 percentage points for two banks and ≈1.2 percentage points for one bank.
- FX funding exposures and sensitivities:
  - USD liabilities = 30 percent of total bank liabilities.
  - Banks’ undrawn FX commitments (credit + liquidity lines) = 67 percent of total commitments at end-September 2023.
  - Counterfactual USD funding shock: combined +220 bps funding cost shock (U.S. short-term +150 bps + 70 bps basis) would reduce some banks’ capital ratios by >0.5 ppt (and in some cases >1 ppt) relative to initial adverse scenario.
  - LCRs: all-currency LCRs generally above 100 percent; USD LCRs below 100 percent for internationally active banks and fall 30-40 percentage points under committed credit-line drawdown scenario.

### Main stress-test findings — insurance sector
- Sample and starting metrics (March 31, 2023):
  - Insurance firms’ aggregate assets in sample = JPY 360 trillion (majority life insurers).
  - Solvency Margin Ratio (SMR): major life insurers = 956 percent; non-life insurers = 840 percent.
  - Economic value-based Solvency Ratio (ESR): life insurers = 226 percent; non-life insurers = 212 percent.
- Adverse scenario impacts:
  - SMR-based top-down: life insurers’ available capital drops by about 48 percent; average life SMR 956 percent → 583 percent; ten of twelve life insurers remain above 200 percent; two do not meet requirement.
  - Non-life: available capital declines by 24 percent; average SMR 840 percent → 686 percent; all non-life insurers remain above 200 percent.
  - ESR-based results: ESR decreases by 67 percentage points (life) and 41 percentage points (nonlife); post-stress ESR for all insurers remains above 100 percent.
- Key drivers and sensitivities:
  - Market risks dominate capital requirements: 70 percent overall; equity risk largest contributor followed by currency risk.
  - Interest-rate sensitivity: a larger increase (+1.6 pp short-term, +2.5 pp long-term) lowers life insurers’ SMR-based available capital by 24 percent on average.
  - FX appreciation (JPY +10 percent) reduces SMR by 3 percent on average; up to 14 percent for some insurers.
- Liquidity stress-test:
  - Large net outflows for life insurers in adverse scenario surge to more than JPY 7 trillion aggregate; allowing reactive management actions generally mitigates net outflows but not always on an individual firm basis.
  - Life insurers’ liquid assets ratio ≈60 percent (baseline) and ≈70 percent (adverse).

### Main stress-test findings — investment funds
- Market structure and size:
  - Total net assets of investment trusts > JPY 300 trillion in 2023 (up from about JPY 170 trillion in 2015).
  - Publicly offered trusts = 60 percent of total net assets; privately offered = about 35 percent.
  - About 54 percent of total assets held in foreign currencies.
- Stress test coverage and sample:
  - Sample: 183 funds with AUM > JPY 50 billion; sample represents 72 percent of total AUM for investment trust products.
- Redemption shock metrics:
  - Median estimated Expected Shortfall (ES): equity funds = 10.2 percent; fixed-income funds = 8.3 percent.
  - Median HQLA = 52 percent of total assets; fixed income funds median HQLA = 68 percent.
  - Time-to-liquidation: ≈75 percent of funds can fully liquidate positions between 7 days and 1 month; none of mixed allocation funds can liquidate within one day.
- Adverse-scenario outcomes:
  - Median estimated redemption shock >20 percent.
  - Under adverse scenario, deposit withdrawals to meet redemptions depend on liquidation method: pro-rata withdrawals ≈ half of waterfall-based withdrawals.
  - Liquidity shortfall under extreme cases could reach up to 18 percent of total assets; average liquid buffer depletion by fund type: equity funds 21 percent; mixed allocation 11 percent; fixed income 5 percent.
  - Price-impact channel: a one standard deviation increase in asset-level vulnerability could imply an 18 percent increase in equity price volatility relative to the median (effect more prominent for small-cap equity and levered funds).

### Main stress-test findings — corporate and household sectors
- Nonfinancial corporates (NFCs):
  - Aggregate annual PDs rise, on average, by 0.6 percentage point in the first year of the adverse scenario.
  - Smaller firms suffer larger PD increases; loans to NFCs (excluding real estate) account for >30 percent of bank exposures in the sample.
- Households and mortgages:
  - Mortgage PD in the adverse scenario: 1.9 percent in the first year; with the “5-year/125-percent rule” incorporated mortgage PD = 1.0 percent in the first year.
  - Mortgage LGDs rise, reaching at most 46 percent in the first year driven by house-price declines and higher interest rates.
- Borrower-based macroprudential policy insights:
  - Calibrated caps: DSTI cap = 35 percent (90th percentile); DTI cap = 6 (90th percentile); LTV cap = 95 percent (60th percentile).
  - DSTI cap at 35 percent more than halves the increase in mortgage PDs under the adverse scenario.
  - Borrower-based tools lower household PDs and LGDs with only small implied reduction in aggregate credit and quarterly credit growth.

### Contagion and interconnectedness analysis
- Domestic network and exposures:
  - System sample: 50 institutions (23 banks, 22 insurers, 5 securities firms).
  - Interbank claims ≈35 percent of banks’ total capital buffers.
  - Insurers’ deposits on banks = 13 percent of insurers’ capital buffers.
- Domestic contagion outcomes:
  - Shock to G-SIBs/D-SIBs induces losses ≈10 percent of system-wide capital buffers (contagion index), transmitted mainly through credit shocks.
  - Insurers produce losses close to 5 percent of system-wide capital buffers when shocked.
  - Institutions generally not simultaneously highly impactful and highly vulnerable; domestic bilateral-exposure-based contagion appears limited.
- Cross-border spillovers:
  - Inward spillovers: assumed write-off + funding shortfalls could reduce Japanese banks’ capital by 17 percent on average for shocks from U.S. banks (LGD = 65 percent; discount rate = 20 percent; funding shortfall rate = 35 percent).
  - Outward spillovers: could be sizeable and depend on liquidity buffers of debtor-country financial systems.

### Climate risk analysis — transition and physical risks
- Exposure and scenarios:
  - Emission-intensive sectors (direct + indirect) ≈13 percent of GDP and ≈22 percent of total bank loans.
  - Banking-sector exposure to emission-intensive sectors ≈ one-fifth of assets on average.
  - NGFS-aligned scenarios: Net Zero 2050 (NZ), Fragmented World (FW), Current Policies (CP).
  - Emissions by 2040 relative to CP: NZ = 60% lower; FW = 30% lower.
- Macro and bank capital impacts:
  - Aggregate banking-system capital ratio changes by 2040:
    - NZ scenario: decline ≈0.6-0.7 percentage points (≈0.03-0.04 percentage points per year vs CP).
    - FW scenario: decline ≈0.3 percentage points by 2040 (≈0.03 percentage points per year in the 2030s).
  - Heterogeneity: regional banks more affected; internationally active banks modestly impacted.
  - Contribution decomposition: about one-third to one-half of capital ratio shifts attributable to RWA changes (rising risk weights due to higher PDs/LGDs).
- Firm-level modeling and uncertainty:
  - Firm micro-sample ≈271,000 nonfinancial firms; Monte Carlo sampling across within-industry emission-intensity distributions (50th, 75th, 90th percentiles).
  - At the 90th percentile of firm emission-intensity uncertainty, large PD and LGD increases for certain emission-intensive sectors (e.g., mining, iron and steel).
- Physical climate risk — flood-focused results:
  - About one-third of physical assets potentially exposed to flooding.
  - Flood hazard projections (SSP 585): 100-year and 1000-year flood depths projected to increase by ≈0.5-0.6 meters by 2080.
  - Koide and others (2022) elasticity: one-meter rise in inundation depth → residential land prices −1.1 percent; commercial land prices −4.8 percent.
  - Average projected decline in commercial land prices ≈3 percent for areas exposed to flood risk; for ≈25 percent and ≈5 percent of commercial land exposed to flood risk declines could exceed 4 percent and 10 percent, respectively.
  - Maximum damage rates (buildings): <=0.49m = 0.189; 0.50~0.99m = 0.253; 1.00~1.99m = 0.406; 2.00~2.99m = 0.592; >=3.00m = 0.800.
  - Maximum damage values (E/m2, 2010) converted to 2020 JPY per 500m2: residential = JPY 2,275 million; corporate = JPY 2,814 million (using lower end of 90 percent CI and adjustments).

### Holistic vulnerability assessment of banking sector
- Multi-dimensional ranking using 22 indicators across balance-sheet, solvency, liquidity, interconnectedness, and climate transition metrics.
- Top three most vulnerable banks represent 5 percent of total assets of the 23-bank sample (two internationally active banks and one domestic bank; no SIBs among top three).
- Cluster findings:
  - Regional banks: on average most vulnerable.
  - Internationally active banks: higher vulnerability due to interconnectedness.
  - Domestic banks: higher forward-looking solvency and liquidity vulnerabilities relative to system average.

### Macrofinancial environment and key statistics (selected)
- Real GDP growth averaged about 2 percent per annum during 2021-2023.
- Inflation: headline and core inflation exceeded BOJ’s 2 percent target since April 2022.
- Nominal wage developments:
  - Base pay rise in 2023 = 2.1 percent (up from 0.6 percent in 2022).
  - First two rounds of 2024 wage negotiations: headline wages +5.3 percent; base pay +3.6 percent.
- BOJ policy changes:
  - BOJ abolished YCC and QQE in March 2024 and ended negative interest rate policy while maintaining gross JGB purchases broadly at current pace.
- Sovereign debt: 240 percent of GDP.
- Real estate and markets:
  - Residential real estate (RRE) estimated overvaluation = 17 percent (average).
  - Commercial real estate (CRE) estimated overvaluation = 30 percent (average).
  - RRE prices increased by more than 11 percent in real terms since 2019.
  - TOPIX and Nikkei 225 rose 25 percent and 28 percent in 2023 (y/y), respectively.
- Banking sector liquidity and deposits:
  - Banks’ liquid assets ≈ one-third of total assets.
  - Share of retail deposits = 58 percent; insured deposits = 60 percent (as of December 2022).
  - BOJ October 2023 FSR: about 30 percent of banks have core deposits with maximum maturity ≥10 years; about 20 percent of banks have core deposits with average maturity ≥5 years; nearly 40 percent of core deposits have remaining maturity >5 years.

### Policy implications and prioritized recommendations (timing codes: I Immediate; ST Short Term; MT Medium Term; C Continuous)
- Strengthen liquidity risk analysis for banks (including JPY and FX exposures) and integrate it into systemic risk assessment. (ST) — FSA, BOJ
- Continue developing and operationalizing bank solvency stress test frameworks. (C) — BOJ
- Enhance models to assess real estate price misalignments and fill data gaps. (C) — BOJ
- Develop and operate an investment fund stress test framework. (MT) — FSA
- Regularly conduct economic value-based solvency ratio (ESR) top-down stress tests for insurers and monitor insurers’ liquidity risk. (MT) — FSA
- Develop models and close data gaps to analyze contagion risks among banks and NBFIs. (MT) — FSA
- Further enhance climate risk models for physical and transition risks in collaboration with other agencies and external researchers. (MT) — FSA, BOJ
- Continue compiling securities-holdings data from banks, including hedging of interest-rate and FX risk. (C) — FSA, BOJ
- Foster interagency collaboration for systemic risk analysis and broaden Common Data Platform coverage. (C) — FSA, BOJ

*Source: EXECUTIVE SUMMARY (content unit 1jpnea2024003).*

### EXECUTIVE SUMMARY _____________________________________________________________________________ 9

### EXECUTIVE SUMMARY

### Introduction and Scope
- Contains an INTRODUCTION with subsections:
  - A. Objectives and Scope
  - B. Macrofinancial Environment
  - C. Financial System Structure
  - D. Financial Sector Soundness

### Systemic Risk Analysis
- SYSTEMIC RISK ANALYSIS with subsections:
  - A. Key Vulnerabilities and Risks
  - B. Macrofinancial Scenarios

### Banking Sector Stress Testing
- BANKING SECTOR STRESS TESTING with modules:
  - A. Bank Solvency Stress Test
  - B. Bank Liquidity Stress Test
- Related technical materials and formulas:
  - Box 1. Loan Loss Calculations and Credit RWAs—Summary of Relevant Formulas
- Scenario and sensitivity components referenced in figures and tables (see Figures 21, 32–40; Tables 2)

### Insurance Sector Stress Testing
- INSURANCE SECTOR STRESS TESTING with modules:
  - A. Insurance Firms’ Balance Sheet Structure
  - B. Insurance Solvency Stress Test
  - C. Insurance Liquidity Stress Test
- Scenario calibrations and risk-free term structures (see Figures 47–56; Tables 3–6)

### Investment Funds Liquidity Stress Test
- INVESTMENT FUNDS LIQUIDITY STRESS TEST with:
  - A. Overview of Market Structure
  - B. Methodology
  - C. Results
- Liquidity-weight and HQLA methodology components (see Figures 57–62; Table 7)

### Corporate and Household Sector Risk Analysis
- CORPORATE AND HOUSEHOLD SECTOR RISK ANALYSIS with:
  - A. Methodology
  - B. Results
- Microdata and modeling components for firms (see Figures 63–66; Tables 8–11)

### Interconnectedness and Contagion Analysis
- INTERCONNECTEDNESS AND CONTAGION ANALYSIS with:
  - A. Methodology
  - B. Results
- Domestic contagion modeling and cross-border spillovers (see Figures 68–71; Appendix V)

### Climate Risk Analysis
- CLIMATE RISK ANALYSIS with:
  - A. Climate Transition Risk Analysis
  - B. Climate Physical Risk Analysis
- Models and scenario frameworks include:
  - IMF-ENV Model and Climate Macro Scenarios (Box 2)
  - ENV-FIBA (Environment-Firm and Bank) model layer referenced in tables (see Figures 72–89; Tables 8, 10–12)
- Sectoral and firm-level impacts, projections for flood hazards, and economic effects (see Figures 74–89; Table 12)

### Holistic Vulnerability Assessment of the Japanese Banking Sector and Recommendations
- HOLISTIC VULNERABILITY ASSESSMENT with:
  - A. Methodology
  - B. Results
  - C. Recommendations
- Composite vulnerability rankings and indicators (see Figures 90–91; Table 13)

### Figures and Tables Inventory (selected)
- Figures cover macroeconomic developments, debt, housing and CRE markets, interest rates, market liquidity, financial system structure and interconnectedness, bank and insurer balance sheet detail, stress test results, climate risk visualizations, and holistic vulnerability outputs (Figures 1–91).
- Tables include Main Recommendations, stress calibration parameters, insurers’ shock allocations, liquidity buffer definitions, HQLA weights, firm microdata and PD estimation results, and data on physical vulnerability (Tables 1–13).

### Appendices and Additional Materials
- APPENDICES I–V covering:
  - I. Additional Tables
  - II. Real Estate Market Analysis
  - III. Nonfinancial Corporate Risk Analysis
  - IV. Household Stress Testing
  - V. Domestic Contagion Modeling
- Glossary of acronyms provided (selected entries include: AEs, AFS, AUM, BMA, BOJ, BU, CCoB, CCyB, CET1, CGE, CLO, COF, CP, CRE, D-SIB, DSTI, DTI, EBIT, ENV-FIBA, ES, ESR, FMC, FSA, FSB, FSAP, FW, FX, FY, GDP, GHG, G-SIB, GVA, HFT, HH, HQLA, HTM, ICE, ICR, ICS, IIR, IMF, IPCC, IRB, IRS, JGB, JPY, JSCC, LCR, LGD, LGL, LMTs, LTI, LTV, MOF, MtM, NACE, NBFIs, NFCs, NFCI, NGFS, NII, NPL, NSFR, NZ, OCI, OTC, PCR, PD, PiT, RAM, RCR, REITs, ROA, RRE, RWA, SMEs, SMR, STD, STA, TD, VaR, WEO, WROR, YCC)

*Source: EXECUTIVE SUMMARY (content unit 1jpnea2024003).*

### EXECUTIVE SUMMARY

### EXECUTIVE SUMMARY

### Financial system resilience and evolving risks
- Japan’s large and globally well-integrated financial system withstood the COVID-19 pandemic, aided by strong capital and liquidity buffers and extensive policy support.
- Credit provision to the private sector has remained robust since the pandemic, supporting a steady economic recovery.
- The financial system stands at a critical juncture amid an evolving macroeconomic environment, with key risks including:
  - Market risk and foreign currency (FX) liquidity risk as the Bank of Japan (BOJ) ended its negative interest rate policy and yield curve control (YCC) framework, potentially allowing higher short and long-term domestic interest rates than those prevalent since 2016.
  - Intensification of regional conflicts and geoeconomic fragmentation, global supply chain disruptions, and significant commodity price volatility that could trigger an abrupt global economic slowdown and a surge in inflation, leading to an increase in foreign and domestic interest rates and tightening of global financial conditions.
  - Structural transformations from climate change, digitalization, and an aging population.

### FSAP approach and scenario-based systemic risk analysis
- The FSAP conducted a comprehensive scenario-based systemic risk analysis including:
  - Solvency and liquidity stress tests for a sample of banks and insurance companies under a baseline and hypothetical adverse scenario.
  - Liquidity stress tests for investment funds to assess resilience to investor redemption shocks.
  - Assessment of potential feedback from liquidity stress to solvency risks for banks.
  - Risk analysis for nonfinancial corporates, households, and the real estate sector to input into bank solvency analysis.
  - Climate risk analysis for the same sample of banks subject to the solvency stress test, assessing transition risks toward net zero greenhouse gas emissions by 2050 and high-level physical risk exposure.

### Main stress-test findings
- Banks
  - The solvency stress test indicates the banking system would be able to maintain its solvency position under the hypothetical adverse scenario comprising high inflation and a sharp economic downturn.
  - In the adverse scenario, the aggregate capital shortfall remains small, though some banks may be susceptible to the stress.
  - Liquidity risks are contained at the system level due to a sufficiently strong liquidity position in Japanese Yen (JPY).
  - Some banks are susceptible due to sizeable FX exposures: asset-side undrawn committed credit lines and liability-side FX wholesale funding and FX swap dependence; rising FX funding costs could adversely feed back to profitability and solvency.
- Insurers
  - Insurers, especially life insurers, are sensitive to an increase in domestic and foreign interest rates, though in aggregate their capital remains well above regulatory capital requirements in the adverse stress testing scenario.
  - A decline in equity prices, rising credit risk, and potential default of counterparties are relevant sources of risk under stress.
  - Insurers are not significantly exposed to liquidity risk in aggregate, but some may face pressures under stress.
- Investment funds
  - The investment funds sector has been growing steadily and appears well positioned to accommodate plausibly sized investor redemption shocks under an adverse scenario.
  - Less liquid funds experience larger redemption shocks and could contribute to fragilities in domestic equity markets under more severe shocks.
- Contagion
  - The Japanese financial system is highly interconnected with G-SIBs and D-SIBs playing important roles.
  - Systemic domestic contagion risks based on bilateral exposures appear limited due to strong capital positions of major financial institutions, but broader market sentiment shifts or valuation changes of common asset holdings could exacerbate risks.
  - Some banks and insurers with large claims on other financial institutions relative to their capital appear vulnerable to contagion risks.
  - Cross-country inward and outward financial spillovers could be sizeable.
- Nonfinancial corporates and households
  - Nonfinancial firms face higher probability of default under the adverse scenario due to a slowdown in economic activity and an increase in interest rates, with the impact more pronounced for smaller firms.
  - Household default rates would rise from a very low level due to presumed rises in unemployment and interest rates.
  - Mortgage loss-given-defaults (LGDs) would rise materially due to the assumed drop in house prices.
  - The industry practice (5-year/125-percent rule) could mitigate some of the impact of a sharp increase in interest rates on mortgage payments.
- Climate risk
  - Banking sector exposure to emission intensive sectors is, on average, about one-fifth of their assets.
  - Banks generally appear resilient to a transition to net zero greenhouse gas emissions by 2050 relative to a “current policies” scenario, though the impact on capital positions varies across banks.
  - High-level physical risk analysis shows about one-third of physical assets are at risk of flooding in Japan, with significant variation across prefectures and expected increases in future flood damage rates in some regions.
  - Available climate adaptation indicators suggest Japan has a strong capacity to cope with extreme weather events.

