## wpiea2024118-print-pdf

## Source details

**Canonical URL:** [wpiea2024118-print-pdf](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024118-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024118-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024118-print-pdf.pdf.json)

---

### Scope and purpose
- Study period: 2000-2022.
- Objective: Examine how demographic aging influences the stability of banking systems across advanced economies.
- Data basis: Demographic projections by the United Nations to 2100 and bank-level data for advanced economies (2000-2022).

### Demographic context (key numeric projections and facts)
- Over two-thirds of the world's population live in nations where fertility rates fall beneath the replacement level of 2.1 births per woman.
- Global median age: currently 30 years; projected to be 36 in 2050 and 42 by 2100.
- Average life expectancy: around 74 in 2022; projected to reach 77 by 2050 and 82 by 2100.
- Share of elderly (above 65): projected to increase from 10 percent at present to 24 percent by 2100.
- World population trajectory: projected to peak by 2090 and then decline (figure text also indicates a peak in 2087 at 10.4 billion).

### Conceptual transmission channels and theoretical mechanisms
- Aging reduces the role of banks in maturity transformation because:
  - Older households and firms demand less credit.
  - Many upcoming pensioners have paid off mortgages, reducing credit sensitivity to interest rates.
  - Older households shift to safer assets, strengthening links between households and the sovereign.
- Banks respond by recalibrating business models, including:
  - Reducing loan offerings.
  - Increasing investments in government bonds and other securities.
  - Diversifying into new financial activities.
  - Expanding internationally.
- Key theoretical mechanisms and illustrative facts:
  - Life-cycle hypothesis (Modigliani, 1975) and the "Wealth Decumulation Puzzle" (Horioka and Ventura, 2022) suggest deposits may remain elevated among retirees due to bequest motives, public pensions, and reluctance to liquidate real estate.
  - A 10 percent rise in the share of individuals aged 60 and above correlates with a 5.5 percent drop in per capita GDP growth (GFSR, 2017).
  - Aging amplifies preference for secure, long-term investments, flattening the yield curve and compressing bank profitability.
  - Behavioral/cohort effects can permanently alter risk preferences (Modigliani, 1986; Jianakoplos and Bernasek, 1998; Mather, 2006; Hasegawa, 2011; Malmendier and Nagel, 2011; Sunde, 2023).
  - Consequent bank strategic adaptations can raise tail risks (consolidation, expansion into unfamiliar products/markets, increased sovereign exposure).

### Empirical approach and data
- Sample: 28 advanced economies, years 2000–2022.
- Econometric approach: fixed effects panel regressions with country and time fixed effects; post-GFC dummy with cut-off in 2010.
- Baseline specifications:
  - 푦푖,푡 = 훼 + 훽1 퐴푔푖푛푔푖,푡 + 퐷푢푚푚푦푖 + 퐷푢푚푚푦푡 + 퐷푢푚푚푦푝표푠푡퐺퐹퐶 + 푒푖,푡
  - 푦푖,푡 = 훼 + 훽1 퐴푔푖푛푔푖,푡 + 훽푚1−3 푀푎푐푟표푖,푡 + 훽푏1−2 퐵푎푛푘푖,푡 + 퐷푢푚푚푦푖 + 퐷푢푚푚푦푡 + 퐷푢푚푚푦푝표푠푡퐺퐹퐶 + 푒푖,푡
- Dependent variables (bank balance-sheet and stability metrics):
  - Deposits to total funding; Loans to assets; Deposits to loans deposits.
  - Risk weight density; Return on equity; Credit loss rates.
  - Z-Scores (World Bank); Price-to-book ratios; Expected Default Frequency (EDFs) (Moody’s); SRisk (NYU).
- Controls: real GDP growth, general economic uncertainty, policy rates, bank credit to GDP, bank leverage, profitability; all lagged by 1 year.
- Aging operationalization:
  - Aging Metric 1: ratio of active population (aged 15-64) to elderly population (aged 65+).
  - Aging Metric 2: population growth compared to a decade earlier and year-on-year growth rates.
  - "Old" countries (Metric 1): Finland, France, Greece, Italy, and Japan.
  - "Old" countries (Metric 2): Finland, Greece, Italy, Japan, and Spain.
  - Aging dummies for 4-year intervals—2010, 2014, and 2018—used to capture progressive intensification of aging.
- Projections rely on UN population forecasts (2022); alternative forecasts (IIASA, IHME) suggest slightly stronger aging and earlier population decline.