### Macrofinancial environment and key statistics
- Real GDP growth averaged about two percent per annum during 2021-2023, driven by inbound tourism, pent-up domestic demand, and strong policy support.
- Headline and core inflation (excluding fresh food) exceeded the BOJ’s 2 percent target since April 2022.
- Nominal wage developments:
  - Workers received a 2.1 percent rise in base pay in 2023 (up from 0.6 percent in 2022).
  - The first two rounds of wage negotiations in 2024 resulted in a 5.3 percent increase in headline wages (3.6 percent in base pay).
- The BOJ abolished the YCC and the Quantitative and Qualitative Easing (QQE) frameworks in March 2024, and ended the negative interest rate policy, while maintaining gross JGB purchases broadly at the current pace.
- Sovereign debt to GDP is 240 percent of GDP, the highest among advanced economies.
- Residential real estate (RRE) prices increased by more than 11 percent in real terms since 2019.
- Stock market moves: TOPIX and Nikkei 225 rose by 25 percent and 28 percent in 2023 (y/y), respectively.
- Elevated interest rate differentials vis-à-vis other major economies have contributed to persistent depreciation pressures on the JPY and increased U.S. dollar funding costs for Japanese firms.

### Policy implications and recommended actions (Table 1: Main Recommendations)
- Immediate, Short Term, Medium Term, Continuous priorities and institutional leads are summarized below (timing codes: I Immediate (within 1 year); ST Short Term (within 1-2 years); MT Medium Term (within 3−5 years); C Continuous).
- Authorities: Further enhance the liquidity risk analysis for banks with, e.g., stress test tools and models for exposures in both JPY and foreign currencies and make it an integral part of systemic risk assessment. (ST) — FSA, BOJ
- Continue developing the bank solvency stress test model framework. (C) — BOJ
- Further enhance models to assess potential vulnerabilities, including price misalignments in real estate markets to complement indicator-based risk assessments. (C) — BOJ
- Develop and operate an investment fund stress test framework. (MT) — FSA
- Regularly conduct economic value-based solvency ratio (ESR) top-down stress test analysis for insurers and continue to closely monitor insurers liquidity risk. (MT) — FSA
- Develop models and fill data gaps to analyze contagion risks in the financial system among banks and nonbank financial institutions (NBFIs) as part of systemic risk monitoring. (MT) — FSA
- Further enhance the climate risk models and analysis for both physical and transition risks, possibly in collaboration with other governmental bodies and external research institutions. (MT) — FSA, BOJ
- Continue compiling data from banks related to securities holdings, including on hedging of interest rate risk and foreign currency risk. (C) — FSA, BOJ
- Continue to foster interagency collaboration for systemic risk analysis, ensuring that both the BOJ and FSA have necessary access to data. (C) — FSA, BOJ
- Continue to enrich the scope of the Common Data Platform in line with the evolving risk environment. (C) — FSA, BOJ

*This note was prepared by Marco Gross (team lead for systemic risk analysis), jointly with Andrea Deghi, Salih Fendoglu, Rui Xu, Mustafa Yasin Yenice, and Jinhyuk Yoo (all at the IMF’s Monetary and Capital Markets Department).*

### 10.      Liquidity in the JGB market has improved as the BOJ has allowed greater flexibility in

### 10.      Liquidity in the JGB market has improved as the BOJ has allowed greater flexibility in the YCC framework.

### Key developments in JGB market liquidity
- After a widening of the 10-year JGB yield target to ±50 bp by the BOJ in December 2022, volatility of interest rates increased and liquidity indicators (such as the bid-ask spread) in the JGB market deteriorated notably (Figure 8).
- The bid-ask spreads were volatile in January 2023 following greater flexibility in the YCC framework in December 2022.
- Historically, a deterioration in JGB market liquidity has been associated with higher market volatility.
- Distortions in the yield curve have declined with greater flexibility in the YCC framework.

### BOJ responses and market functioning
- In response to liquidity deterioration after December 2022, the BOJ made unscheduled purchases of JGBs, and the liquidity indicators improved.
- During 2023, the increased flexibility allowed by the BOJ in the conduct of YCC helped to smooth the JGB yield curve and improved the functioning of the JGB market.

### Market activity and related indicators
- JGB trading volumes have risen amid higher yields.
- The BOJ provided USD funding amid strains in offshore USD funding markets in March 2020.
- The Japanese stock market has boomed in 2023.

*Source: IMF staff summary from chapter 10 of the provided Japan country report content.*

### 18.      The financial system is characterized by a high degree of interdependence between

### 1jpnea2024003 - 18. The financial system is characterized by a high degree of interdependence between

### Financial–real sector interdependence and asset holdings
- Banks and nonbank financial institutions (NBFIs) have large exposures to each other and to corporate and household sectors.
- Banks and NBFIs hold a notable share of JGBs, though their exposure has declined considerably by about 15 and 10 percentage points, respectively, in the last decade.
- On average, about one-fifth of banks’ assets constitute securities; the share is larger for domestic banks.
- Insurance companies, especially life insurers, hold a sizable share of securities in total assets.

### Cross-border exposures, foreign assets, and FX funding risks
- The share of foreign assets in total assets: banks = 14 percent; NBFIs = 20 percent (the latter increasing by 6 percentage points over the last decade).
- Most foreign assets held by banks are loans and debt securities; insurers hold sizable foreign debt and equities.
- To obtain foreign currency (FX) funding, banks mostly rely on unsecured wholesale funding and FX swaps, making them highly sensitive to an increase in foreign interest rates.
- Overseas exposure of Japanese financial institutions has declined since 2022 due to the sharp rise in U.S. interest rates and FX funding and hedging costs.
- USD liabilities amount to 30 percent of total bank liabilities.
- Banks have undrawn FX commitments through credit and liquidity lines equal to 67 percent of total commitments at end-September 2023.
- For the ten banks categorized as “major banks” by BOJ (2023), foreign loans constitute about one-third of total loans.

### Financial sector soundness: capital, liquidity, and profitability
- Banks entered the pandemic with generally strong capital and liquidity positions; these deteriorated only modestly after the pandemic owing to massive fiscal and monetary stimulus and supportive financial sector measures.
- Banks’ capital ratios remain well above the regulatory minimum, though some recent declines occurred due to valuation losses from overseas securities as foreign interest rates rose.
- Nonperforming loan (NPL) ratios have remained low and fairly stable since the pandemic.
- Banks have maintained sizeable JPY liquidity buffers, with about one-third of total assets being liquid assets.
- As of December 2022, the share of retail deposits = 58 percent; insured deposits = 60 percent, in the Japanese banking sector.
- BOJ’s October 2023 FSR: about 30 percent of banks have core deposits with a maximum maturity of 10 years or more; about 20 percent of banks have core deposits with an average maturity of 5 years or more; nearly 40 percent of core deposits have a remaining maturity of over 5 years.
- Banking profitability has remained structurally weak in a low interest rate, subdued growth environment; low net interest margins have constrained domestic and regional banks’ profitability.
- Internationally active banks, particularly G-SIBs, saw profitability aided in 2022 by a rise in foreign interest rates and JPY depreciation.
- Insurance sector solvency and metrics (March 2023, stress test sample):
  - Solvency margin ratio (SMR): major life insurers = 956 percent; non-life insurers = 840 percent.
  - Economic value-based solvency ratio (ESR): life insurers = 226 percent; non-life insurers = 212 percent.
- Insurance sector pressures: higher FX hedging costs; decrease in unrealized gains on securities; higher natural disaster expenses affecting non-life insurers; annualized premiums for new businesses have risen and returned to pre-COVID-19 levels.
- Ongoing reduction of insurance policies with legacy high guaranteed interest rates has contributed to internal capital generation.

### Real estate exposures and household leverage
- Real estate markets appear overvalued in some areas:
  - Residential real estate (RRE) estimated overvaluation = 17 percent (average).
  - Commercial real estate (CRE) estimated overvaluation = 30 percent (average).
- About 11 percent of banks’ total outstanding credit constitutes CRE loans.
- About 15 percent of banks’ total outstanding credit constitutes retail mortgage loans.
- Of retail mortgage loans, more than three-fourth constitute floating-rate mortgages (BOJ, 2023).
- Households’ debt-service-to-income (DSTI) ratios have remained broadly stable, but the share of housing loans with DSTIs exceeding 30 percent = 27 percent in March 2022.
- Loan-to-income (LTI) ratio at origination has increased from 3.6 in 2015 to 4.2 as of end-March 2022, driven primarily by young-age borrowers (BOJ, 2022).

### Key vulnerabilities and transmission channels
- Three principal vulnerability sources:
  - Significant exposure to domestic and foreign securities held under mark-to-market accounting.
  - Notable FX exposure, including large USD liabilities and reliance on FX swaps and unsecured wholesale funding, creating susceptibility to rising USD funding costs and rollover/run-off risk; banks’ LCR in USD terms rests well below 100 percent.
  - Potential overvaluation in real estate markets (RRE and CRE).
- Additional risk factors:
  - A large share of floating-rate loans: corporate loans = 55 percent; housing loans = 75 percent (BOJ, 2023).
  - Mark-to-market valuation losses on securities and loan default risk if interest rates rise.
  - Material off-balance-sheet FX exposures via undrawn commitments.

### Adverse global downturn scenario and systemic implications
- Potential triggers: intensification of regional conflicts and geoeconomic fragmentation; global supply chain disruptions; significant commodity price volatility leading to an abrupt synchronized global slowdown and inflationary pressures with sharp increases in foreign and domestic interest rates.
- Transmission and impacts under an adverse global downturn:
  - Valuation losses on debt securities held under mark-to-market accounting from rising domestic and foreign interest rates.
  - Increased default risk among leveraged firms and highly indebted households.
  - Potential significant price corrections in real estate markets commensurate with estimated overvaluation levels.
  - Interest rate risk could morph into material default risk given large shares of floating-rate loans.
  - Sharper overseas credit risks for banks amid a global slowdown.
  - Liquidity risk from notable FX liabilities and sizeable undrawn off-balance-sheet commitments if foreign funding costs increase.

*Source: IMF staff assessment in the provided chapter section.*

### 29.      Japan further confronts several ongoing structural transformations. Climate-related

### 1jpnea2024003 - 29.      Japan further confronts several ongoing structural transformations. Climate-related

### Structural transformations and systemic challenges
- Japan is among the largest carbon emitters globally and has pledged to reduce GHG emissions by 46 percent from 2013 levels by 2030, and to achieve net zero GHG emissions by 2050.
- Climate-related transition risks are highly relevant given that almost one-fifth of bank loans are to emission intensive sectors (including sectors depending on inputs from emission-intensive upstream industries).
- Japan has a high degree of exposure to physical climate risk, though it is considered to have a strong adaptive capacity to handle such risks.
- Digital transformation in the financial sector is rapid, offering efficiency gains but raising competitiveness challenges for incumbent financial institutions and increasing cybersecurity risks.
  - More than 200 cyberattacks against critical infrastructure occurred in 2021, the highest level in the past five years (BOJ, 2022).
  - There has been a sharp increase in cyberattacks against Japanese businesses since Russia’s invasion of Ukraine (Teikoku Databank survey).
- Demographics: Japan’s population is aging and shrinking, posing long-term challenges to the economy and the financial system; regional banks in less-populated prefectures have limited business opportunities on average.

### Financial sector exposures and vulnerabilities
- Securities and foreign exposure:
  - Banks have a high share of securities in banking sector assets, particularly domestic banks, though this share has been trending down.
  - The share of FX-denominated securities in total securities averages about 30 percent but is generally higher for internationally active banks.
- Valuation losses and interest rate risks:
  - Amid a sharp rise in foreign interest rates since 2022, Japanese banks registered record valuation losses from foreign bonds, amounting to nearly JPY 3 trillion as of end-March 2023, although overall valuation net gains on securities holdings remained positive given rising stock prices.
  - In an adverse scenario, a further increase in overseas interest rates could weigh on banks’ balance sheets, especially for banks with limited or no interest rate hedging.
  - In such an adverse scenario, the rise in domestic interest rates could partly occur because of an increase in sovereign risk premia, given the high level of sovereign debt in Japan.
- Overseas lending and CLOs:
  - The share of foreign loans in major banks’ total loans has nearly doubled over the last decade and is up by about 5 ppt since the last FSAP.
  - The majority of overseas loans are investment grade (about 60 percent).
  - Major banks increased overseas collateralized loan obligation (CLO) investment since 2016, with over 99 percent concentrated in AAA-rate tranches. As of 2019, CLO holdings amounted to about JPY 20 trillion, less than one-tenth of the size of banks’ overseas loans and credit products; most (almost three-fourths) are held to maturity.
- Commercial real estate (CRE) exposure:
  - Direct exposure of Japanese major banks to overseas CRE market via loans or securities holdings is limited in aggregate—less than 2 percent of their total assets, or less than 5.3 percent of total loans (BOJ, 2023)—though some institutions could be more exposed.
  - Indirect risks from overseas CRE markets could arise through price spillovers to domestic CRE markets, which could be relevant given more material domestic CRE exposure.
- Real estate market signals:
  - House prices have been rising above trend in recent years while affordability has declined.
  - In CRE, the industrial and residential segments have been booming and capitalization rates have remained compressed.
  - Real estate markets appear to be moderately-to-highly overvalued, increasing the likelihood of sizeable corrections.
- Bank foreign currency funding and exposures:
  - USD liabilities are dominated by market funding (i.e., non-deposit funding).
  - Japanese banks' unused off-balance sheet commitments in foreign currency have trended down but remain sizeable relative to total credit lines.
  - Banks’ asset-side FX exposures are dominated by loans and securities. Regional banks hold a higher share of their FX assets in securities than major banks, exposing them to higher market risk but benefiting liquidity. Regional banks make more material use of FX swaps than major banks.

### Macrofinancial scenarios used for stress testing
- Horizon and alignment:
  - Scenarios span a three-year horizon from 2024-2026.
  - Baseline scenario aligned with the IMF’s October 2023 World Economic Outlook (WEO).
  - Adverse scenario represents a global downturn with a global financial market downturn, combining risks defined in the Risk Assessment Matrix (RAM).
- Adverse scenario features:
  - Originates from a global commodity price shock alongside equity market and term premia shocks and implies some appreciation of the JPY.
  - Provokes a sharp increase in inflation and a large negative output gap—comparable in magnitude to that observed in Japan during the COVID-19 pandemic.
  - Domestic interest rates are assumed to initially react to the rise in inflation, but further increases are contained by the wide output gap and a gradual decline in inflation.
  - Weaker aggregate demand and higher interest rates raise NFC sector default probability, pressuring profitability and interest payments; combined with falls in real wages and higher interest rates, private sector default risk rises, real estate demand declines, and a notable correction in real estate prices is triggered.
- Severity and plausibility metrics:
  - The adverse scenario meets the severe yet plausible criterion.
  - The standard deviation (STD) multiple for the adverse scenario regarding real GDP growth is 1.8x when measured based on 2-year cumulative GDP growth under the adverse scenario relative to the historical mean and STD of GDP growth in Japan.
  - Relative to the growth rate of 2023 (instead of to historical mean growth), the adverse STD multiple amounts to 2.2x.
- Sensitivity analyses conducted on top of the initial adverse scenario include:
  - Interest rate sensitivity: short-term rates (1Y sovereign bond yields) move to 1.5 percent (instead of 1 percent in year 1 under the initial adverse scenario); 10Y JGB yields move to 3 percent (2.25 percent in year 1 under the initial adverse scenario).
  - GDP growth sensitivity: Real GDP growth is assumed to drop further to -3.2 percent and -1 percent in the first two years (compared to -1.8 percent and -0.5 percent under the initial adverse scenario, respectively).
  - Additional sensitivity variants include further pronounced shocks to Japanese short- and long-term interest rates, GDP growth, and FX valuation shocks.

### Banking-sector stress testing: methodology and coverage
- Coverage:
  - Top-down solvency stress test covered internationally active and domestic banks, including regional banks.
  - Twenty-three banks, including thirteen regional banks, were covered, representing 82 percent of the banking system’s total assets as of 2023Q1.
- Model features:
  - The solvency stress test model captured credit risk, market risk (including interest rate risk), and other profit and loss drivers, with dynamic balance sheets.
  - The model explicitly accounted for interest rate risk hedging and FX risk hedging using detailed bank data and a fine-grained set of portfolio types by investment category and numerous economic sectors.
  - Tax credit (deferred tax assets), which can be applied under Japanese capital regulation, were not accounted for in the bank solvency stress test; this omission implies adverse scenario impacts on banks’ solvency ratios would be to an extent less adverse if the tax credit mechanism were accounted for.
- Credit risk module:
  - Employed a “model-flows-imply-stocks-philosophy”: sub-models provide default rate and implied default flow projections from which NPL stocks and other metrics are calculated.
  - Required parameters: cure rates, write-off rates, and gross loan growth assumptions (informed by macrofinancial scenario).
  - Write-off rates held constant at bank-portfolio specific initial values. Cure rates held constant in baseline and stressed under the adverse scenario.
- Loss given default (LGD) and RWAs:
  - Structural LGD models: household LGDs projected primarily as a function of residential house prices; corporate and consumer credit segments used Frye-Jacobs methodology allowing LGDs to move as a function of PDs.
  - Dynamics of credit risk weighted assets (RWAs) driven by default rates, implied default migration effects, downturn LGDs, and general loan growth.
  - Twenty-two of the twenty-three banks follow the internal ratings-based (IRB) approach (either A- or F-IRB); one bank follows the standardized approach.
  - Basel risk weight formulas employed for performing IRB bank-portfolios. PD through-the-cycle (TTC) inputs used as a smooth function of bank-portfolio-specific default rate paths with smoothing factor set to 0.2.
  - Downturn LGDs kept at initially observed levels under both baseline and adverse scenarios.
  - “Pseudo-T0” risk weights implied to generate “pseudo RWA” starting points; changes from pseudo starting points attached to banks’ observed T0 RWA starting points to obtain RWA paths at bank-portfolio level.
  - Risk weights for nonperforming IRB exposure held constant at initially observed portfolio values. For the standardized approach bank, risk weights held constant for performing and nonperforming exposures so overall RWA effects arise from loan growth and default migration.

*Source: IMF staff and data as presented in the provided content unit.*

### Box 1. Japan: Loan Loss Calculations and Credit RWAs—Summary of Relevant Formulas

### Box 1. Japan: Loan Loss Calculations and Credit RWAs—Summary of Relevant Formulas

### Balance sheet and exposure projection
- Gross exposures, 퐿퐿_t, projected at bank-portfolio level using gross loan growth trajectory 푔푔_t (mean quarterly growth rate).
- Dynamic vs. static balance sheet:
  - 퐿퐿_{t+1} = 퐿퐿_t (1 + 푔푔_{t+1})
- Exposures and related metrics projected at quarterly frequency.

### NPLs, performing exposures, and default dynamics
- Nonperforming exposure stock evolution:
  - 푁푁푁푁퐿퐿_{t+1} = 푁푁푁푁퐿퐿_t (1 − 푊푊푊푊푊푊푊푊_{t+1} − 퐶퐶푈푈푊푊𝑅𝑅 푊_{t+1}) + 푁푁𝑃𝑃_{t+1} (퐿퐿_t − 푁푁푁푁퐿퐿_t)
  - 푊푊푊푊푊푊푊푊_t: NPL write-off rate (quarterly flow window)
  - 퐶퐶푈푈푊푊𝑅𝑅 푊_t: quarterly cure rate (percentage of NPLs moving back to performing)
  - 푁푁𝑃𝑃_t: quarterly default rate
- Performing exposure stock implied as residual:
  - 푁푁퐿퐿_{t+1} = 퐿퐿_{t+1} − 푁푁𝑁𝑁퐿퐿_{t+1}
- Note on cure rates: Given limited historical data, judgmental stress assumptions used. For first and second years, cure rates fall to 0.75x and 0.85x, respectively, the level of the baseline at bank-portfolio level and equal the baseline cure rates in the third year.

### Provision stocks and provision coverage ratios (PCR)
- Provision stock for performing exposures:
  - 푁푁𝑃𝑃_{t+1}^{푃푃푃푃} = 푁푁퐶퐶푊푊_{t+1}^{푃푃푃푃} × 푁푁퐿퐿_{t+1}
- PCR for performing exposures evolves with smoothing parameter 0.2:
  - 푁푁퐶퐶푊푊_{t+1}^{푃푃푃푃} = 푁푁퐶퐶푊푊_t^{푃푃푃푃} + 0.2 ( 푁푁𝑃𝑃_{t+1} × 퐿퐿퐿퐿𝑃𝑃_{t+1} − 푁푁𝑃𝑃_t × 퐿퐿퐿퐿𝑃𝑃_t )
  - Rationale: reflects partial movement from time-invariant to forward-looking, time-varying provisioning based on expected losses.
- Provision stock for nonperforming exposures:
  - 푁푁𝑃𝑃_{t+1}^{푁푁푃푃푃푃} = 푁푁퐶퐶푊푊_{t+1}^{푁푁푃푃푃푃} × 푁푁𝑁𝑁퐿퐿_{t+1}
- PCRs for NPLs floored at LGDs from Year 1 forward and move in line with PiT LGD trajectories:
  - 푁푁퐶퐶푊푊_{t+1}^{푁푁푃푃푃푃} = max( 푁푁퐶퐶푊푊_{t=0}^{푁푁푃푃푃푃}, 퐿퐿퐿퐿𝑃𝑃_{t+1} )
  - Implication: potential under-provisioning for NPLs is “corrected” in year 1 (additional one-off loss and capital impact), and provisioning dynamics reflect LGD thereafter.

### Capital impacts from provisions and write-offs
- Capital impact from performing exposures:
  - 퐿퐿퐿퐿_{t+1}^{푃푃푃푃} = 푁푁𝑃𝑃_{t+1}^{푃푃푃푃} − 푁푁𝑃𝑃_t^{푃푃푃푃}
- Capital impact from nonperforming exposures:
  - 퐿퐿퐿퐿_{t+1}^{푁푁푃푃푃푃} = 푁푁𝑃𝑃_{t+1}^{푁푁푃푃푃푃} − 푁𝑁𝑃𝑃_t^{푁𝑁푃푃푃푃} + 푊푊푊푊푊푊_{t+1} × 퐿퐿퐿퐿퐿퐿_{t+1} × 푁𝑁𝑁𝑁퐿퐿_t
- Total periodic capital impact from loan losses for bank b:
  - ∆퐶퐶퐶퐶𝑊_{b,t+1} = ∑_{p=1}^{푃} 퐿퐿퐿퐿_{p,t+1}^{푃푃}
  - Dependences: change in risk parameters, change in performing exposure balances, change in volumes and LGDs for outstanding NPLs, and additional impact from write-offs if LGL differs from provision coverage at write-off time.

- Expanded nonperforming exposure capital-impact decomposition (terms A and B):
  - 퐿퐿퐿퐿_{t+1}^{푁푁푃푃푃푃} = 푁𝑁𝑃𝑃_{t+1}^{푁𝑁푃푃푃푃} − 푁𝑁𝑃𝑃_t^{푁𝑁푃푃푃푃} + 푊푊푊푊푊푊_{t+1} × 푁푁𝐶𝐶푊𝑊_{t+1}^{푁푁𝑃𝑃푃푃} × 푁𝑁𝑁𝑁퐿퐿_t  (A)
    + 푊푊푊푊푊푊_{t+1} × ( 퐿퐿퐿퐿퐿퐿_{t+1} − 푁푁𝐶𝐶푊𝑊_{t+1}^{푁푁𝑃𝑃푃푃} ) × 푁𝑁𝑁𝑁퐿퐿_t  (B)
  - Term A ensures write-off driven falls in provision stock are not mistaken as releases (profits).
  - Term B captures additional gain/loss from gap between LGD at time of write-off (LGL) and provision coverage just before write-off.
  - The short form is recovered as 푁푁𝐶𝐶푊𝑊_{t+1}^{푁푁𝑃𝑃푃푃} cancels out.

- Loss-given-loss (LGL) assumption and transformation from LGD:
  - 퐿퐿퐿퐿퐿퐿_{t+1} = 퐿퐿퐿퐿𝑃𝑃_{t+1} / (1 − 훼 × 퐶퐶퐶퐶퐶퐶퐶퐶𝑊𝑊)
  - 훼 set to 0.5 (not 1) because a portion of the NPL stock is already closer to write-off and hence its LGD and provision coverage already closer to the LGL.

### Credit risk RWAs and sources of change
- IRB performing exposure risk weights (RWs) projected using IRB RW formulas.
- RWs for nonperforming exposures held constant.
- STA risk weights for performing and nonperforming exposures held constant.
- Three sources for changes in credit risk RWAs for a bank:
  - Gross loan growth under a dynamic balance sheet
  - IRB performing RW changes
  - Migration effects

### Empirical inputs and model scope
- Initial risk parameter distributions across banks and portfolios depicted in Figures 22-31.
- Sample: 23 banks subject to solvency and liquidity stress test.
- Historical data span for econometric models: 2002Q1-2023Q1 (23 models for interest income rates (IIRs) and 23 cost-of-funding (CoF) models).
- Cost-of-funding variable: four-quarter trailing sum of quarterly interest expense flows divided by four-quarter average of liabilities.
- Cure rate stress assumptions: first year 0.75x baseline, second year 0.85x baseline, third year equals baseline.

### Key results and numeric findings
- Aggregate Common Equity Tier 1 (CET1) ratio of the Japanese banking system falls by 510 basis points in the first year of the adverse scenario but remains well above the regulatory minimum.
- Valuation losses contribute to a decline in the banking system capital ratio by 6.3 percentage points in the first year of the adverse scenario.
- Loan losses grow three-fold under the adverse scenario relative to the baseline scenario.
- Sample/models and data notes:
  - 23 banks modeled for IIR and CoF.
  - Data span for models: 2002Q1-2023Q1.
  - Figures referenced: Figures 22-31 for initial risk parameter distributions; Figure 32 for pass-through coefficients and sensitivities.

### Additional modeling notes (interest income, market risk, other P&L)
- Net interest income projections:
  - IIR defined as four-quarter trailing sum of quarterly interest income flows divided by the four-quarter trailing average of interest-bearing assets.
  - Predictor variables for IIR models: banks’ own CoF spread (spread to 1Y JGB), 1Y JGB yield, 10Y JGB yield, 10Y US treasury spread to 10Y JGB, 10Y euro area spread to 10Y JGB.
  - Nonperforming exposures assumed to not generate interest income in solvency stress test.
- Cost-of-funding (CoF) model predictor variables (five core variables and expected coefficient signs):
  - (1) short-term market/policy rates (including the 1Y sovereign yield) [+]
  - (2) 1Y US treasury bond yield [+]
  - (3) realized volatility metric for the Japanese stock market [+]
  - (4) BOJ’s total asset growth [−]
  - (5) banks’ individual regulatory capital ratio [−]
  - Solvency-funding cost feedback found economically significant and positively correlated with banks’ wholesale funding shares.
- Market risk module central formula: modified duration-based valuation for bonds accounting for interest rate hedging (Macaulay durations, current yields, interest rate sensitivities collected at bank-portfolio level, hedges partially represented via instrument-specific interest rate swaps).
- AFS filter: assumed not present for domestic banks in the stress test; starting point capital ratios adjusted to reflect unrealized gains at onset.
- Other P&L:
  - Net fee and commission income (NFCI) modeled via panel Bayesian Model Averaging (BMA) relating NFCI/TA to macrofinancial indicators.
  - “Other operating expense” (mostly salary) held constant as ratio to total assets gross of provisions.

*Source: IMF staff box on loan loss calculations and credit RWAs (Box 1), from the Japan staff report.*

### 45.      Overall losses are dominated by valuation losses from bonds and equity, and a sizeable

### 45.      Overall losses are dominated by valuation losses from bonds and equity, and a sizeable portion of those in turn stem from foreign investments.

### Loss composition and contributions
- Combined loan losses and valuation losses: bond and equity investments contribute 90 percent; loan book contributes 10 percent.
- Loan book share of loan book and bond and equity investments: 60 percent.
- Loan book loss contributions:
  - Domestic exposures: 79 percent.
  - Foreign exposures: 21 percent.
- Bond and equity investments loss contributions:
  - Domestic exposures: about 50 percent.
  - Foreign exposures: about 50 percent.
- Held-to-maturity (HTM) portfolios exclusion: For bonds, these shares exclude the HTM portfolios of banks, which amount to about 20 percent of total bonds.

### NPLs and loan loss dynamics
- Aggregate NPL ratio (system level):
  - Initial: 0.8 percent.
  - Third year (under adverse scenario): 2.5 percent.
- Domestic banks and regional banks NPL ratios:
  - Initial: 1.5 percent.
  - Third year: 3.6-3.8 percent.
- NPLs peak in the third year because defaults accumulate and outflows from NPL stock due to write-offs do not outweigh inflows from defaulting exposures over the stress test horizon.
- Loan loss rate (cost of risk) dynamics:
  - System aggregate loan loss rate:
    - Initial year: 0.1 percent.
    - First year of adverse scenario: 0.6 percent (six-fold rise).
  - Domestic banks and regional banks first-year loan loss rates: 1.0-1.1 percent.
- Portfolio-level contributions to loan losses (approximate ranges):
  - Domestic nonfinancial corporate lending: around 35-40 percent.
  - Domestic mortgages: 16-20 percent.
  - Domestic consumer credit: less than 5 percent.

### Capital adequacy outcomes and buffer use
- Banks facing capital shortfalls under adverse scenario:
  - Three banks (CET1 hurdle 4.5 percent).
  - Four banks (total capital hurdle 8 percent; domestic banks’ core capital ratio hurdle 4 percent).
  - These banks amount to:
    - 0.04 percent of 2022 nominal GDP (three banks, CET1 shortfall).
    - 0.05 percent of 2022 nominal GDP (four banks, total capital shortfall).
- Cluster membership and asset share:
  - Three of the four banks with total capital ratio below the hurdle rate belong to the regional bank cluster.
  - The four banks represent five percent of total assets of the 23 banks in scope of the stress test.
- Capital Conservation Buffer (CCoB) effects:
  - CET1: eight banks would consume their CCoB.
  - Total capital ratios: twelve banks would consume their CCoB.
  - The three and four banks with capital shortfalls would fully deplete their CCoB as they fall further below the hurdle rate.
- Notes on measurement:
  - For the domestic bank cluster, core capital ratio is used and was adjusted to take unrealized gains/losses at the onset into account.
  - The AFS filter was “switched off” for domestic banks for comparability with international banks (for which no AFS filter is in place).
  - Deferred tax credit effects were not accounted for; their inclusion would reduce the impact of the adverse scenario to an extent.
  - Additional surcharges on capital requirements for G-SIBs and D-SIBs in the sample were included in the hurdle rates.

### Sensitivity analysis — interest rates, hedging, and AFS filter
- Interest rate sensitivity analysis:
  - Banking system aggregate capital ratio drops by an additional 1.1 percentage points from the initial adverse scenario to reach 8.9 percent.
  - Driver: larger valuation losses on securities through higher interest rates, and further rising loan losses as interest rate risk morphs into default risk.
  - Domestic banks and regional banks are more sensitive due to higher share of security holdings.
  - Count of banks with total capital below hurdle rate doubles to eight; five belong to regional bank cluster.
  - System-wide total capital shortfall rises from 0.05 percent of GDP to 0.3 percent of GDP.
- Interest rate hedging counterfactual:
  - Switching off hedging would make aggregate banking system capital ratio fall by an additional 130 basis points in the first year compared to the initial adverse scenario.
  - Overall shift in capital ratios would be 640 basis points.
  - Ten banks would see their total capital ratios fall below the hurdle rates.
  - Note: hedging-related data may to an extent be incomplete.
- Activating the AFS filter for domestic banks:
  - Starting point for the banking system moves from 15.1 percent to 14.7 percent, reflecting valuation gains difference for domestic banks at the outset.
  - Capital ratio drop under adverse scenario:
    - With AFS filter on: 340 basis points.
    - Initial setting with AFS filter deactivated: 510 basis points.
  - For the underlying domestic banks sample, adverse scenario results are notably more benign when AFS filter is on.

### FX positions and USD funding sensitivity
- Net open FX positions:
  - Majority of banks: small relative to capital.
  - Two banks: net open FX positions relative to capital of about 40 percent at end-FY2022.
  - These two banks are driven by mismatch between USD assets and liabilities and are either not or only partially FX hedged.
  - Aggregate assessment: banking system judged well balanced in terms of currency mismatches.
- USD funding cost counterfactual:
  - Shock informed by historical distribution of JPY-USD currency basis swap spreads, set to +70 bps (last observed in 2017).
  - Shock assumed to prevail over rising “base” U.S. interest rates in the adverse scenario.
  - Combined shock: U.S. short-term (1Y sovereign) rate shift in adverse scenario (+150 bps from year 0 to year 1) coupled with additional 70 bps spread shock implies combined +220 bps funding cost shock for Japanese banks’ USD funding.
  - Impact:
    - For some selected banks of the 23, capital ratios would fall by more than 0.5 ppt (and in very selected cases more than 1 ppt) relative to the initial adverse scenario.
  - Note: a notable share of the regional bank cluster is classified as “internationally active” banks.

### Bank liquidity stress test — methodology
- Framework:
  - Cash flow-based liquidity stress test accounting for solvency-liquidity interactions.
  - Horizon: six months; weekly time steps (24 weeks).
  - Assumes hypothetical, severe liquidity outflow assumptions and a sales hierarchy for securities.
  - Securities under fair value accounting are sold first; amortized cost securities (HTM, and AFS for domestic banks) sold afterwards if necessary.
  - Liquid asset buffers defined as:
    - Reserve holdings at the central bank.
    - Cash in own vaults.
    - Interbank placements (loans and deposits from asset side).
    - Equity investments.
    - Derivative-related inflows.
    - Bonds.
- Data inputs and calibration:
  - Detailed supervisory data: maturity profiles, asset-side bond exposures’ duration, interest rate hedging information.
  - Run-off rates for liabilities informed by Liquidity Coverage Ratio (LCR) run-off calibration for a “base case” stress calibration (Table 2).
  - Equity price shock and interest rate shifts aligned with solvency stress test.
  - First-year solvency shocks assumed to instantaneously apply at onset of liquidity simulation.
  - Sales hierarchy note: Trading and available-for-sale securities sold first if reserves insufficient; followed by HTM assets (and AFS for domestic banks) if needed.
  - Liquidity stress test does not take account of banks’ management actions such as raising additional private/public funding.
  - LCR-type run-off rates were modified because in the liquidity stress test they apply to contractually maturing portions per week; hence run-off rates need to be higher than under the LCR.

- Run-off contribution calibration (banking system, first 30 days):
  - Unsecured wholesale funding: about 50 percent of total liability-driven outflows.
  - Liability-side derivatives (including FX swaps; about half U.S. dollar-related): 30 percent.
  - Retail unsecured funding: 10 percent.
  - Secured wholesale funding and drawdowns from credit and liquidity lines: residual about 10 percent.
  - For domestic banks and regional banks:
    - Derivatives contribution does not exceed 5 percent.
    - Unsecured retail funding contribution: 35 percent (domestic banks) and 20 percent (regional banks).

### Bank liquidity stress test — results
- System resilience:
  - Japanese banks generally resilient against hypothetical severe liquidity stress based on all-currency liquidity stress test results.
  - Five banks would face a liquidity shortfall, accumulating to 1.7 percent of initial liquid asset buffers among the 23 banks.
  - The five banks represent 7 percent of total assets of the 23 banks in scope.
  - They include three regional banks.
- Liquid asset buffer depletion:
  - Various liquid asset buffers are depleted to a significant extent in aggregate.
  - Held-for-trading securities are generally small and are ranked first in the sales hierarchy; they are depleted fully at the system level.
- Feedback from liquidity to solvency:
  - For two out of 23 banks, selling mark-to-market securities to honor outflows implies a decline in capital ratios of more than 2 percentage points.
  - For one additional bank, the impact amounts to a decline of 1.2 percentage points.
  - One bank experiences a sizeable impact (larger than a decline of 2 percentage points) without a liquidity shortfall because sale of securities suffices to avert its liquidity shortfall.
  - Two banks with liquidity shortfalls do not face feedback to capital because the bonds they sell are under fair value accounting already.

*Source: IMF staff calculations and supervisory data as presented in the provided content.*

### 60.      A bank liquidity risk-related sensitivity analysis considers a hypothetical case whereby

### 1jpnea2024003 - 60.      A bank liquidity risk-related sensitivity analysis considers a hypothetical case whereby

### Bank liquidity stress-test key findings
- Under a hypothetical case where run-off rates for liabilities double compared to the initial stress calibration:
  - The number of banks with a liquidity shortfall rises to twelve.
  - The liquidity shortfall accumulates to about 6 percent of initial liquid asset stocks.
- In the baseline liquidity stress test described:
  - Five banks fail the liquidity stress test.
  - The first (fifth) bank faces a liquidity shortfall after three (18) weeks.
  - The liquidity shortfall accumulates to 1.7 percent of initial liquid asset stocks by the end of the 6-month horizon.
- Aggregate buffer depletion observations:
  - Various liquid asset buffer types are depleted to a significant extent in aggregate over the banking system.
  - HFT exposures are generally small for all Japanese banks; they are ranked first in the sales hierarchy under stress and hence are depleted fully at system level.
- Notes on reserves and sales hierarchy:
  - Reserves include reserves at central bank, cash in vaults, sight and term deposits at other banks, repo backflows, and backflows from derivatives.
  - For domestic banks, AFS bond holdings are subsumed by HTM bond categories.
  - Impacts pertain to the banking system sum over banks; use of reserves and equity, which are sorted before bonds in the sales hierarchy, do not need to reach 100 percent before other liquid asset buffers (bonds) are used, reflecting underlying bank heterogeneity.
- Rationale:
  - More severe run-off scenarios better reveal relative liquidity risk of banks while abstracting from more severe impacts in absolute terms.