### Main empirical findings — assets and liabilities (Section IV.A and Appendix 3)
- Descriptive patterns:
  - The “oldest” countries show a differential reduction in demand for loans, particularly over the past five years.
  - Share of deposits in total funding has been marginally lower in these older countries but remains on an upward trend.
  - Excess deposits relative to loans observed across aging metrics for the eldest countries.
  - Influence of aging on balance-sheet adjustments is more pronounced with Aging Metric 1 than Metric 2.
- Regression evidence (selected coefficients and averages preserved as reported):
  - Reported coefficient b1 (aging effect on loan-to-asset ratios) in formulas (1) and (2): -0.04 (see Table 2).
  - Deposits-to-total funding (avg: 60%): differential coefficient (β1) reported -0.02 to -0.03; Table 2 reported -0.04***; Table A3.1 estimated -0.09*** (various specifications).
  - Loans-to-assets (avg: 55%): differential coefficient (β1) reported -0.02*** to -0.03***; Table A3.1 estimated -0.04***.
  - Deposits-to-loans (avg: 99%): Table 2 reported 0.05*** to 0.08***; Table A3.1 shows 0.07*.
  - Projection text preserved: "Since the average aged countries will approximately be as 'old' in 20235 as the oldest advanced economies at present (Figure 4), we can expect a general drop of the loan to asset ratios and the deposits to loan ratios, respectively."
- Other asset-composition signals:
  - Aging linked to a higher share ownership of securities and government securities; foreign claims have grown differentially for the oldest economies.

### Main empirical findings — risk taking and stability (Section IV.B–D and Annex III)
- Risk weight density (avg: 48.7%):
  - Table 3 summary coefficients: -0.015(*) to -0.02(*) for Metric 1; -0.03** to -0.04* for Metric 2.
  - Table A3.1 estimated risk weight density: -0.05*** in some specifications.
- Credit loss rates (avg: 1.5%):
  - Table 3: 0 to 0.002** for Metric 1; not significant for Metric 2.
  - Table A3.1 loan loss rate: 0.004***.
- Return on Equity (avg: 13.8%):
  - RoE generally not significant; some alternative specifications show significance.
- Z-Score (avg: 18%):
  - Descriptive/regression: Z-Scores for banks in aging countries are found to be 2.5 to 3 percentage points lower than in median aged economies (text preserved).
  - Annex III econometric summary: Coefficient for Aging (Metric 1): -0.024** to -0.03***; Metric 2: Not significant.
  - Compared with the youngest economy, the differential effect would about 5-6 percentage points, or one third of the average Z-Score metric across all advanced countries and over time.
  - Extrapolation: Z-Scores for the median aging country in 2035 would be 2.5-3 percentage points lower compared to 2022.
- Market-based tail-risk metrics:
  - EDF (avg: 0.68%): Coefficient for Aging (Metric 1): 0.004* to 0.005** in Table 4; Table A3.1 EDF: 0.0033**.
  - SRisk to total assets (avg: 3.1%): Coefficient for Aging (Metric 1): 0.016(*) to 0.04**; Table A3.1 SRisk: 0.015(*).
  - Banks in aging countries exhibit higher tail risks during severe stress despite lower average risks: one-year EDFs show comparable normal-time risk but pronounced susceptibility in stress; SRisk shows banks in aging countries exhibit a 1.5 percentage points higher SRisk to asset metric in some results.
- Asset-allocation and international exposure:
  - Shift toward securities (especially government bonds) documented; effect statistically significant in oldest three or five economies in some specifications.
  - No pronounced differential impact on foreign claims relative to GDP in aggregate, though some specifications show positive differential for oldest three economies.
- Interaction and marginal effects (Table A3.6):
  - Aging dummy x loans to assets: ZScore +(*); Price-to-book ratio +**; SRisk to total assets -***.
  - Aging dummy x deposits to loans: Price-to-book ratio -***; EDF -***; SRisk to total assets +***.
  - Aging dummy x risk weight density: SRisk to total assets -***.
  - Aging dummy x return on equity: EDF -***; SRisk to total assets +**.
  - Aging dummy x share of government securities: EDF -***; SRisk to total assets -(*).

### Robustness checks and limitations
- Robustness:
  - Results hold when controlling for macro-financial conditions and bank characteristics.
  - Excluding Greece and Italy does not materially affect most results; coefficients on EDF and SRisk decrease but often remain significant.
  - Varying onset cut-off years (2010, 2014, 2018) supports findings.
  - Excluding pandemic years (2020-2022) keeps findings resilient; some trends more pronounced in recent years.
  - Focusing on three most aged countries hints at stronger looming risks.
- Limitations:
  - Causal attribution is challenged by the gradual predictable nature of aging versus volatile banking crises and multiple interacting factors.
  - Limited number of observations and need for more granular bank-level data constrain identification of precise behavioral shifts and exposures.
  - Use of UN projections may be conservative relative to IIASA and IHME forecasts.