*Sources: BOJ; FSA; and IMF staff calculations.*

### FX liquidity risk for banks
- Size and composition of FX exposure:
  - Liabilities in USD alone represent 30 percent of total liabilities.
  - Foreign currency funding is sizeable in particular for internationally active banks: it represents about 30 percent of their total liabilities, while for domestic and regional banks it amounts to a rounded 10 percent.
- USD funding characteristics and risks:
  - USD funding is obtained largely through FX swaps and unsecured wholesale funding, making banks susceptible to rollover risks and the risk of rising USD funding cost.
  - The cost of USD funding for Japanese banks has been rising in line with rising U.S. policy and market interest rates since March 2022, but the spread for Japanese banks remains particularly compressed since the pandemic, after only an occasional, short-lived shock for about a week in March 2020.
- Bank-cluster differences (regional vs. major/domestic):
  - Regional banks:
    - Appear to face more pronounced liquidity risks in some dimensions, including through FX exposure.
    - Face a lower liquid asset ratio than other banks.
    - Have smaller reserve ratios than other bank clusters.
    - Appear to make yet more use of FX swaps than major banks.
    - On the positive side: higher share of stable (rather than unstable) retail deposits; use more secured borrowing from the BOJ; FX off-balance sheet exposures smaller than for other banks.
  - Liability structure notes:
    - Retail and wholesale unsecured funding dominates liabilities in JPY and currencies other than the USD.
    - Retail unsecured is dominant for domestic banks.
    - Regarding USD funding, derivatives (largely FX swaps) and unsecured wholesale funding dominate at banking system level. For domestic banks, secured wholesale funding weighs more strongly than unsecured wholesale funding.
- LCR and off-balance-sheet findings:
  - All-currency LCRs generally rest well above the 100 percent reference value.
  - Domestic banks are somewhat better placed in terms of median LCRs, with larger cross-bank heterogeneity.
  - Off-balance sheet exposure from credit and liquidity lines are low for domestic banks and largest for international banks. On a USD basis, internationally active banks are more exposed than other clusters.
- Counterfactual committed credit-line drawdown scenario:
  - Assuming an outflow of yet undrawn USD committed credit lines:
    - Some Japanese banks would see their liquidity position weakened substantially, while their all-currency LCRs would remain above 100 percent.
    - The liquidity pressure would be generally confined to banks with large committed, yet undrawn credit and liquidity lines—tending to be large internationally active banks, mostly G-SIBs.
    - Their all-currency LCRs would remain above 100 percent, falling from 150 percent at the outset to 120 percent on average under the counterfactual.
    - The U.S. dollar LCRs of these banks rest below 100 percent at the outset and would fall between 30-40 percentage points under the counterfactual drawdown scenario.

### Insurance sector: balance-sheet structure (as of March 2023)
- Aggregate and sectoral asset composition:
  - Aggregate assets of insurance firms in the sample amount to JPY 360 trillion, majority attributable to life insurers.
  - Major life insurers:
    - Invested about 45 percent of their assets in JGBs and bonds of municipalities and public sector entities.
    - Invested another 17 percent in corporate bonds.
  - Non-life insurers:
    - Exposure to government bonds at 20 percent.
    - Notable exposure to equities that comprised 40 percent of their total assets (half of which toward foreign equities).
  - Foreign bonds account in aggregate for 15 percent of insurers’ portfolio, with large variation across insurers (some life insurers: foreign bonds >60 percent of total fixed-income investments).
  - Corporate bond quality: 97 percent of securities rated at BBB or above.
- FX exposures and hedging:
  - FX-denominated assets:
    - Life insurers: 22 percent of assets.
    - Non-life insurers: 33 percent of assets.
  - FX-denominated liabilities:
    - Life insurers: 8 percent.
    - Non-life insurers: 7 percent.
  - Hedge ratios:
    - Life insurers: about 45 percent of the total foreign asset position hedged.
    - Non-life insurers: 23 percent hedged.
  - In response to rising FX hedging costs, life insurers reduced foreign bond holdings and shifted to domestic long-term bonds and foreign stocks.

### Insurance sector: duration and climate exposure
- Duration:
  - Median asset duration for major life insurers: 12 years.
  - Median net liabilities duration for major life insurers: 17 years.
  - An increase in long-term interest rates makes it attractive for insurers to increase holdings of very long-term JGB instruments to extend asset duration and reduce interest-rate risk.
- Non-life insurance climate exposure:
  - Non-life insurers cover wind and water damage in property insurance:
    - Wind damage accounts for almost 100 percent of property insurance.
    - Water damage accounts for around 70 percent.
  - Climate change poses significant risk due to potential increased severity and frequency of such damage.
  - Property insurance is typically renewed annually, allowing premiums to be adjusted more frequently despite long-term climate risk horizons.

### Insurance solvency stress-test methodology and high-level results
- Sample and coverage:
  - Stress tests conducted on twelve of the largest life insurance companies and ten of the largest nonlife insurance companies, representing 72 percent and 93 percent of premiums in the respective sectors.
  - Data at solo-level of consolidation, current as of March 31, 2023.
- Approaches and metrics:
  - Stress tests assessed credit and market risks under:
    - Current solvency margin ratio (SMR).
    - A market-consistent valuation using the Economic value-based Solvency Ratio (ESR).
  - Top-down solvency stress tests:
    - Conducted by FSAP team under SMR and by authorities under ESR.
  - Bottom-up (BU) stress tests:
    - Conducted by insurers under SMR with same scenarios as top-down.
  - Shocks applied instantaneously at reference date (March 31, 2023).
- Additional shocks specific to insurance exercise:
  - Simultaneous default of the largest banking counterparty and of the largest nonfinancial corporate counterparties.
  - Lapse risk: a (non-permanent) mass lapse event leading to 20 percent of insurance policies being discontinued in the bottom-up stress test exercise.
- Aggregate solvency metrics and risk contributions:
  - Participating firms’ aggregate initial solvency ratio stands above the regulatory threshold of 200 percent, with wide dispersion across individual insurers.
  - Average SMR:
    - Life insurers: about 955 percent.
    - Nonlife insurers: 840 percent.
  - Average (median) ESR:
    - Life insurers: 225 percent.
    - Nonlife insurers: 211 percent.
  - In capital requirement computation:
    - Market risks dominate, contributing to 70 percent of capital requirements overall and 81 percent for life and non-life insurers, respectively.
    - Among market risk components, equity risk contributes most across insurers, followed by currency risk.
    - Life insurance risk is the second largest contributor to capital requirement in life insurers.

### Insurance stress-test scenario calibrations (selected macrofinancial shocks)
- Equity Scenario:
  - Japan -27%
  - United States, Euro area -32%
  - Other advanced economies -25%
  - Emerging economies -35%
  - Holdings in related undertakings -15%
- Currencies Scenario:
  - JPY effective exchange rate +3.4%
  - JPY-USD: Nominal +6.5%
  - JPY-EUR: Nominal +2.8%
  - JPY-AUD: Nominal +1.5%
- Real estate Scenario:
  - Residential, domestic -12.1%
  - Commercial, domestic -22.6%
  - Residential, other countries -15.0%
  - Commercial, other countries -25.0%
- Corporate bond spreads Scenario (non-financials):
  - AAA +50bp
  - AA +60bp
  - A +90bp
  - BBB +140bp
  - BB or lower +220bp
  - Unrated +200bp
- Financials corporate bond spreads Scenario:
  - AAA +80bp
  - AA +100bp
  - A +140bp
  - BBB +200bp
  - BB or lower +250bp
  - Unrated +210bp
- Risk-free interest rates Scenario:
  - JPN, 1 year +1.1 p.p.
  - JPN, 10 years +1.75 p.p.
  - USD, 1 year +1.5 p.p.
  - USD, 10 years +2.5 p.p.
  - Other, 1 year +1.5 p.p.
  - Other, 10 years +1.5 p.p.
- Other investments Scenario:
  - Structured notes and collateralized securities -5%
  - Other investments, not classified elsewhere -5%
- Notes on calibration:
  - A positive value of JPY-USD implies an appreciation of JPY relative to USD.
  - Input variables not included in the macrofinancial scenario are calibrated using a non-parametric methodology that simulates forward in time the variables while conditioning on risk factors assumed in the adverse scenario.

### Additional insurance-sector shocks (explicit calibrations)
- Lapse risks:
  - Mass lapse event assumed to imply that 20 percent of insurance policies for which discontinuance would result in an increase in insurance liabilities are discontinued. No permanent shock is assumed.
- Simultaneous default write-off/write-down assumptions:
  - 100 percent write-off for equity exposures, guarantees for liabilities.
  - 50 percent write-down for bonds and loans.
  - 15 percent write-down for deposits, loaned securities, and lease assets.

*Sources: BOJ; FSA; insurance firms’ data submissions; and IMF staff calculations.*

### 74.      To complement the short-term perspective of an instantaneous shock, insurance firms

### 1jpnea2024003 - 74.

### Methodology and Projection Framework
- Participating insurance firms provided a three-year projection of:
  - gross written premiums, investment returns, average guaranteed interest rate, and net surplus.
- Projections aligned with macrofinancial scenarios (e.g., GDP and interest rate assumptions).
- Market prices of asset-side investments were assumed to remain constant after shocks at the beginning of year 1; only interest payments received, dividends, and rental incomes contribute to investment returns.
- Reactive management actions (e.g., altering underwriting standards, changing reinsurance, de-risking by selling equity/high-yield corporate bonds and buying sovereign bonds) were not modeled; balance sheets were assumed static.

### Assumptions Regarding Capital Standards
- Solvency Margin Ratio (SMR) formula:
  - Solvency margin ratio = Total Amount of Solvency Margin / (Total amount of risks x 1/2) x 100
- Regulatory SMR thresholds and supervisory actions:
  - SMR ≥ 200 percent: problems concerning management soundness requiring potential remedial actions are understood not to arise.
  - Category 1 (SMR between 100 percent and 200 percent): submission and implementation of a business improvement plan.
  - Category 2 (SMR between 0 percent and less than 100 percent): submission and implementation of a plan for adequate solvency and potential measures including prohibition/limitation of dividends, policy dividends, distribution of surplus to policyholders, change in premium calculation for new policies, prohibition/limitation of directors’ bonuses, limitation of other operating costs.
  - Category 3 (SMR less than 0 percent): total suspension of operation for a limited period.
- Economic Solvency Ratio (ESR) regime:
  - Required capital is calculated per risk category with prescribed methodology calibrated at 99.5 percent VaR over one year (corresponding to a 1-in-200 years event).
  - Under the economic value-based regime, the relevant supervisory threshold for first remedial action would be an ESR of 100 percent.
  - Many required capital measures: changes in net assets under prescribed stresses; factor-based approaches for some exposures; natural catastrophe risk based on stochastic models.

### Stress Test Implementation (Top-Down)
- Adverse scenario shocks applied to investment assets and insurance liabilities.
- Haircuts applied to market value of directly held assets; investment fund holdings stressed by underlying asset-class shocks (no look-through to individual securities).
- Fixed-income assets re-valued with stressed term structure for each major currency.
- Interpolation/extrapolation of reference interest rate structure performed using the Smith-Wilson extrapolation method; adverse scenario entails a parallel shift of the entire term structure for all other currencies.
- In economic value-based balance sheet, liabilities evaluated market-consistently; qualifying capital resources adjusted to economic value-based net asset.

### Top-Down Results — SMR-Based Outcomes
- Aggregate available capital drops by about 48 percent under the adverse scenario (life insurers).
- Life insurers:
  - Average SMR declines from 956 percent before stress to 583 percent in the adverse scenario.
  - Ten of twelve life insurance companies remain above the statutory SMR requirement of 200 percent; two do not meet the requirement.
  - Almost three quarters of the decline in available capital is attributable to assumed domestic equity price decline and increase in domestic and foreign interest rates.
- Nonlife insurers:
  - Available capital declines by 24 percent under the adverse scenario.
  - Average SMR declines from 840 percent to 686 percent.
  - All nonlife insurers remain above the statutory SMR hurdle rate.

### Bottom-Up Results and Comparison with Top-Down
- Bottom-up (BU) results broadly aligned with top-down (TD) results:
  - Ten of twelve life insurers remain above the 200 percent SMR requirement under the adverse scenario (BU).
  - Counterparty and credit risks have considerable impact on the most affected insurers.
- Differences between BU and TD:
  - Post-stress available capital in the BU exercise is in aggregate slightly lower than TD.
  - Required capital is lower after stress in BU relative to TD:
    - Reduction in required capital relative to TD amounts to 17 percent for life insurers.
    - Reduction in required capital relative to TD amounts to 11 percent for nonlife insurers.
  - (Footnote comparison: in TD analysis required capital on average is 12 percent as calculated by reduction in risk of portfolio exposures without lapse risk.)

### Profitability and Three-Year Projections
- Net investment returns (spread between investment returns and guaranteed interest rates) decline markedly in the first year of the adverse scenario, contributing to lower insurance profits.
- Net surplus projections:
  - Life companies: net surplus suffers under the adverse scenario and improves only gradually after the shock (recovery over three years).
  - Non-life firms: continued impact on net surplus; already declining in baseline due to natural disasters, rising COVID-19 related payments, and increased auto insurance loss ratios for the largest firms.
- No significant effect of the adverse scenario on gross premium projections in additional projections.

### ESR-Based Results (Economic Value-Based)
- Under ESR, Japanese insurers broadly resilient under the adverse scenario (re-calculation focused on selected risk modules; market risk module capital charges proportionately adjusted).
- ESR changes:
  - ESR decreases by 67 percentage points for the average life insurer.
  - ESR decreases by 41 percentage points for the average nonlife insurer.
  - Post-stress ESR for all insurers remains above the prescribed hurdle rate of 100 percent.
- Aggregate capital resources:
  - Decrease by 30 percent in aggregate across insurers.
  - Larger decline for the group of nonlife insurers: about 44 percent decline.
- Drivers under ESR:
  - Decline in equity prices and credit shocks mainly drive changes in available capital.
  - Life insurers: increase in long-term interest rates tends to reduce liabilities, partially offsetting valuation losses on securities.
  - Nonlife insurers: decline in equity prices dominates, contributing about 60 percent to overall change in capital resources.
- Assets-to-liabilities ratio:
  - Not materially impacted for the average life insurer under ESR adverse scenario.
  - Decreases by 19 percentage points for the average non-life insurer.

### Key Risks Highlighted
- Rise in interest rates beyond assumed adverse scenario may trigger policyholder surrenders of legacy policies to buy new policies with higher guaranteed rates, potentially causing:
  - Large losses when insurers liquidate bond portfolios backing surrendered policies amid rising rates.
  - Compressed underwriting profitability if higher inflation increases claim payments and premium rate increases are not possible.
  - These surrender-related risks may be more prevalent for nonlife insurers where underwriting profitability is more sensitive; life policies have surrender costs and individual protective features that may reduce switching.
- Limitations in top-down model: granularity of supervisory data limits recognition of financial hedges, stop-loss arrangements, financial reinsurance; reactive management actions and dynamic de-risking strategies not modeled.

### Sensitivity Analysis and Single-Factor Shocks
- Single-factor shocks considered separately from macrofinancial scenario; cover biometric risks, catastrophic events, and additional market shocks.
- Biometric shocks for life business:
  - Longevity: a permanent 20 percent decline in mortality rates.
  - Pandemic event: temporary 35 percent increase in disability and morbidity rates over one year, and temporary 10 percent increase in mortality rates over one year.
- Catastrophic events for non-life insurers:
  - Great Kantō earthquake (Japan, 1923); The Ise Bay Typhoon; catastrophic events in 2018 (torrential rains in western Japan, typhoons, earthquakes) as prescribed by the FSA.
  - Participants to provide reinsurance recoveries from top five reinsurers for each catastrophic event.
- Additional market shocks:
  - Increase in key domestic interest rates: +1.6 pp in short-term rate (one-year government bond), +2.5 pp long-term rate (10-year government bond).
  - Appreciation/depreciation of JPY relative to USD, AUD, EUR: +/- 10 percent.

### Sensitivity Findings
- Large sensitivity to interest rates:
  - A larger increase in domestic interest rates (+1.6 pp short-term, +2.5 pp long-term) lowers life insurers’ SMR-based available capital by 24 percent on average.
- FX sensitivity:
  - Appreciation of JPY by 10 percent reduces SMR by 3 percent on average; for some insurers impact as large as 14 percent.
  - Variation explained partly by differing FX hedging across insurers; hedge ratio generally moderately high.
- Nonlife insurers remain relatively limited in sensitivity to the more severe interest rate scenario compared with life insurers.

*Source: IMF staff calculations and companies’ submissions as presented in the referenced chapter.*

### 92.      The sensitivity analysis on ESR confirm the results of the main stress test exercise. An

### 1jpnea2024003 - 92.      The sensitivity analysis on ESR confirm the results of the main stress test exercise. An

### Sensitivity analysis and solvency results
- Sensitivity analysis on ESR confirms the results of the main stress test exercise.
- Additional rise in domestic interest rates under the sensitivity analysis:
  - Provides more income from investments in yen-denominated bonds, partially mitigating higher hedging costs and fluctuations in stock prices.
  - Can be potentially positive for some insurers investing in longer dated government bonds.
- Aggregate solvency position under presumed pandemic and catastrophic natural events (individually) remains resilient.
  - Median sensitivity across insurers to biometric shocks is close to zero (Figure 55).
  - On average, solvency margin for life insurances after the assumed pandemic shock decreases by one percent and increases by 0.2 percent following a longevity shock.
  - Overseas catastrophic events lead on average to a 7 percent change in capital sources for non-life insurers.
  - A repetition of an event such as Great Kantō could cause losses of about 9 percent of the available capital for Japanese non-life insurers exposed to that region.
- Japan Earthquake Reinsurance:
  - Could face almost 60 percent change in its capital resources.
  - Largest reinsurance recoveries reported up to JPY 12 trillion for events as severe as Great Kantō’s earthquake, but less for other historic catastrophic events.
- Note on single-factor shocks for insurers: additional single-factor shocks cover biometric risks, catastrophic events as well as selected market shocks. (Banks’ sensitivity analyses assume short-term and long-term interest rates increasing to 1.5 percent and 3 percent.)

*Source: IMF staff calculations based on companies’ data submissions.*

### Insurance liquidity stress test — methodology
- Liquidity stress test complements solvency risk analysis with in-depth examination of liquidity sources and needs under the main adverse scenario and insurance-specific shocks.
- Variation margining due to interest rate swaps (IRS) exposures assessed using transaction-level data submitted by insurers participating in the solvency stress test exercise.
  - Data sample: reported transactions from only ten life insurers and three nonlife insurers; insurers required to report a minimum of 70 percent (in notional amount) of overall transactions active at the reference date.
- Two liquidity assessment approaches:
  - Stock-based approach: liquidity needs and sources from assets and liabilities at a reference date; applies haircuts to assets and liabilities.
  - Cash flow approach: compares projected or realized liquidity sources and needs over 90 days from the reference date.
- Variation margining analysis scenarios:
  - One-day abrupt market movement without allowance for collateral transformation.
  - Prolonged market turmoil (2-week market movement) allowing collateral transformation of liquid assets.
  - Interest rate shocks examined vary from 50 to 250 basis points; parallel shifts in risk-free rate curves for all currencies are simulated.
- Definitions of liquidity buffers:
  - Scenario 1 (1-day market movement): cash only.
  - Prolonged turmoil: cash and tradable high-quality liquid assets (HQLA), with gradually broadened definitions of tradable assets (Table 4).
- Reference dates and baseline conversions:
  - Reference date for stock-based approach: March 31, 2023.
  - Cash flow approach baseline uses in- and outflows recorded over April 2023 to June 2023 period.
- Two assumptions for liquidity positions:
  - Fixed (no reactive management actions; only nondiscretionary actions in place at reference date).
  - Constrained/relaxed (allows reactive post-stress management actions aligned with stressed scenario).

### Insurance-specific shocks and allocation across modules
- Mass lapse, mortality, pandemic morbidity, increases in frequency and severity, reinsurance in-flows, reduction in written premia, and catastrophic events included.
- Examples from Table 5 (selected prescribed shocks and calibrations):
  - Mass lapse shock: Instantaneous discontinuance — 10%; Term insurance 20%; Endowments 20%.
  - Reinsurance in-flows: Annuities in deferral phase — ; Annuities in pay-out phase — ; Pure unit-linked contracts (without financial guarantees) -10%; Unit-linked contracts with financial guarantees 20%.
  - Written premia shocks: Disability 20%; Health -10%.
  - Line-of-business shocks to severity and frequency (direct business and accepted non-proportional reinsurance):
    - Medical expense insurance: Shock to Severity 2% ; Shock to Frequency 15%
    - Income protection insurance: Shock to Severity 2% ; Shock to Frequency 15%
    - General liability insurance: Shock to Severity 2% ; Shock to Frequency 15%
    - Credit and suretyship insurance: Shock to Severity 2% ; Shock to Frequency 15%
    - Legal expenses insurance: Shock to Severity 2% ; Shock to Frequency 15%
    - Miscellaneous financial loss: Shock to Severity 2% ; Shock to Frequency 15%
    - Non-proportional health, casualty, and property reinsurance: Shock to Severity 2% ; Shock to Frequency 15%
- Allocation of shocks across stress test modules (Table 6):
  - S = Solvency risk analysis; L = Liquidity risk analysis.
  - Market shocks: Adverse macrofinancial scenario allocated to S, L, and sensitivity analysis.
  - Underwriting shocks: Mass Lapse (life only) allocated to S, L; Mortality (life only) allocated to L and sensitivity analysis; Pandemic morbidity and increase in cost of claims (nonlife only) allocated to L; Increase in frequency L; Increase in severity S, L; Reinsurance in-flows L; Reduction in written premia L; Catastrophic events S; Other shocks include default of relevant counterparties allocated to S.

### Liquidity assessment indicators and implementation details
- Liquidity indicators used:
  - Ratio of liquid assets to total assets under narrow and broad definitions.
  - Estimated net cash-flow, customizable by business or product portfolio.
  - Liquid asset to net flows ratio (ratio of estimated liquid assets after haircuts to net flows) as a sustainability indicator.
- Cash flow adjustments:
  - Inflows adjusted for written premiums (existing and new business).
  - Life business technical cash outflows include surrender cash outflows and account for impact on lapses.
  - Non-life business cash outflows for claim settlement incorporate rise in cost of claims.
- Timing of frequency and severity shocks in liquidity analysis:
  - Payments incurred up to the reference date: subject solely to severity shock.
  - Payments incurred after the reference date: subject to both frequency and severity shocks.
  - Shock to frequency is excluded from solvency ratio calculations and determination of the liability stock in liquidity analysis.

### Liquidity stress test results
- Margin calls and interest rate swaps:
  - Cash margin calls on IRS following a 250 basis points increase in domestic interest rate could be met by drawing on about 20 percent of cash equivalents.
  - Variation margin calls likely do not lead to sector-wide phenomenon given limited size of interest swap positions across the market, but notable firm-level differences exist.
  - Cash positions of at least two insurers could be inadequate relative to IRS positions under severe one-day market movements.
- Stock-based approach findings:
  - Share of liquid assets is relatively high.
  - Life insurers’ liquid assets ratio stands at about 60 (70) percent under the baseline (adverse) scenario (Figure 56).
  - Haircuts applied to assets reflect loss of value and fire-sale externalities; haircuts to liabilities reflect ease of redeeming a liability.
  - Encumbrance levels for life insurers’ high-quality sovereign bond holdings are not trivial relative to the size of such exposures.
- Flow-based approach findings:
  - Large net outflows projected under adverse scenario, surging to more than JPY 7 trillion in aggregate for life insurers.
  - Allowing for reactive management actions, including asset sales, limits net outflows; net outflows generally met once tradable securities are considered part of the liquid buffer stock, but not always on a case-by-case basis.
  - Forced sale of assets comprises primarily domestic sovereign bonds: for the median life insurer, sales of domestic sovereign bonds represent 60 percent of total sales, with significant cross-sectional variation.
- Supervisory implications:
  - Heterogeneity in impact across insurers underlines importance for supervisors to monitor liquidity risks.
  - Insurers with the largest negative net flows under the adverse scenario include companies with higher solvency risk.
  - Conducting stress testing exercises to assess cash outflows under different market liquidity conditions is instrumental for liquidity risk analysis and monitoring.

*Source: IMF staff calculations and analysis (content based on insurers’ data submissions and IMF staff methodologies).*

### 112.      The asset management sector in Japan has grown steadily in recent years. Total net

### 1jpnea2024003 - 112.

### Growth and Market Structure of Japan’s Investment Trusts
- Total net assets of investment trusts stood at more than JPY 300 trillion in 2023, up from about JPY 170 trillion in 2015.
- Publicly offered investment trusts account for 60 percent of the total net assets of investment trusts; privately offered investment trusts account for about 35 percent.
- Investment returns and flows:
  - Sustained growth supported by robust investment returns—10 percent for all equity investment trusts.
  - Continuous net fund inflows; about 61 percent of the total cash inflow into publicly offered investment trusts was directed towards global equities.
  - Approximately 54 percent of the total assets were held in foreign currencies.
- Governance and delegation:
  - Nearly 90 percent of actively managed investment trusts have delegated their investment decisions to asset management firms located overseas.
- Product composition and sectoral considerations:
  - Majority of publicly offered investment trusts invest mainly in traditional equity, debt securities, and their associated derivatives.
  - Open-ended investment trusts (majority of publicly offered trusts) have portfolios comprised mainly of products with short periods between contract and delivery.
  - According to the Bank of Japan’s flow of funds statistics, about four percent of the Japanese depository financial institutions’ assets comprised investments in securities investment trusts, as of December 2022.
  - The real estate fund sector, particularly REITs, has been challenged by increased office vacancies and declining rents in the Tokyo area after the COVID-19 pandemic, weakening the Tokyo Stock Exchange REIT Index.

### Stress Testing Methodology
- Objective:
  - Assess resilience of individual funds to significant, yet plausible, investor redemption shocks given potential liquidity mismatches between assets and liabilities.
- Sample and data:
  - Sample predominantly based on selected investment funds from an FSA survey launched in 2023.
  - Sample comprises investment funds with total AUM of more than JPY 50 billion, amounting to 72 percent of total AUM for investment trust products.
  - Additional selection used commercial data sources; time-series monthly net flows and portfolio composition obtained from commercial data sources for the period 2006-2023.
  - Final number of investment funds used in the stress test exercise: 183.
  - Market characteristics: highly concentrated with a few large investment funds; leverage less common among mixed allocation funds and more widespread among funds with investments concentrated in the U.S.
- Four-step stress test:
  1. Set a redemption shock using various value-at-risk approaches and a macroeconomic model relating fund flows to macrofinancial variables.
  2. Calculate liquidity buffers on the asset side of the balance sheet.
  3. Assess resilience based on available liquidity to meet redemption shocks under specific asset liquidation assumptions.
  4. Estimate contributions of funds’ illiquidity to asset price fragilities by measuring impact on market price volatility for specific asset classes.
- Liquidity management tools (LMTs):
  - Stress test does not consider use of LMTs for two main reasons: (1) uncertainty regarding their use and impact due to limited empirical work; and (2) stress tests aim to assess resilience without mitigating measures. Availability of LMTs at the fund level could potentially mitigate risks.

### Modelling Assumptions and Calibration
- Redemption shock calibration:
  - Historical distribution approach using Value-at-Risk (VaR) percentiles (e.g., 1st percentile) and expected shortfall (ES).
  - ES defined as the average value of net flows below the VaR; formal definitions preserved in source.
- Homogeneity vs. heterogeneity assumptions:
  - Homogeneity: each fund within a specific fund type faces the same-sized redemption shock.
  - Heterogeneity: redemption shock calibrated separately for each fund based only on its historical data.
  - The heterogeneity assumption is used as a benchmark; homogeneity also examined.
- Model-based approach:
  - Satellite model translates macrofinancial scenarios into net flows at the fund level.
  - Model estimated for the period 2006-2023 and projects net flows using values of macrofinancial variables in the adverse scenario.
  - Model inputs include: ΔVIX, Δ3M (monthly change in 3-month interbank rate), Δ10Y (change in 10-year domestic sovereign yield), Term spread, ΔNikkei 225, ΔS&P500, change in Japanese corporate bond spreads, and Δnominal effective exchange rate.
- Liquidity buffer measurement:
  - Two approaches: time-to-liquidation (IOSCO guidelines, assuming no fire-sale discounting) and HQLA-based liquidity bucket (inspired by Basel III LCR).
  - HQLA computation splits portfolio by asset class and ratings, defines eligible assets, and applies haircut values (liquidity weights).
  - Example weight treatment preserved: corporate debt securities rated above BBB- eligible; 15 percent reduction applied to corporate bonds with AA- or higher ratings; 50 percent reduction applied to bonds in the A+ to BBB- range.
  - Liquidity index formula preserved: LP_LIQUIDITY_y,t = sum_k (omega_k × s_t,k) where omega_k is liquidity weight and s_t,k is share of security k in percent of total net assets.
- Liquidation strategies:
  - Waterfall (horizontal slicing): use HQLA before less liquid assets.
  - Pro-rata (vertical slicing): liquidate proportionally across assets to maintain portfolio composition.
  - Managers may use a mix of approaches in stress periods.
- Resilience metrics:
  - Redemption Coverage Ratio (RCR) = liquid assets available over net outflows (redemption shocks).
    - If RCR > 1: portfolio sufficiently liquid to cope with redemption scenario.
    - If RCR < 1: liquidity profile may worsen when redemption scenario occurs.
  - Liquidity shortfall when RCR < 1: additional assets to be sold, expressed as percentage of fund’s TNA.
- Funding-channel assessment:
  - Scenario assumes funds exposed to redemption pressures may withdraw deposits, potentially creating liquidity challenges for depository banks; effect measured simply by assuming all deposits are with a depository bank.
- Price impact and asset price vulnerability:
  - Fund-level illiquidity measure constructed as weighted average of bid-ask spreads of assets held by the fund.
  - Asset price vulnerability then calculated as weighted average of investing funds’ illiquidity with weights representing funds’ relative holdings of the asset.

### Stress Test Results and Key Statistics
- Historical net outflows:
  - Redemption shock calibrated using the first percentile of historical net outflows to capture tail risk events.
  - Net outflows were more pronounced during the global financial crisis relative to the COVID-19 market turmoil in March 2020.
  - Net outflows generally larger for equity funds relative to fixed income and mixed allocation funds.
- Expected Shortfall (ES) findings:
  - Median estimated ES is 10.2 percent for equity funds and 8.3 percent for fixed-income funds.
  - Large heterogeneity across funds: ES distribution skewness with troughs up to negative 50 percent.
- Time-to-liquidation and liquidity buffers:
  - Majority of publicly offered investment trusts invest mainly in equity, debt securities, and associated derivatives; funds hold large share of products with short contract-to-delivery periods (1 to 3 days).
  - About 75 percent of the funds can fully liquidate their portfolio position between seven days up to a one-month horizon.
  - Results similar for equity and fixed income funds; mixed allocation funds tend to face longer periods of liquidation.
  - None of the mixed allocation funds reported positions that can be liquidated within one day.
- High Quality Liquid Assets (HQLA):
  - Median HQLA is 52 percent of the total assets position.
  - Fixed income funds have the largest median HQLA at 68 percent, with considerable variation within that group.

*Source: IMF staff calculations and source text.*

### 134.      Overall, the redemption shocks in the adverse scenario are typically larger than the

### 1jpnea2024003 - 134. Overall, the redemption shocks in the adverse scenario are typically larger than the

### Investment funds: redemption shocks and resilience
- Median estimated redemption shock of more than 20 percent.
- Under the adverse scenario a more significant percentage of investment funds must withdraw deposits to cope fully with the redemption shock.
- Deposit withdrawal (bank) outcomes depend on liquidation method:
  - On a pro-rata basis, total bank deposit withdrawal tends to be about half of the withdrawal needed when adopting a waterfall approach (selling HQLA first before less liquid assets).
- Stress test headline results:
  - Most Japan-domiciled funds in the sample are resilient to the considered shock.
  - Under more severe redemption shocks, the share of funds with lower liquidity buffers can widen, increasing liquidity shortfalls.
  - Liquidity shortfall associated with outflows consistent with the adverse scenario could reach up to 18 percent of total assets.

### Fund-type heterogeneity in liquid buffer depletion
- Average liquid buffer depletion by fund type:
  - Equity funds: 21 percent (on average).
  - Mixed allocation funds: 11 percent (on average).
  - Fixed income funds: 5 percent (on average).
- Presence of money market mutual funds with large cash and cash equivalent positions (more than 40 percent of the total portfolio) contributes to the more limited impact on fixed income funds.

### Sensitivity analysis on equity weight and fund failure rates
- Equity weight for HQLA computation varied from 30 to 70 percent (parameter range 30–70).
- Lower equity weight implies lower quality of equity holdings in the HQLA computation.
- Assuming a larger percentage of low-quality equity holdings yields up to 28 percent of the funds may not meet redemption shocks.
- Under the homogeneity assumption (same shock across funds), stress test outcomes were no worse for the sample.

### Price impact and asset fragility
- Assets held by less liquid funds are more susceptible to selling pressure during large redemptions, evidenced in COVID-19 and early monetary policy normalization episodes.
- Quantitative estimate: a one standard deviation increase in asset-level vulnerability of funds domiciled in Japan could imply an 18 percent increase in equity price volatility (relative to the median).
- Effect is more prominent for small-cap equity and can be exacerbated by funds with greater leverage.

### Implication for monitoring and supervision
- Vulnerabilities from fund redemptions could adversely affect specific asset markets concentrated with such holdings.
- Recommendation: strengthen systemic risk monitoring of the investment funds sector through broader data coverage and regular liquidity stress testing as part of offsite monitoring of firms.

*Sources: Bloomberg; Lipper; Factset; and IMF staff calculations.*

### Corporate and household sector risk analysis — methodology overview
- NFC PDs and LGDs:
  - Projected using a model anchored in firm-level data from Moody’s Orbis.
  - Fixed effect firm-level panel regression using Moody’s KMV default rates as primary input.
  - Structural simulation at firm-micro level for predictors (cash to debt ratios, leverage = debt over assets, interest coverage ratios, etc.).
  - Firm LGDs derived using an embedded Vašíček model structure.
- Household PDs and LGDs:
  - Anchored in Japan Household Panel Survey (JHPS/KHPS) data.
  - Structural micro-macro simulation model of Gross and Poblacion (2017) and Gross and others (2022).
  - Model simulates each household’s balance sheet dynamics (including employment status) for 2024–2026 conditional on macrofinancial scenarios.
  - Model accounts for Japanese mortgage market features: full-recourse system, high share of floating-rate mortgages, and industry practice of the “5-year/125 percent rule.”

### Corporate and household sector risk analysis — key results
- NFC sector:
  - Aggregate annual PDs rise, on average, by 0.6 percentage point in the first year of the shock under the adverse scenario.
  - PD increases more under sensitivity analysis (additional interest rate increase) but remain below global financial crisis levels due to sizeable cash buffers accumulated during the pandemic.
  - Loans to NFCs (excluding real estate) account for more than 30 percent of bank exposures in the sample.
  - Smaller firms experience larger PD increases relative to large firms, driven by higher initial leverage and lower interest coverage and cash buffers.
- Household sector:
  - Mortgage probability of default (PD) in the adverse scenario reaches 1.9 percent in the first year.
  - When taking the “5-year/125 percent rule” into account, mortgage PD reaches one percent in the first year.
  - Mortgage LGDs rise under adverse scenarios, reaching at most 46 percent in the first year.
  - LGD increases are driven mainly by a material drop in residential house prices and higher interest rates (larger discounting of expected recovery value).
  - Mortgage PD trajectory is somewhat lower when considering the 5-year/125 percent rule.

### Policy analysis — borrower-based macroprudential tools (DSTI, LTV, DTI)
- Grid-based assessment:
  - Examined full range of caps; LTV caps primarily affect LGD, DSTI caps primarily affect PD; PDs react notably to all caps due to positive correlation between DSTIs and LTVs across households.
- Concrete cap calibrations informed by ratio distributions:
  - DSTI cap set at 35 percent (corresponds to 90th percentile of the sample).
  - DTI cap set at 6 (corresponds to 90th percentile of the sample).
  - LTV cap set at 95 percent (corresponds to 60th percentile of the LTV distribution).
- Impact of borrower-based tools:
  - DSTI cap at 35 percent more than halves the increase in mortgage PDs under the adverse scenario (with or without the 5-year/125 percent rule).
  - Tighter borrower-based policy caps induce lower mortgage PDs and LGDs across the examined grid.
  - Introducing borrower-based tools lowers household PDs and LGDs while not having a significant impact on credit growth.
  - Implied reduction in aggregate credit is small; decline in quarterly credit growth is less than (figure-based quantification presented in source).

*Source: IMF staff estimates; household and corporate model inputs and scenarios as described in the text.*

### 1.5 percentage points.

### 1jpnea2024003 - 1.5 percentage points.

### Mortgage policy caps: calibration, effects, and design considerations
- The calibrated policy caps are illustrative and not “optimal”; more granular credit registry data would be required to inform cap levels and pacing meaningfully.
- Points authorities should consider to mitigate unintended consequences:
  - Keep a microprudential policy perspective, which could impact real estate markets and complement macroprudential policy.
  - Be mindful of possible leakages.
  - Consider phased implementation of measures, such as through “speed limits.”
- Model mechanics and behavioral responses:
  - Under the 5-year rule, a fraction of floating-rate mortgages is repriced.
  - Households can endogenously lower mortgage PDs at the expense of lower consumption.
  - Policy caps induce a mild impact on private credit growth.
- Note: The PD metric (middle chart) is a debt weighted population aggregate, based on the underlying household-level PDs in the model.
- Source of policy-cap methodology reference: Gross and Poblacion (2017).

### Interconnectedness and contagion analysis — Methodology
- Network mapping:
  - Disaggregated data on gross bilateral exposures (loans, short-term money placements, equity, bonds, and banks’ undrawn commitments) are aggregated for network mapping.
  - Nodes represent financial institutions, colored by sector (banks, insurers, securities firms) and, among banks, by G-SIB/D-SIB, internationally active (excluding G-SIBs/D-SIBs), or domestic bank.
  - Edges colored by source node; layout based on Fruchterman-Reingold algorithm.
  - Corroborating statistics: (1) number of direct links (sum of inward and outward linkages); (2) gross bilateral claims of each sector as percent of total sectoral capital buffers.
- Contagion structural model (domestic):
  - Follows Covi, Gorpe, and Kok (2021).
  - Trigger: hypothetical failure of a financial institution inducing credit shocks (losses on claims depending on LGD rates) and funding shocks (fire-sale losses).
  - Fire-sale losses depend on liquidity surplus (high-quality liquid assets in excess of net funding outflows), available-for-sale asset volumes, and a dynamic discount rate (depends on share of central bank eligible assets and system-wide fire sales, up to a cap).
  - Institutions fully realize losses impacting capital; defaults occur if capital or liquidity buffers are insufficient. Simulations run until no further defaults.
  - Bilateral exposures and exposure-specific LGDs sourced from proprietary data sources.
  - Further calibration details provided in Appendix V.