### Summary interpretation and main conclusion
- Bifurcated impact of aging on banking stability:
  - Average solvency risk: tends to decrease in aging countries, attributed to diminished demand for loans and more conservative asset allocations.
  - Tail risk: increases, attributed to banks channeling excess deposits into unfamiliar and potentially risky activities as they search for yield.
- Heterogeneity:
  - Larger, global banks may be better able to adapt via diversification and geographic expansion.
  - Smaller, regional banks face greater challenges adapting to aging and shrinking local populations.
- Consistency:
  - Findings align with prior regional/domestic studies (Japan, United States) showing reduced net interest margins, lower credit demand, and shifts toward securities and other activities.

### Policy implications and recommended actions
- Supervisors and regulators should:
  - Use scenario analysis to assess vulnerabilities from demographic shifts.
  - Monitor balance-sheet adjustments away from traditional maturity transformation toward securities and unfamiliar activities.
  - Pay particular attention to smaller regional banks that may lack diversification options.
  - Consider prudential measures where appropriate without weakening solvency profiles; keep fiscal and other backstops in view.
- Banks should:
  - Reassess strategic orientations for deposit funding as wealth transfers from baby boomers to millennials and Generation Z occur.
  - Evaluate demand shifts and risk profiles associated with new business lines and international expansion.
- Forward-looking initiatives:
  - Examples from Japan, Europe, and other Asia countries illustrate proactive private and public responses to demographic and financial shifts.
- Research agenda:
  - Explore implications of aging in emerging market economies.
  - Conduct detailed international bank-level analyses on effects of aging on banks’ business models and risk postures.

*International Monetary Fund — excerpts from "Aging Gracefully: Steering the Banking Sector through Demographic Shifts" (wpiea2024118-print-pdf).*

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

### Executive Summary

### Scope and purpose
- Study period: 2000-2022.
- Objective: Examine how demographic aging influences the stability of banking systems across advanced economies.
- Data basis: Demographic projections by the United Nations to 2100 and bank-level data for advanced economies (2000-2022).

### Demographic context (key numeric projections and facts)
- Over two-thirds of the world's population live in nations where fertility rates fall beneath the replacement level of 2.1 births per woman.
- Global median age: currently 30 years; projected to be 36 in 2050 and 42 by 2100.
- Average life expectancy: around 74 in 2022; projected to reach 77 by 2050 and 82 by 2100.
- Share of elderly (above 65): projected to increase from 10 percent at present to 24 percent by 2100.
- World population trajectory: projected to peak by 2090 and then decline (figure text also indicates a peak in 2087 at 10.4 billion).

### Conceptual transmission channels
- Aging reduces the role of banks in maturity transformation because:
  - Older households and firms demand less credit.
  - Many upcoming pensioners have paid off mortgages, reducing credit sensitivity to interest rates.
  - Older households shift to safer assets, strengthening links between households and the sovereign.
- Banks respond by recalibrating business models, including:
  - Reducing loan offerings.
  - Increasing investments in government bonds and other securities.
  - Diversifying into new financial activities.
  - Expanding internationally.
- These strategic shifts can open new revenue streams but also create potential new risks and tail exposures.

### Main empirical findings
- Bifurcated impact of aging on banking stability:
  - Average solvency risk: decreases in aging countries, attributed to diminished demand for loans (a historical precursor to banking crises).
  - Tail risk: increases, attributed to banks channeling excess deposits into unfamiliar and potentially risky activities.
- Heterogeneity across banks:
  - Larger, global banks may be better able to adapt via diversification and geographic expansion.
  - Smaller, regional banks face greater challenges adapting to aging and shrinking local populations.
- Consistency with prior regional/domestic studies:
  - Aligns with findings from studies of Japan and the United States where aging reduced net interest margins, lowered credit demand, and prompted shifts toward securities and other activities.

### Caveats and limitations
- Causal attribution is challenging due to:
  - The gradual and predictable nature of aging versus the volatile nature of banking crises.
  - Multiple interacting factors affecting bank stability.
  - Relatively recent acceleration in demographic aging and the limited time span for observable effects.
- UN projections are used for comparability; alternative forecasts (IIASA, IHME) point to slightly stronger aging and earlier population decline, implying this analysis may marginally underestimate future aging impacts.