- Cross-border contagion model:
  - Uses a simpler structural model (Espinosa-Vega and Sole, 2010) at the country-level (consolidated banking system).
  - Bilateral exposures sourced from BIS Locational Banking Statistics (consolidated ultimate borrower basis); aggregate capital from Fitch/Bankscope.
  - Given limited cross-country data, LGD, discount rate, and funding short-fall rate assumed homogenous across countries; distress threshold and provisions assumed zero; results do not account for liquidity buffers.

### Domestic contagion — key results
- System structure:
  - The Japanese financial system is highly interconnected, with G-SIBs and D-SIBs central in the domestic network.
  - Analysis sample: 50 financial institutions (23 banks, 22 insurers, 5 large securities firms).
- Sectoral bilateral claims relative to capital buffers:
  - Interbank claims average 35 percent of banks’ total capital buffers.
  - Banks’ claims on securities firms amount to 9 percent of banking sector’s total capital buffers.
  - Insurers’ deposits on banks are 13 percent of insurers’ total capital buffers.
  - Deposits, followed by loans, are the main types of bilateral exposures.
- Contagion impact summary:
  - A shock to G-SIBs/D-SIBs induces losses of about 10 percent of system-wide capital buffers (contagion index), transmitting primarily through credit shocks rather than funding shocks.
  - Insurers are the second most impactful sector, exerting losses close to 5 percent of system-wide capital buffers.
  - Vulnerability at cluster-average level appears moderate, up to one percent of own capital buffers.
- Cross-sectional institution-level findings:
  - A shock to a bank can exert losses on other banks that can reach as high as 14 percent of the banking sector capital buffers.
  - The average loss a bank can face due to individual shocks to other banks reaches close to 5 percent of its own capital buffers.
  - An insurer’s average loss due to failure of others reaches 30 percent of own capital buffers (though the loss is small in JPY terms).
- Overall assessment:
  - Domestic systemic contagion risks due to bilateral on-balance sheet exposures appear limited; institutions are generally not both highly impactful and highly vulnerable simultaneously.
  - BOJ liquidity provision against collateral appears helpful in containing systemic contagion risks.
- Caveat: Results based solely on bilateral balance sheet exposures; do not reflect potential changes in market sentiment or valuations in commonly held assets in response to shocks considered.

### Cross-border spillovers — key results
- Inward spillovers:
  - Cross-border inward spillovers from failure of foreign financial institutions could be sizeable, strongest emanating from the U.S.
  - In an assumed scenario where a portion of Japanese banks’ claims on U.S. banks are written off and moderate funding shortfalls occur, the capital of Japanese banks would, on average, decline by 17 percent.
  - Exposure to French banks and, to a lesser extent, Canadian banks also exerts notable inward spillovers.
  - The degree of inward spillovers has been broadly stable over the last decade.
  - Model assumptions for cross-border results: LGD = 65 percent, discount rate = 20 percent, funding shortfall rate = 35 percent. Inward spillovers driven primarily by assumed LGD given Japan’s creditor status.
- Outward spillovers:
  - Cross-border outward spillovers operate mainly through counterparties facing funding shortfalls and could be sizeable.
  - Geographic distribution of outward spillovers aligns with Japan’s importance as a creditor for recipient economies.
  - Overall impact depends on liquidity buffers held by exposed financial institutions of debtor countries.
- Caveat: Cross-country model does not consider liquidity buffers of counterparties; including such buffers could imply lower losses.

### Climate risk analysis — transition risk focus
- Relevance and context:
  - Transition risks are highly pertinent for Japan given its status as a major carbon emitter and government plans to reduce GHG emissions to net zero by 2050.
  - Japan has relied mainly on fossil fuels since the Fukushima nuclear accident in 2011; fossil fuel power generation has remained above 70 percent.
- Power generation shares (2021, Agency for Natural Resources and Energy in Japan):
  - LNG 34.4 percent, coal 31.0 percent, oil 7.4 percent, solar 8.3 percent, hydropower 7.5 percent, nuclear 6.9 percent, others 4.4 percent.
  - Current fossil fuel power generation level: 73 percent (lower than 2012 peak of 89 percent).
- Emission-intensive sectors:
  - Seven emission intensive industries (ranging from electricity and gas to basic metal and other non-metallic mineral products) collectively contribute about 80 percent of Japan’s total CO2 emissions.
  - Direct and indirect emission-intensive sectors together represent approximately 13 percent of GDP and account for about 22 percent of total bank loans in Japan.
  - Analysis focuses on both direct and indirect emission-intensive sectors; direct emission intensive sectors include mining and quarrying, electricity and gas, and water transport. Sectors relying on emission-intensive inputs include fabricated metal, transport equipment, and paper products.
- Comparative position:
  - Among G7 countries, Japan stands out with the second-highest output share of sectors with high emission intensity, behind Canada.

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

### 163.      The government of Japan has pledged to substantially reduce greenhouse gas

### 1jpnea2024003 - 163.      The government of Japan has pledged to substantially reduce greenhouse gas

### Japan’s climate commitments and policy framework
- Japan set an interim target to reduce GHG emissions by 46 percent from 2013 levels by 2030, with an objective of achieving net-zero GHG emissions by 2050.
- Green Transformation Promotion Act enacted in May 2023 and a comprehensive strategy based on the law:
  - Emphasizes upfront investment of JPY 20 trillion over the next decade in decarbonization initiatives.
  - Funding to be largely provided through issuance of Japan Climate Transition Bonds.
  - Plans to introduce a carbon levy on fossil fuel supplies from FY2028.
  - Emissions trading system in high-emission industries starting from FY2026; allowance auctioning to power generation companies to be gradually phased in from FY2033.
- Current carbon tax (Tax for Climate Change Mitigation) implemented in 2012:
  - Set at JPY 289 (approximately $2) per ton of CO2 equivalent, with certain exemptions.
  - Japan’s effective carbon price level is the second lowest among G7 peers, following the United States (OECD, 2021).

### Scenario framework and macro scenarios used for transition risk analysis
- Integrated micro-macro simulation links IMF-ENV (global CGE) model with a micro simulation layer for firms and banks:
  - Micro layer covers approximately 271,000 nonfinancial firms from Japan, focused on emission-intensive sectors.
  - Bank module comprises 22 banks (one bank excluded for lacking emission-intensive portfolios).
- Three focal NGFS-aligned scenarios:
  - Net Zero 2050 (NZ): global warming limited to 1.5°C above pre-industrial levels; global net-zero CO2 emissions by around 2050.
  - Fragmented World (FW): delayed and divergent policy response; currently implemented policies maintained until 2030; thereafter, countries with net-zero targets achieve an 80 percent reduction only by 2050, while others continue current policies.
  - Current Policies (CP): only currently implemented policies; serves as the reference scenario.
- Emissions outcomes by 2040 (relative to CP):
  - Total emissions in Net Zero 2050 are 60% lower by 2040 compared to Current Policies.
  - Total emissions in Fragmented World are 30% lower by 2040 compared to Current Policies.
- Macro modeling assumptions:
  - Carbon price paths endogenously derived to meet NGFS emission and temperature targets.
  - Fiscal neutrality maintained: carbon tax revenues must be offset by increased government expenditures, reduced taxes elsewhere, or both.
  - Two revenue-recycling rules modeled:
    - Rule 1: all carbon tax revenues transferred to households.
    - Rule 2: half of revenues allocated to feed-in tariffs for renewable energy and half transferred to households (aims to emulate a component of Japan’s GX policy).
  - Required carbon tax rates are lower under Rule 2 than Rule 1.
- Macro outcomes by 2040:
  - NZ scenario exhibits slightly larger adverse effects compared to FW; Rule 2 mitigates some negative impacts.
  - Impact on employment remains relatively subdued.
  - CP scenario incorporates chronic physical risk: estimated reduction of GDP levels of 2.7 percent by 2040. NZ reduces that risk by around 25 percent (equivalent to a 0.7 percent higher GDP level in 2040 relative to CP). FW reduction of chronic physical risk is negligible.

### IMF-ENV model description and key assumptions (Box 2)
- IMF-ENV is a global recursive dynamic computable general equilibrium (CGE) model:
  - Captures structural change and sectoral impacts from energy and climate shocks/policies.
  - Allows simulation of energy demand/supply, GHG emissions, macro variables, sectoral outcomes, and trade.
  - Capital stocks have vintages; firms’ production and behavior differ in short and long run.
- Model structure and database:
  - Nested CES production functions; Armington specification for international trade.
  - Built on GTAP 10 Power database: country-specific I-O tables for 141 countries and 65 commodities.
  - Current version: 36 activities, 28 sectors, 26 country/regions (includes all G20 countries and 5 aggregated regions).
  - Electricity generation separated into 8 power sources: coal, natural gas, oil (diesel), hydro, nuclear, solar, wind and others (e.g., geothermal, biomass).
- Scenario-specific modeling notes:
  - For NZ, assumes progress in green technology, including increased electric vehicle penetration and carbon capture contributing approximately 18 percent of total emission reductions, amounting to 536 million tons.
  - Reduced GDP losses from chronic physical risk incorporated via adjustments in overall Total Factor Productivity growth.

### Sectoral macro impacts and transmission to micro layer
- Sectoral patterns by 2040 under NZ vs CP:
  - Emission-intensive sectors with large declines in GVA relative to CP: natural gas, petroleum and coal, chemical products, iron and steel, and air transport.
  - Electricity sector overall grows, but sub-sectors diverge: coal and gas power see reductions in GVA; renewable energy experiences growth.
  - Labor costs vary similarly to GVA but with less pronounced changes.
- Sectoral variables used to guide micro simulation: sectoral output, intermediate input, GVA, labor costs, and emissions.
  - Some sectors (e.g., land transport, business services) show positive output growth relative to CP but declining GVA due to larger increases in intermediate input.

### FIBA model layer — firm module: data sources and construction
- Primary microdata: Moody’s/Orbis balance sheets and income statements (2005-2023), panel of between 150,000 to 280,000 firms annually.
  - “No double counting” consolidation principle: firms included at highest consolidation level; lower-level subsidiaries excluded if parent company included.
- T0 database constructed to mitigate pandemic distortions:
  - Contains pre-pandemic data for flow variables (e.g., revenues) and latest data for stock variables (e.g., total assets).
  - Emission-intensive sectors represent 5.6 percent of total number of firms in sample, but account for 42 percent of total assets and 26 percent of total sales in the firm sample.
  - According to Japan Industrial Productivity database, these sectors comprise 13.4 percent of total GVA.
- PD (probability of default) data:
  - Moody’s KMV PD data for approximately 1,900 listed Japanese firms (2005–2020), converted to yearly averages.
  - PDs merged with Moody’s/Orbis dataset and used to estimate firm-fixed effects panel model; historical PDs for unlisted firms imputed from the model.
- Scope 1 emissions:
  - Firm-level reported emission intensities from Intercontinental Exchange (ICE), measured in ton CO2/USD million revenue, covering about 360 Japanese firms.
  - For firms without reported emissions, industry-average values for NACE Level 4 industries in the Asia-Pacific region from ICE were used.
  - Comparative exercise: industry-average ICE intensities correlated with 45-sector emission intensity data for Japan (correlation coefficient of 75 percent excluding electricity and gas sector).
- Sample coverage in micro layer:
  - Firm sample used in micro simulation encompasses approximately 271,000 nonfinancial firms (with emphasis on emission-intensive sectors).
  - Bank sample in bank solvency stress test includes 22 banks (one excluded for lacking emission-intensive portfolios).

### Firm P&L modeling and transmission of carbon costs
- Linkages between IMF-ENV sectoral outputs and firm-level variables:
  - Firms’ sales revenues modeled to move proportionally with sectoral output growth.
  - Cost of employees linked to growth of sectoral labor income (sectoral employment × equilibrium wage).
  - Operating expenses net of employee costs linked to sectoral intermediate input (output minus GVA).
  - Direct costs from a firm’s Scope 1 emissions incorporated into earnings:
    - Firm CO2 emissions calculated as emission intensity × sales revenues and assumed to follow the sectoral emission path.

### Key quantitative findings and modeling outputs
- Policy and investment commitments:
  - JPY 20 trillion upfront investment over the next decade for decarbonization.
  - Carbon levy planned from FY2028; emissions trading starts in FY2026 with phased allowance auctioning from FY2033.
- Sample and model scales:
  - Micro simulation layer: approximately 271,000 nonfinancial firms.
  - Bank module: 22 banks (one excluded).
  - ICE emission coverage: about 360 Japanese firms.
- Scenario impacts and numbers:
  - NZ: global net-zero CO2 emissions by around 2050; total emissions 60% lower by 2040 vs CP.
  - FW: delayed mitigation starting 2031; total emissions 30% lower by 2040 vs CP.
  - Chronic physical risk in CP: GDP levels reduced by 2.7 percent by 2040; NZ reduces this risk by around 25 percent (0.7 percent higher GDP level in 2040 vs CP).
  - Carbon capture contribution under NZ: approximately 18 percent of total emission reductions, amounting to 536 million tons.

*Source: IMF staff—chapter excerpt from 1jpnea2024003*

### 174.      Income tax expense flows during the simulation are computed based on net income

### 1jpnea2024003 - 174.      Income tax expense flows during the simulation are computed based on net income

### Income tax treatment in the simulation
- Income tax expense flows are computed based on net income before tax using industry-specific relevant rates.
- The tax rates are informed by the micro data themselves (Figure 77) and applied when earnings before tax for any given firm and per time period along the simulation horizon are positive.
- When firms’ pre-tax earnings are negative, they do not pay taxes.
- Tax credits are not considered.

### Channels through which carbon prices affect firm profitability
- Macro channel: carbon taxes induce changes in sectoral variables that map to firm-level variables (sector → firm transmission).
- Micro channel: carbon taxes directly affect firms’ profitability through direct emission costs incurred by firms.

### Uncertainty in firm-level emissions and Monte Carlo module
- Significant within-industry variation in firm emission intensities motivates explicit treatment of uncertainty (Capelle and others (2023) finding referenced).
- Residualized emission intensity is represented as the logarithm of CO2 emissions over revenues in megatons per million USD after controlling for industry-year fixed effects.
- In Japan, the 75th and 90th percentiles of “residualized” emission intensities are about 2 and 3 times greater than the median in the sample.
- A Monte Carlo simulation module is embedded based on the estimated Kernel density of firm emission intensity within an industry and the industry average emission intensity from ICE.
- Impact on individual firm risk metrics is assessed at the 50th, 75th, and 90th percentiles.

### Firm balance sheet dynamics and cash shortfall treatment
- Firms’ periodic post-tax profits are added to cash and cash equivalents and to total assets.
- If a firm’s cash holdings become negative, an amount equivalent to the shortfall is added to the firm’s short-term debt.
- Implicit assumptions: value of nonfinancial assets remains constant; outstanding debt is continuously rolled over.

### Probability of Default (PD) model specification and inputs
- PD model: firm-fixed effects panel econometric equation with logit-transformed firm-level PDs on the left-hand side.
- Logit transform notation and inverse sigmoid are used to guarantee predicted default rates in [0-1].
- Right-hand-side regressors reflect stock and flow features of Merton-style logic:
  - LEV: stock-stock leverage ratio = (short-term debt + 0.5 × long-term debt) / total assets.
  - ICR: interest coverage ratio = EBIT / interest expense.
  - EBITR: EBIT / assets ratio.
  - CDR: cash / short-term debt metric.
- Additional macrofinancial variables were not included because they feed structurally through the other regressors.

### PD estimation results (pre-pandemic sample)
- Estimation sample period: 2005-2019 for listed firms.
- Coefficient estimates (Table 11):
  - LEV: 3.534*** (standard error .0917)
  - ICR: -0.01*** (standard error 0.0002)
  - EBITR: -0.791*** (standard error .1568)
  - CDR: -0.016*** (standard error .0037)
  - _cons: -6.031*** (standard error .0327)
- Observations: 26,793
- Within R2: .173
- Notes: Standard errors in parentheses. *** p<.01, ** p<.05, * p<.1. ICR, EBITR, and CDR are winsorized at 5 (95) percent, while LEV is winsorized at one.
- Quantitative role: normalized coefficient for leverage (normalized by its standard deviation) is about twice as large as the normalized coefficients for all other regressors combined.

### Embedding PDs and TTC anchoring for unlisted firms
- Intercept adjustment implemented when embedding PD equation in the FIBA module to anchor model-implied firm PDs.
- For unlisted firms not in estimation sample, industry-averages of estimated firm fixed effects were used.
- Historical PDs and through-the-cycle (TTC) PDs (long-term averages of PiT PDs) computed for all unlisted firms.

### LGD, credit spreads, and aggregation to bank inputs
- LGD model component computes credit spreads endogenously and feeds into bank capitalization assessment.
- Credit spreads calculated at the firm level using a loan pricing formula, capturing the feedback loop between firms’ credit risk and cost of debt.
- Aggregate firm credit risk indicators computed (debt-weighted PDs, LGDs, credit spreads) and reaggregated to industry-level risk indicators for the bank module.
- PDs, LGDs, and credit spreads at firm industry levels are reaggregated using each firm’s debt as the weighting factor to align with bank loan portfolios.

### Heterogeneity in initial firm risk characteristics (Figure 78 summary)
- High initial leverage sectors: fishery (2), electricity (13), iron and steel (8), other manufacturing (18), business services (22).
- Low initial leverage sectors: gas (14), mining and quarrying (3), chemical products (6).
- Weak initial ICR sectors: electricity (13), water transport (15), paper products (4), fishery (2).
- Weak EBITR sectors: agriculture (1), water transport (15), paper products (4), collective services (21).
- Weak cash/short-term debt sectors: iron and steel (8), non-ferrous metals (9), paper products (4).
- Initial PDs vary substantially across and within sectors; emission-intensive sectors with relatively high initial PDs include non-metallic minerals (7), iron and steel (8), non-ferrous metals (9), fabricated metal products (10), electronic equipment (11), water transport (15). Non-emission-intensive sectors with elevated PDs include water supply (20) and services (21-22).

### Bank module inputs, loan portfolio construction, and dynamics
- Industry exposure data as of end-March 2023 sourced from banks for 22 banks, focused on emission-intensive and emission-input-reliant sectors (BOJ classification of Loans and Bills Discounted by Sector).
- Supplemented with banks’ publicly available financial reports to cover entire nonfinancial corporate sector; constructed 17 distinct portfolio segments used to compute scenario-conditional credit losses and interest income.
- Aggregation of PDs, LGDs, and credit spreads to these 17 portfolios uses firm debt weights.
- Loan portfolio composition and exposure to emission-intensive sectors vary across banks; internationally active banks have higher exposure to emission-intensive sectors.
- Average loan exposure to emission-intensive sectors in the sample: roughly 30 percent (compared to BOJ loan statistics share of 22 percent).

### Bank dynamics, credit loss accounting, and assumptions
- Dynamic balance sheet approach: accounts for deleveraging/leveraging of shrinking/thriving industries under scenarios; interest income evolves with industry performance.
- Rising credit spreads and LGDs are assumed priced into lending rates for vulnerable industries.
- Base lending rate: debt interest rate for Japanese corporate sector around about one percent in 2022; lending rate composed of base lending rate plus credit spread.
- General loan growth aligned with a rolling multi-year average of scenario-based sectoral GVA growth for each industry.
- Nonperforming loan (NPL) stock evolution for bank b’s exposure to sector n at horizon h:
  - NPL_n,b,h = NPL_n,b,h−1 × (1 − WROR − CURER) + NP_n,b,h × (L_n,b,h−1 − NPL_n,b,h−1)
  - Write-off rate (WROR) set to 100 percent and cure rate (CURER) set to zero percent for climate risk model application (interpretation: MKMV PDs closer to ultimate bankruptcy rates).
  - Performing exposures computed as residual of gross exposures (driven by g) and NPL stock.
- Model-consistent provision stocks for NPL portfolios:
  - PROV_n,b,h_NPN = LLPR_n,b,h × NPL_n,b,h
  - Loan loss flows: LL_n,b,h = PROV_n,b,h_NPN − PROV_n,b,h−1_NPN + WRITEOFF × LLPR_n,b,h × NPL_n,b,h−1
  - Loan loss flows change banks’ capital numerator period by period.

### Cross-bank heterogeneity and risk parameter anchoring
- PDs and LGDs at bank level for the NFC portfolio used as anchor points to reflect structural cross-bank risk differences.
- Regional banks in the sample tend to exhibit higher risk parameters (TTC PDs and LGDs) than internationally active banks (Figure 79).
- Alternative approach using PDs and LGDs at bank-portfolio level was partly tested and yielded quantitatively similar impacts on banks’ capitalization.
- NFC PDs and LGDs of each bank updated during simulation to align with firm-industry PD and LGD trajectories via micro firm debt-weighted aggregates and a logistic transformation/attachment scheme.

### Regulatory risk weights and pass-through assumptions
- All banks’ risk weights modeled following the IRB approach; recomputed using Basel risk weight formula for corporate segment with TTC PDs and downturn LGDs as inputs.
- Pass-through assumptions:
  - 100 percent pass-through from point-in-time PDs to regulatory TTC PDs (rationale: long-term horizon of climate risk analysis).
  - 20 percent pass-through assumed for the solvency stress test.
- Downturn LGDs held constant at initially observed levels.

### Results highlights and interpretation caveats
- Multiple heterogeneities shape outcomes: (1) heterogeneous sectoral impacts via emission intensity and inter-industry linkages; (2) initial risk characteristics across and within industries; (3) differential loan exposures across banks; (4) varying bank risk profiles.
- Focus of results on median (50th percentile) outcomes from the Monte Carlo simulations related to firm emission intensity uncertainty.
- Under policy scenarios versus CP scenario, PDs, LGDs, and credit spreads vary significantly across sectors:
  - Notable impacts in emission-intensive sectors such as paper products, iron and steel, fabricated metal industries (Figure 80).
  - Some emission-intensive sectors (chemical products, petroleum and coal) show only modest PD increases due to favorable initial conditions despite substantial sectoral GVA declines.
  - Non-emission-intensive sectors like other services and construction experience somewhat higher increases in PDs and LGDs.
- Monte Carlo findings on uncertainty:
  - No noticeable differences between 50th and 75th percentiles.
  - At the 90th percentile, large increases in both PDs and LGDs for certain emission-intensive sectors (e.g., mining and iron and steel).
  - This underscores sensitivity of firms’ credit risk to emission intensity and raises concerns for firms lacking emission data if their true emission intensity is substantially worse than reported.

*Source: IMF staff calculations as presented in the supplied content unit.*

### 195.      The overall impact on bank capital appears not too sizeable at the system level, while

### 1jpnea2024003 - 195.      The overall impact on bank capital appears not too sizeable at the system level, while

### Aggregate banking system capital impacts under climate scenarios
- Under the Net Zero 2050 (NZ) scenario, the aggregate capital ratio for the banking system is estimated to decrease by about 0.6-0.7 percentage points by 2040, translating to an annual decline of around 0.03-0.04 percentage points per annum when compared to the CP scenario.
- Under the FW scenario, the banking system’s aggregate capital ratio is projected to decline by 0.3 percentage points by 2040.
- In the FW scenario, mitigation efforts are designed to commence from 2031 onwards, yielding an estimated annual decline in capital ratios during the 2030s of about 0.03 percentage points per year, which is comparable to the average annual rate observed in the NZ scenario.

### Heterogeneity across banks and contributions to capital shifts
- Internationally active banks are modestly impacted because of the more favorable risk profiles of their borrowers.
- Regional banks in the sample experience the most pronounced effects.
- Non-emission intensive sectors, such as other services, account for approximately 30 percent of the shift in the banking system capital ratio.
- About one-third to one-half of the capital ratio shift can be attributed to changes in risk-weighted assets (RWAs), partly due to rising risk weights resulting from rising PDs and LGDs in emission-intensive sectors.
- The contribution analysis decomposes a shift in bank capital ratios into:
  - a change in interest incomes (∆VAI) relative to RWAs,
  - a change in loan losses (∆LL) relative to RWAs,
  - and a residual change related to RWAs (δRWACs).
- Significant uncertainty in firms’ emission intensity mainly affects bank capital ratios through direct emission-intensive sectors; outcomes show no notable differences between the 50th and 75th percentiles, but large differences at the 90th percentile driven by increased loan losses in emission-intensive segments.

### Dynamic balance sheets, industry shifts, and bank profitability
- Two industries examined in detail under Scenario 5 (Net Zero 2050 with Rule 2, 50th percentile of firms’ risk metrics) versus Scenario 1 (Current Policies):
  - Industry #6: Chemical (Shrinking in S5 relative to S1) — cumulative flows until 2040 (S5 minus S1) [JPY million]:
    - Static: Interest income 9,623; Loan loss -29,468; Sum -19,845
    - Dynamic: Interest income -218,213; Loan loss 21,322; Sum -196,890
  - Industry #14: Electricity and Gas (Growing in S5 relative to S1) — cumulative flows until 2040 (S5 minus S1) [JPY million]:
    - Static: Interest income -2,153; Loan loss 14,963; Sum 12,811
    - Dynamic: Interest income 598,422; Loan loss -348,908; Sum 249,514
- Under a dynamic balance sheet, the banking system’s capital ratio in the NZ scenario further declines by 0.1 percentage points compared to a static balance sheet where zero gross loan growth is assumed.
- More positive loan growth puts downward pressure on capital ratios; less positive or negative growth exerts upward pressure on capital ratios due to effects on RWAs.
- Some individual banks see improved capital ratios under a dynamic balance sheet compared to a static balance sheet because reduced loan losses and lower RWAs in shrinking industries may outweigh foregone interest income.
- Banks with riskier portfolios (e.g., higher through-the-cycle PDs) tend to perform better under a dynamic balance sheet relative to peers.

### Pricing credit risk versus adopting dynamic balance sheets
- Adopting a dynamic balance sheet leads to a modest decline in the aggregate capital ratio, approximately 0.05 percentage points.
- The increase in lending rates for industries experiencing an uptick in credit risk enables banks to generate a higher interest income relative to RWAs, by approximately 0.15 percentage points.
- The increase in interest income (≈0.15 percentage points) is three times as large as the impact of moving from a static to a dynamic balance sheet (≈0.05 percentage points).

### Caveats and model limitations
- The firm module lacks elements related to firm entry and exit dynamics.
- The credit risk assessment of borrower firms is based on an estimated PD model using the stock and flow-oriented Merton approach; certain sectors may demand sector-specific considerations for more accurate evaluation of credit risk during the transition.
- Additional layers of heterogeneity (e.g., quality of physical capital, intensity of knowledge in green technology) may affect results.
- More granular bank-industry specific risk parameters would be instrumental for refining the analysis using models of this kind.

### Climate physical risk overview and flood-focused analysis
- Japan’s overall physical risk score is categorized as “very high” based on Moody’s 427 dataset under no global mitigation policies over the next two decades.
- Individual hazards: risks associated with hurricanes and sea level rise are deemed exceedingly high; flood risk is classified as high.
- Flood-related disasters (storms, floods, landslides) account for over 70 percent of major natural disasters in Japan.
- Flood damages have been increasing in recent years, indicating growing vulnerability.
- Japan’s adaptive capacity is strong according to the Climate-driven INFORM Risk indicators; vulnerability remains relatively low compared to other countries due to flood control manuals, regular updates to “Expected Flood Inundation Area,” and innovative flood prevention measures (for example, a 6.3-kilometer underground discharge channel).

### Flood-damage estimation methodology and data
- Hazard projections:
  - Flood hazard dataset from MS&AD LaRC-Flood®Project with four hazard maps covering inundation depths for 100-year and 1000-year hazards for current climate and projected 2080 under SSP 585.
  - The dataset covers approximately 17.4 percent of Japan's total land area; resolution is 15 arcseconds (~500 square meters).
- Geolocational exposure:
  - A proxy dataset (normalized product of gridded nightlight intensity and gridded population density) from Eberenz and others (2020) used to infer distribution of physical assets; resolution 30 arcseconds (~1 square kilometer).
  - Prefecture-level data for private corporate capital stock and household net capital stock used to downscale national totals for 2021 exposure estimates.
- Vulnerability:
  - Damage functions by flood depth for buildings in Japan taken from Flood Control Economic Survey Manual (MLIT, 2020).
  - Adjustments: a depth of 0.45 meters for subfloor flooding based on Japan’s “Building Standard Law”; correction for number of stories in buildings (correction factor = 1 for up to two floors; for >2 floors factor = 2 divided by average number of floors).
- The analysis maps inundation depth and physical asset exposures to estimate damages and assess regional disparities in impacts across prefectures.

*Source: IMF staff calculations and the cited analysis in the referenced content unit.*

### 209.      Maximum flood damage values are typically considered in the literature reflecting

### Maximum flood damage values are typically considered in the literature reflecting

### Data on Physical Vulnerability
- Damage Rate by Flood Depth for Buildings (Damage rates (buildings)):
  - <=0.49m: 0.189
  - 0.50 ~ 0.99m: 0.253
  - 1.00 ~ 1.99m: 0.406
  - 2.00 ~ 2.99m: 0.592
  - >=3.00m: 0.800
- Maximum Damage Values (Structure, land-based) — Building Class / Maximum damage values (E/m2, 2010) / Uncertainty range:
  - Residential: 105.7 (76.1, 161.8)
  - Commercial: 164.7 (113.7, 222.4)
  - Industrial: 106.7 (74.7, 163.2)
- Notes:
  - The lower end of the 90 percent confidence interval from the proposed maximum damage values were selected, given Japan’s high capacity for climate adaptation.
  - Values were adjusted for price changes between 2010 and 2020, and for currency conversion.
  - In 2020, the resulting maximum damage value for residential buildings is JPY 2,275 million per 500m2, while for corporate buildings, it is JPY 2,814 million per 500m2.

### Results: Asset Exposure, Climate Projections, and Property Price Impacts
- Exposure and baseline findings:
  - Approximately one-third of the physical assets in Japan are potentially exposed to flooding risk.
  - Roughly 60 percent of the country’s physical assets fall outside the scope of flood risk assessment mainly due to locations such as mountainous regions.
  - Around 4 to 12 percent of physical assets are subject to inundation depths of less than 45cm and are not assessed prone to flood risk given the implementation of Japan’s “Building Standard Law.”
  - In certain prefectures, more than half of the physical assets are exposed to flood risk.
- Climate change projections (SSP 585 scenario):
  - The flood depths for a 100-year and 1000-year hazards are projected to increase, on average, by about 0.5-0.6 meters by 2080, when compared to current hazard maps.
  - More than 25 percent of the areas exposed to flood risk is anticipated to experience inundation depths exceeding 0.8 meters for 1000-year flood events.
  - Assets exposed to flood depths greater than 3.45 meters are projected to rise.
  - Geographical concentration: certain prefectures, particularly those in the Tohoku region, are poised to experience a relatively significant increase in inundation depth.
- Property price impacts:
  - Koide and others (2022) find that a one-meter rise in inundation depth is associated with a 1.1 percent decrease in residential land prices and a 4.8 percent decline in commercial land prices.
  - The projected increase in inundation depth could imply a decline in commercial land prices by, on average, about 3 percent.
  - Prices may drop by more than four percent and ten percent, respectively, for about 25 percent and 5 percent of the commercial land exposed to flood risk.
  - Several prefectures may experience an average decline in total commercial land values exceeding 5 percent for the areas exposed to flood risk.
- Flood damage rates:
  - Expected damage rates in 2080 surpass the present rates by, on average, about 20 percent.
  - In absolute terms, the average damage rate for 1000-year floods rises from 28 percent to 32 percent by 4 percentage points.
  - For 100-year floods, the damage rate, on average, increases by 5 percentage points.
  - Distributional effects:
    - For 100-year floods, the level of the 75th percentile rises, suggesting areas already exposed to significant flood risk may face more severe disasters.
    - For 1000-year floods, the level of the bottom 25th percentile moves up, indicating wider areas could be at risk of severe flood damage.
  - Geographical vulnerability: several prefectures, particularly in the Tohoku and Kanto regions, exhibit heightened vulnerability.

### Financial Sector Implications and Prefecture-Level Correlations
- Prefecture-level regression:
  - There is a negative correlation between average damage rates and the ratios of bank loans to the total private assets of firms and households.
  - Specifically, a 5 percentage points increase in damage rates is associated with a 1.5-3 percentage points lower overall loan-to-asset ratios.
  - Interpretation caveat: data shows a disproportionately high share of loans allocated to Tokyo (nearly half of the loan share) versus Tokyo’s ~20 percent share in overall GDP, which may lead to underestimation of actual exposure to physical risks.

### Holistic Vulnerability Assessment of the Japanese Banking Sector — Methodology and Results
- Methodology:
  - Banks were ranked across multiple vulnerability dimensions: initial balance sheet characteristics (2023Q1/FY2022), solvency risk, liquidity risk, interconnectedness, and climate transition risk.
  - A composite weighted-average ranking was computed using 22 individual indicators with judgmentally set weights; substantially more weight assigned to forward-looking metrics based on solvency, liquidity, and contagion analyses.
- Results:
  - The top three most vulnerable banks based on the composite indicator represent five percent of total assets of the sample of the 23 banks.
  - These include two internationally active banks and one domestic bank; no SIBs are among the top three.
  - Cluster patterns:
    - Internationally active banks are, on average, more vulnerable than domestic banks.
    - Regional banks are on average the most vulnerable.
    - Domestic banks have composite rankings above the banking system average but are below the average for forward-looking solvency and liquidity dimensions.
  - Decomposition:
    - Internationally active banks are overall more vulnerable due to greater interconnectedness.
    - Domestic banks appear more vulnerable in solvency and liquidity stress test results.

### Recommendations for Systemic Risk Analysis and Supervisory Practices
- Strengthen stress testing and model frameworks:
  - The BOJ should continue enhancing its macro stress test model, e.g., by developing its credit risk module further.
  - Liquidity risk analysis for banks should be enhanced with stress testing tools and models for exposures in both JPY and FX and made an integral part of systemic risk assessment.
- Broaden systemic risk coverage:
  - Broaden systemic risk analysis to cover stress testing of investment funds; FSA should enhance data collection and quality assurance for supervisory reporting by investment funds.
  - For insurers, consider ESR top-down analysis across combinations of market liquidity conditions and macroeconomic shocks, and a more comprehensive liquidity risk analysis; supervisory reporting could be expanded to include granular derivative positions.
  - Formally assess potential contagion risks (including NBFIs) as part of systemic risk monitoring.
- Real estate and climate analyses:
  - Continue assessing vulnerabilities in real estate markets using granular information on real estate loans and borrower characteristics; consider formal modeling to assess price misalignments.
  - Authorities should continue to develop analytical capacity to analyze climate-related risks, e.g., by accounting for dynamic balance sheets; BOJ’s recent top-down scenario analysis of transition risk is noted as commendable.
- Interagency collaboration and data enhancements:
  - FSA and BOJ could explore interagency collaboration with access to rich, granular data for systemic risk analyses.
  - Interagency collaboration should be strengthened for climate risk-related analysis, collaborating with other government agencies and research institutions to leverage expertise in climate and natural disaster modeling.
- Data and model improvements for flood risk analysis:
  - Obtain geolocational exposures of individual financial institutions or reliable estimates to evaluate potential impacts on financial resilience.
  - Refine analysis with more granular data: higher-resolution flood hazard maps, municipality-level data on average number of stories in buildings.
  - Access to hazard maps for various return periods would permit estimation of expected aggregate damage rates across Japan.
  - Use natural disaster models together with localized climate scenarios to probabilistically evaluate flood risk across Japan.

*Sources: Flood Control Economic Survey Manual (MLIT, 2020); Huizinga and others (2017); and IMF staff calculations.*

### 221.      To broaden and deepen systemic risk analysis, further improvements with regard to

### 1jpnea2024003 - 221.      To broaden and deepen systemic risk analysis, further improvements with regard to

### Data-collection initiatives and recent progress
- Common Data Platform: jointly run by the FSA and the BOJ; aims to collect transaction-level loan data from a large number of banks; the platform has been operational since FY2023. (Reference note 97)
- Surveys and councils: progress via surveys (including “Investment Fund Survey”) and the newly established Financial Monitoring Council (FMC). (Reference notes 97, 98)
- Recent climate scenario work: The FSA and BOJ conducted the first bottom-up scenario analysis in August 2022, covering both transition and physical risks. BOJ researchers published a top-down scenario analysis of transition risk (Abe and others, 2023). (Reference note 96)

### Identified data gaps and recommended scope enhancements
- Enrich the scope of the Common Data Platform and integrate related databases for more comprehensive risk assessment.
- Collect granular data on loan exposures to:
  - enhance credit risk analysis and modeling as input for bank solvency stress testing; and
  - model and assess real estate-related vulnerabilities and support policy making, especially once integrated with detailed borrower-level information.
- Strengthen the FSA’s collection of data on NBFIs and ensure the scope of the “Investment Fund Survey” allows for comprehensive risk assessment.
  - The Investment Fund Survey is a welcome step but: it is in pilot phase, does not cover all funds, and currently provides limited information necessary for comprehensive fund-level stress tests. (Reference note 99)
- Close remaining gaps related to climate data, including enhanced disclosure mechanisms to gather granular emissions data across a wide spectrum of firms.

### Climate-related data and geospatial exposure needs
- Enhanced disclosure mechanisms should be implemented to gather granular data that provides insights into the emissions of a wide spectrum of firms.
- To evaluate implications of physical climate risks for individual financial institutions, compile geospatial exposure data that:
  - includes not only the location of a firm's headquarters but
  - also considers the location of investments through which the financial institution would be exposed to physical risk.