### Policy implications and recommended actions
- Vigilant policy-making and strategic foresight are critical as populations age and, in some cases, decline.
- Supervisors and regulators should:
  - Use scenario analysis to assess underlying vulnerabilities arising from demographic shifts.
  - Monitor banks’ balance-sheet adjustments away from traditional maturity transformation toward securities and unfamiliar activities.
  - Pay particular attention to smaller regional banks that may lack diversification options.
- Banks should:
  - Reassess strategic orientations for deposit funding as wealth transfers from baby boomers to millennials and Generation Z occur.
  - Evaluate demand shifts and risk profiles associated with new business lines and international expansion.
- Forward-looking initiatives by private and public entities (examples cited from Japan, Europe, and other Asia countries) are useful proactive responses to demographic and financial shifts.

*International Monetary Fund — Executive Summary from "Aging Gracefully: Steering the Banking Sector through Demographic Shifts"*

### Section IV presents our empirical findings and robustness tests. Section V concludes with policy

### IV. Empirical findings: Lower average, but higher tail risk

### II. Theory: ambiguous impact of aging on bank stability
- Conceptual channels are multifaceted; aging affects households, firms, and banks through balance-sheet composition, risk preferences, and macro-financial conditions.
- Life-cycle hypothesis (Modigliani, 1975) implies hump-shaped wealth accumulation, but empirical "Wealth Decumulation Puzzle" (Horioka and Ventura, 2022) finds less wealth depletion among retirees and in some cases slight wealth increases, implying bank deposit levels are likely to remain elevated in aging societies.
- Factors sustaining deposits among the elderly: bequest motives, substantial public pension schemes, reluctance to liquidate real estate.
- Demographic shifts could later induce more pronounced decumulation as baby-boomer wealth transfers to Generations X, Y, and Z, and rising dependency ratios could pressure public pensions and private saving decumulation.

Key theoretical mechanisms:
- Structural change in banks’ balance sheet (balance sheet mismatch)
  - Reduced investment activity in aging economies -> lower demand for credit (Jones, 2022).
  - Examples: regulatory lending-age constraints (e.g., Slovakia; Jurca et al., 2020); corporate deleveraging in Japan (Kim et al., 2023).
  - A 10 percent rise in the share of individuals aged 60 and above correlates with a 5.5 percent drop in per capita GDP growth (GFSR, 2017).
  - Aging amplifies preference for secure, long-term investments, flattening the yield curve and compressing bank profitability.
  - Deposit-to-loan ratio secularly rises; transitory phase poses asset-liability mismatch risks.
  - Non-bank financial institutions may absorb intermediation activity (insurance, asset management), shifting risk landscape (beyond scope of banking-focused analysis).
- Changes in risk-taking behavior
  - Risk aversion typically increases with age (life-cycle theory of risk); older cohorts prefer liquidity and safer investments (Modigliani, 1986; Jianakoplos and Bernasek, 1998; Mather, 2006).
  - Behavioral and cohort effects (e.g., exposure to prolonged macro slumps) can permanently alter risk preferences (Hasegawa, 2011; Malmendier and Nagel, 2011; Sunde, 2023).
  - Bank reactions: pivot to fee-based and fixed-income assets, tailor retirement-related products, face asset-quality pressure (mortgage-backed assets) particularly for small, localized banks.
- Impact of banks’ balance sheet changes
  - Aging and higher savings with lower credit demand depress equilibrium interest rates, narrowing spreads (Summers, 2013; Rachel and Summers, 2019).
  - Low-rate environments can induce banks to seek higher yields via riskier assets, longer maturities, eased lending standards, or higher leverage (Bubeck et al., 2020).
  - Strategic adaptations and associated channels of risk:
    - Consolidation of banks -> “too big to fail/too complex to fail” concerns.
    - Expansion into new products -> operational and market risks (structured credit, CDS).
    - Increased sovereign exposure -> stronger sovereign-bank nexus, limiting fiscal backstop capacity.
    - International expansion -> cross-border credit, FX, country, and geopolitical risks.
    - Slower adoption of new technologies among smaller banks -> operational and client-retention risks.

Concluding theoretical insight:
- Aging may lower average risk (more conservative behavior) but encourage bank strategies that raise tail risks as banks search for yield in a low-growth, low-rate environment.