### Institutional coordination and burden reduction
- The Financial Monitoring Council (FMC), comprising senior officials from the FSA and BOJ, aims to improve coordination in monitoring risks and streamline/combine data templates to reduce burden on financial institutions. (Reference note 98)

*Source: 1jpnea2024003 - 221.*

### Appendix I. Table 7. Scenario Profile for Other Jurisdictions

### Appendix I. Table 7. Scenario Profile for Other Jurisdictions

### Table values (as presented)
- Source: IMF staff.
- United States: 100101.5103.3105.598.498.3100.7
- China: 100104.2108.4112.998.8100.9103.2
- United Kingdom: 100100.6102.6104.795.294.396.1
- Germany: 100100.9102.9104.996.896.398.1
- France: 100101.3103.2104.997.997.899.7
- Spain: 100101.7103.8105.796.596.7100.1

### Additional numeric blocks (as presented)
- United States: 4.95.03.72.66.44.52.7
- China: 0.40.40.40.41.00.50.3
- United Kingdom: 5.65.85.04.67.15.74.1
- Germany: 3.73.73.42.96.65.73.6
- France: 3.32.92.52.56.34.93.0
- Spain: 3.53.33.03.16.55.33.6

### Further numeric blocks (as presented)
- United States: 3.64.03.83.76.15.33.6
- China: 2.72.62.72.76.65.42.7
- United Kingdom: 4.04.24.34.46.55.43.3
- Germany: 2.42.62.72.75.74.82.4
- France: 3.03.33.53.66.35.53.2
- Spain: 3.43.94.03.87.66.23.4

### Table context labels appearing in the source
- Long-Term Interest Rates
- 2023
- Baseline
- Adverse
- Variables
- Countries
- Real GDP
- Short-Term Interest Rates

*Source: IMF staff.*

### Appendix II. Real Estate Market Analysis

### Appendix II. Real Estate Market Analysis

### Construction of the Price Acceleration Indicators
- Methodology:
  - Detection of explosive behavior in real estate prices uses Augmented Dickey-Fuller (ADF) tests.
  - Generic ADF regression (notation preserved):  
    Δyt = C r1,r2 + β r1,r2 yt−1 + Σ_{j=1}^{k} ψ r1,r2 j Δyt−j + ε t, ε ~ N(0, σ r1,r2 2)
  - yt denotes a generic time series (log house price or price-to-rent ratio); Δyt for j=1,...,k are differenced lags; εt is the error term.
  - C1 and C2 denote fractions of the total sample specifying starting and ending points of a subsample period; k is the maximum number of lags.
  - Null hypothesis under interest: unit root in yt corresponding to β r1,r2 = 0. Test statistic: ADF r1 r2 = β r1,r2 / sSC (standard statistic ADF0 1 with C1=0, C2=1).
- Recursive procedure to improve detection in boom-bust dynamics:
  - SADF(r0) = sup_{r2 ∈ [r0,1]} CP F0 r2, with limit distribution under H0: sup_{r2 ∈ [r0,1]} ∫ W L W_{r2 0} (∫ W^2_{r2 0})^{1/2}, where W is a Wiener process and window size C̅ = C2 − C1.
  - Rejection requires SADF statistic exceed right-tailed critical value from limit distribution.
- Date-stamping exuberance (allowing start and end to change):
  - BSADF(r0) = sup_{r1 ∈ [0, r2 − r0]} PCPF_{r1 r2}
  - Origination date re* = inf_{r2 ∈ [r0,1]} {C2 : PCPF_{r2}(C0) > sS C[r2 T]_α}, where sS C[...]_α is the 100(1-α) percent critical value of the SADF on [C2 T] observations and α is chosen significance level.
  - Termination date is first observation after re for which BSADF falls below its critical value.

### Housing and CRE Price Overvaluation
- Two alternative approaches to estimate misalignment (overvaluation):
  1. Error Correction Model (ECM; “benchmark model”):
     - Levels regression: yt = β1 + β2 Xt + C t
     - Engel two-step procedure:
       - First-stage OLS to estimate β1, β2; obtain residuals.
       - Second-stage OLS with first difference:  
         Δyt = β3 + β4 ΔXt + α(yt−1 − β1 − β2̂ m t−1) + εt
       - α is coefficient on estimated lagged residual; β2̂ is long-run coefficient; β4̂ is short-run coefficient.
     - Order of integration tested via augmented ADF test; cointegration required for consistency.
     - Explanatory variables capture mainly demand-side factors; supply assumed relatively inelastic in short run but affects long run.
  2. Alternative OLS with affordability:
     - Δyt = β0 + β1 ΔXt + θ C_{AFFORDANCE}_{t−1} + εt (affordability acts as anchor; affordability measured as price-to-income ratio or deviation from historical trend in ECM specification).
     - Affordability used as benchmark/error correction mechanism reflecting long-run equilibrium determined by fundamentals.
- Covariates included in both approaches:
  - Changes in income per capita
  - Short- and long-term interest rates
  - Credit growth
  - Growth in equity prices
  - Growth in fraction of working-age population
  - Affordability measure based on price-to-income ratios
  - Construction costs as supply-side proxy
- Base levels and misalignment computation:
  - Levels of house prices in years from 1997 to 2001 used as alternative base levels.
  - Estimated misalignment calculated as average over these base years.
  - For CRE, model adds rental growth, vacancy rate, and capital inflows; affordability proxy is misalignment of capitalization rates.
- Empirical outputs referenced:
  - Results from the two models for house and CRE prices provided in Figure I.1 (overvaluation measures; estimated range derives from different base price levels for years 2000-15; average misalignment is mean over different base years).

### Measuring Downside Risks to Real Estate Prices (House-Price-at-Risk and CRE at Risk)
- Purpose:
  - Determine likelihood of price correction using a house price at-risk model based on Adrian and others (2023).
  - House price-at-risk corresponds to downside risks defined as forecast house price growth at the 5th percentile of the house price distribution.
  - Relevance: helps forecast downside risks to GDP growth beyond simpler measures of house price imbalances; adds to early-warning models for financial crises.
- Estimation approach:
  - Two-step panel quantile regression following Canay (2011).
  - Step 1: fixed effect estimation of (log annualized) changes in real house prices and construction of demeaned dependent variable: Δ^h Ȳ_{L,t+h,τ} ≡ Δ^h Y_{L,t+h,τ} − α̂_{L,h,τ}.
  - Step 2: quantile regression for each quantile q and horizon h; quantile function obtained as solution to optimization problem minimizing quantile loss function ρ_τ.
  - House-Price-at-Risk defined as HaR_{i,h}(1−τ | X_t) where Pr(Δ^h Ȳ_{L,t+h,τ} ≤ HaR_{i,h}(1−τ | X_t)).
- Main control variables for house price-at-risk:
  - Past growth in house prices
  - Financial conditions
  - Real GDP growth
  - Credit boom indicator
  - Overvaluation indicator (degree of deviation from fundamentals)
- CRE price-at-risk baseline specification:
  - Similar approach; controls include past growth in CRE prices, CRE price misalignment, GDP growth, credit-to-GDP growth, capital-flow-to-GDP ratio, monetary aggregates, and vacancy rates.
  - Specification based on Deghi, Mok and Tsuruga (2021).
- Empirical outputs referenced:
  - Figure II.2 shows probability densities for one-year-ahead and three-year-ahead house and CRE price growth distributions (cumulative).
  - Notes: forecast density estimates assume the GDP growth path projected by the WEO; filled circles indicate the price decline with a 5 percent probability (5th percentile) in an adverse scenario.
  - Horizons explicitly referenced: One-Year Ahead and Three-Years Ahead for both house and CRE price-at-risk models.

### Nonfinancial Corporate Risk Analysis — Data and Methodology
- Data sources and coverage:
  - Moody’s/Orbis database: historical balance sheet and income statement data for nonfinancial firms.
  - Sample excludes entities in finance and insurance (NACE Rev. 2, 64-66), public administration (84), and activities of households (97-98).
  - Dataset spans an 18-year time frame, 2005 to 2022.
- Consolidation and “no double counting” principle:
  - Firms included at highest consolidation level; lower-level subsidiaries excluded if parents were Japanese firms included in sample.
  - Relevant consolidation codes: C2, C1, U1, U2. Preference order: C2-C1-U1. U2 not used when consolidated results available.
  - “Filing type” used to remove remaining duplicates per consolidation code per firm and year.
- Filters applied to clean dataset (variables and conditions preserved):
  - Cash_and_cash_equivalent (>0)
  - Total_assets (>0)
  - Debt_holding* (>0)  [Note: Debt_holding = Loans + Creditors + Long-term_debt]
  - Long_term_debt** (>0)
  - Costs_of_goods_sold** (≥0)
  - Costs_of_employees** (≥0)
  - Operating_revenue_turnover (>0)
  - Operating_P_L_EBIT (≠0)
  - P_L_before_tax (≠0)
  - P_L_after_tax (≠0)
  - Sources: Moody’s; and Orbis. Notes: ** filtering applied only to climate risk analysis.
- Moody’s/KMV PDs:
  - Daily one-year PDs converted to annual by taking annual averages per firm and aligned with closing date of firm financial statements; PD averages merged with Moody’s/Orbis dataset by firm IDs.
- PD estimation and imputation:
  - Firm fixed effects econometric model (equation 1) used to estimate determinants of PD; dependent variable = logit transformation of 1-year PD.
  - Independent variables include leverage, interest coverage ratio (ICR), earnings before interest and taxes over total asset ratio (EBITR), cash over short-term debt ratio (CDR), and real GDP growth.
  - Estimated model used to impute PDs for unlisted firms (without Moody’s/KMV PDs).
- Out-of-sample projections and stress:
  - Orbis does not cover 2023; firm performance for 2023 estimated based on macro variables and industry-level trends.
  - Firm performance simulated in 2024-2026 under baseline and various stress scenarios used for bank stress test.
  - Key variables forecasted: revenues, cost of goods sold, other operating expense, financial revenue, interest paid, tax expense.
  - Assumptions:
    - Revenues grow in line with nominal GDP.
    - Cost of goods sold has constant elasticity to sales (elasticity estimated historically).
    - Other operating expense grows with inflation.
    - Financial income affected by financial market performance and sales revenue.
    - Tax rate assumed constant (zero if earnings negative).
    - Interest rates are firm-specific and endogenous to PD and LGD due to credit spread.
- LGD modeling:
  - Vašíček model structure per Frye and Jacobs (2012):  
    LGD_{t0+h} = Φ(Φ^{-1}(PD_{t0+h}) − k) / PD_{t0+h}; where k = (Φ^{-1}(PD) − Φ^{-1}(PD × LGD)) / sqrt(1 − ρ)
  - LGD trajectory linked to PD trajectory at firm level.
  - Aggregate nonfinancial corporate sector PD (or LGD) in 2024-2026 obtained by average of PD (or LGD) weighted by each firm’s debt level.
- Climate risk analysis:
  - Constructed a “T0” database to anchor microsimulation and mitigate pandemic distortions:
    - Pre-pandemic data up to March 2020 used for flow variables (e.g., revenues).
    - Most recent data used for stock variables (e.g., total assets).
    - For flow variables, averages over up to three years considered when available.
  - “T0” database retained only active firms; firms that defaulted, merged, or were resolved were removed.
  - Missing value imputation:
    - Cost of employees: industry median wage at NACE Level 2 × firm's number of employees.
    - Interest expenses: industry median interest rate × firm's debt holdings.

### Household Stress Testing — Data
- Microdata source:
  - Japan Household Panel Survey (JHPS/KHPS) compiled by University of Keio (KHPS since 2004; JHPS since 2009).
  - Survey includes household- and household-member level data: age, education, marital and employment status, labor income, unemployment benefits; household-level financial assets, liabilities, income/expense flows, mortgage loan details (type, maturity, loan amount, house/plot value at origination).
  - Survey compared to PSID (U.S.), ECHP (Europe), and HFCS (Europe).
- Wave and timing:
  - Latest available wave used: January 2021; provides information for the year 2020 or end-2020 depending on question.
  - Several income and expenditure variables extrapolated to 2022 or end-2022 using macroeconomic variables for future projections.
- Variables:
  - Full list of variables used provided in Table IV (referenced in source).

*Italic: Source: 1jpnea2024003 - Appendix II. Real Estate Market Analysis (PDF chapter).*

### Appendix IV. Table 1 . Japan: Survey-based variables used in the Household Stress Testing

### Appendix IV. Table 1 . Japan: Survey-based variables used in the Household Stress Testing

### Survey variables and coverage
- Primary data sources: JPHS/ KHPS; and IMF staff calculations.
- Survey scope:
  - # households: 4,817
  - # household members: 14,318
  - One-fourth of households hold mortgage debt.
  - 38 percent hold either mortgage or consumer debt.
- Key survey variables (as listed in the appendix):
  - Assets: Current value of house (IN), Current market value of bonds (B), Current market value of stocks (S), Total financial assets (TFA)
  - Liabilities/debt: Outstanding balance of mortgage debt (D_M), Outstanding balance of non-mortgage debt (D_NM), Total Debt (D)
  - Income flows: Household income total, quarterly, gross of tax (I); Household income total, quarterly, net of tax (I_2); Labor income (gross of tax) from employment or self-employment, public/private pension income (net of tax), quarterly (C_E); Rental income, quarterly (RIR); Interest Income, quarterly (II); Other regular income, quarterly (OI); Unemployment benefit, net of tax, quarterly (INC_U)
  - Expense flows: Total expense, quarterly (AT); Rental expense, quarterly (OER); Living expense, excl. annuities and rent, quarterly (E); Living expense excluding rent - other expenses (E_2)
  - Household/member identifiers and demographics: Household ID (HH_ID), Household member ID (HM_ID), Household members' household IDs (HM_HH_map), Household member level (HM Level), Age (AGE), Marital status (MAR), Level of education (EDU), Gender (GEN)
  - Mortgage specifics: Year of 1st mortgage origination (Myear); Duration of 1st mortgage at origination in years (MiniDur); Rate type of total debt (DType); Annuity for mortgage debt, quarterly (A_M); Annuity for non-mortgage debt, quarterly (A_NM); Living expenses (excl. Annuities and rent) as share of gross income (Etol)
- The survey data matches well the 2020 Census on geography, gender, age, marital status, and aggregate unemployment rate.

### Summary statistics and cross-country comparisons
- Loan-to-value and debt service statistics (median household):
  - Median initial LTV (mortgages): 90 percent
  - Median current LTV (mortgages): 69 percent
  - Median current DSTI (total debt): 17 percent
- Comparative notes:
  - Japan figures (JHPS&KHPS) are lower than U.S. or European averages in the share of households with debt, but LTV at origination/current appear higher compared with peers.
- Loan-to-income (LTI) evolution:
  - Current LTI: increased from below 2.3 in 2016 to 2.6 in 2021.
  - LTI at origination (or related measure): increased from below 4 percent in 2015 to 4.3 in 2020.
  - Increase in LTI mainly driven by younger age groups, which constitute close to 30 percent of outstanding loans.
  - Share of floating rate mortgages for the young age group: 80 percent.

### Extrapolation of household-level data to 2022 (aggregate macro inputs)
- Relative changes between 2020 and 2022 (Relative to 2020):
  - House price: 6% (2021), 15% (2022)
  - Household holding government bonds: -5% (2021), -6% (2022)
  - Stock price: 27% (2021), 20% (2022)
  - Household debt to income: 5% (2021), 3% (2022)
  - Gross disposable income: -3% (2021), -1% (2022)
  - CPI (rent): 0.1% (2021), 0.1% (2022)
  - CPI: -0.2% (2021), 2.3% (2022)
  - Household consumption: 1% (2021), 6% (2022)
  - Nominal Wage Income of Employees: 1% (2021), 2% (2022)
  - Households: Assets: Currency & Deposits: 3% (2021), 6% (2022)
- Narrative summary from text:
  - Between 2020 and 2022: aggregate house prices (including plots) increased by 15 percent, aggregate stock prices by 20 percent, overall consumer price index by 2.3 percent, household debt-to-income by 3 percent, household consumption by 6 percent, nominal wages by 2 percent, household assets (currency and deposits) by 6 percent. Household holdings of government debt declined by 6 percent, gross disposable income declined by 1 percent, and rent prices remained broadly stable.

### Modeling framework and simulation steps
- Model used: IDHBS+ (Integrated Dynamic Household Balance Sheet+).
- Main simulation steps:
  - Build macrofinancial scenarios (baseline and adverse) taken from bank solvency stress testing.
  - Simulate employment status at household member level (excluding retirees and students) using logistic regression linked to characteristics (age, education, marital status, gender). The regression:
    - Suggests individuals with 4-year university degrees, or who are male or married are more likely to be employed; age appears not significant.
    - AUROC suggests the model can predict about 72 out of 100 cases correctly.
  - Employment Status Simulator uses logistic coefficients, dynamically updates age, adjusts intercept to match aggregate unemployment rate, and assumes persistence of the error term to match aggregate unemployment duration (3 quarters, the OECD).
  - Simulate household balance sheets using employment outcomes and evolution of assets, income, consumption, debt repayments.
- Financial assets evolution and debt service mechanics:
  - Households earn interest on deposits and certain incomes, receive employment income or unemployment benefits, and gain/lose from bond/stock revaluations.
  - Debt service: constant payment for fixed-rate loans (or until maturity), and moves with short-term interest rates for floating-rate debt. Interest and principal payment flows evolve non-linearly with principal share rising over time.
  - Stock and bond holdings re-valued based on aggregate stock market return and risk-free interest rate path assuming average bond duration.
  - Interest income on deposits linked to 3-month money market rate.
- Default and loss mechanisms:
  - Default occurs when financial assets become negative (full-recourse system).
  - Nominal consumption assumed to move with inflation path via a cointegrating relationship.
  - Loss-given-default (LGD) for mortgages computed via discounted expected recovery value of the underlying house, with a cure rate of 5 percent.

### Key calibration parameters (2022)
- Income tax rate: 21.6 percent.
- Unemployment benefit replacement rates (relative to previous gross-of-tax income):
  - 52.5 percent in the first year,
  - 31.5 percent in the second year,
  - 18.9 percent in the third year of unemployment.
- Unemployment duration: 3 quarters.
- Deposit rate: 0.1 percent.
- External anchors for mortgage portfolio parameters (from bank solvency data templates):
  - PD anchor: 0.3 percent.
  - LGD anchor: 25 percent.

### Logistic regression results (employment probability)
- Dependent variable: Prob(Employed)
- Coefficients (with p-values in parentheses):
  - Marital Status (Single = 1): -0.28 (0.08)
  - Education (No Uni. Degree = 1): -0.626 (0.00)
  - Gender (Male = 1): 1.642 (0.00)
  - Age: -0.002 (0.64)
  - Intercept: 3.074 (0.00)
- AUROC: model predicts about 72 out of 100 cases correctly.

### Simulation grid and sensitivity analysis for mortgage PDs
- Grid boundaries (5 equally spaced points) used to cover historical ranges and adverse scenarios:
  - Unemployment rate: from 1 to 6 percent
  - Interest rates: from 0 to 8 percent
  - Compensation per employee growth: from -6 to 8 percent
  - Stock price growth: from -50 to 60 percent
  - Nominal consumption growth: from -1.5 to 6 percent
- Polynomial analysis findings for mortgage PDs:
  - Higher unemployment rate, higher interest rates, lower nominal wage growth, lower stock price growth, and higher consumption growth imply higher mortgage PDs.
  - Mortgage PDs are more sensitive to changes in unemployment, interest rates, and consumption, and notably less sensitive to stock price growth given low household investment in stocks.
- Mortgage PD polynomial estimates (coefficients and model fit):
  - Intercept: -6.71
  - Unemployment rate (URX): 20.79
  - Unemployment rate (URX) ^2: 91.46
  - Short-term interest rate (STN): 94.4
  - Short-term interest rate (STN) ^2: -590
  - Compensation per employee growth: -16.32
  - Compensation per employee growth ^2: 3.61
  - Stock price growth: -0.13
  - Stock price growth ^2: 0.11
  - Consumption growth: 18.54
  - Consumption growth ^2: 84.28
  - R^2 (reported in table): 0.96 (and other R^2 entries shown: 0.39, 0.86, 0.93, 0.95 for subplots)

### Drivers of mortgage LGDs
- Mortgage LGDs are tightly linked to house price growth, cure rates, and interest rates.
  - Decline in house price growth or cure rates, or an increase in interest rates, implies higher mortgage LGDs.
- Mortgage LGD polynomial estimates (coefficients and fit):
  - Intercept: -0.69
  - House price growth: -4.11
  - House price growth ^2: -2.22
  - Cure rate: -1.54
  - Cure rate ^2: -0.81
  - Short-term interest rate (STN): 7.38
  - Short-term interest rate (STN) ^2: -11.73
  - R^2 values shown for these regressions: 1 / 1 / 1 (as reported in table layout).

*Sources: JPHS/ KHPS; and IMF staff calculations.*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2024/english/1jpnea2024003.pdf_