### III. Data and model specifications
- Empirical hypotheses:
  - Hypothesis 1: Aging generally lowers bank financial stability risk.
  - Hypothesis 2: Aging gives rise to potential tail risks, although the average risk might decrease.
- Sample and period:
  - 28 advanced economies, covering the years 2000 to 2022.
- Econometric approach:
  - Fixed effects panel regressions with country (i) and time (t) fixed effects.
  - Post-GFC dummy with a cut-off in 2010 to capture structural changes after the Global Financial Crisis.
- Baseline specifications (as presented):
  - 푦푖,푡 = 훼 + 훽1 퐴푔푖푛푔푖,푡 + 퐷푢푚푚푦푖 + 퐷푢푚푚푦푡 + 퐷푢푚푚푦푝표푠푡퐺퐹퐶 + 푒푖,푡   (1)
  - 푦푖,푡 = 훼 + 훽1 퐴푔푖푛푔푖,푡 + 훽푚1−3 푀푎푐푟표푖,푡 + 훽푏1−2 퐵푎푛푘푖,푡 + 퐷푢푚푚푦푖 + 퐷푢푚푚푦푡 + 퐷푢푚푚푦푝표푠푡퐺퐹퐶 + 푒푖,푡   (2)
- Dependent variables (Y_i,t): metrics capturing banks’ balance sheet changes, risk taking, and bank stability (sourced mainly from FitchConnect and other vendors).
  - Metrics to capture changes in bank balance sheets:
    - Deposits to total funding
    - Loans to assets
    - Deposits to loans deposits
  - Metrics to capture banks’ risk taking:
    - Risk weight density (ie risk weight per unit of asset)
    - Return on equity
    - Credit loss rates
  - Bank stability metrics:
    - Banks' Z-Scores (World Bank) — CAMEL-type metric; Z-score estimated as (ROA+equity/assets)/sd(ROA); sd(ROA) is the standard deviation of ROA.
    - Banks’ price to book ratios
    - Banks' Expected Default Frequency (EDFs) (Moody’s)
    - SRisk — market-based metric of worst-case losses (New York University)
- Controls:
  - Macro vector: real GDP growth, general economic uncertainty, policy rates, level of bank credit to GDP.
  - Bank vector: bank leverage, profitability (including credit losses).
  - All control variables lagged by 1 year.
- Aging operationalization:
  - Aging Metric 1: ratio of active population (aged 15-64) to elderly population (aged 65+).
  - Aging Metric 2: population growth compared to a decade earlier and year-on-year growth rates.
  - "Old" countries defined as five most aged advanced economies by Aging Metric 1: Finland, France, Greece, Italy, and Japan.
  - By Aging Metric 2 the eldest set becomes: Finland, Greece, Italy, Japan, and Spain.
  - Sensitivity checks: alternative definitions, varying “old” group composition (e.g., excluding Greece and Italy).
- Further notes:
  - Aging dummies for 4-year intervals—2010, 2014, and 2018—used to capture progressive intensification of aging.
  - Projections rely on UN population forecasts (2022); alternative forecasts suggest potentially faster aging (analysis possibly conservative).

### IV.A Empirical findings: Impact of aging on banks' assets and liabilities
Descriptive patterns:
- The “oldest” countries (by active population to elderly ratio: Finland, France, Greece, Italy, Japan) show a differential reduction in demand for loans, particularly over the past five years (Figure 5, blue dashed lines).
- This reduction in loan demand persists even when excluding Italy and Greece from the oldest-country group.
- Share of deposits in total funding mix has been marginally lower in these older countries but remains on an upward trend (Appendix 2); younger countries may currently be in a wealth accumulation phase (e.g., Japan, Figure 3b).
- Historically low interest rates encouraged shifts from deposits to alternative assets (e.g., real estate), but higher policy interest rates may reverse this trend and enhance liquidity.
- Across both aging metrics, the eldest countries have experienced excess deposits relative to loans — a broader phenomenon in advanced economies as they embark on aging trajectories.
- The influence of aging on balance-sheet adjustments is more pronounced when using the active-population-to-elderly ratio (Aging Metric 1) than when using ten-year cumulative population growth (Aging Metric 2).

Regression evidence:
- Regression analyses corroborate descriptive trends: banks in aging countries exhibit lower loan-to-asset ratios.
- Reported coefficient for aging effect on loan-to-asset ratios:
  - The coefficient b1 in formula (1) and (2) is -0.04 (see Table 2).

Summary empirical interpretation:
- Aging is associated with a decline in loan demand and a buildup of deposits, raising deposit-to-loan ratios and prompting banks to reallocate assets and adapt business models.
- While average measures of risk may decline with increased conservatism, behavioral adjustments by banks seeking yield can increase exposure to tail risks.

*Source: IMF Working Paper "Aging Gracefully: Steering the Banking Sector through Demographic Shifts" (excerpts from Sections II–IV and model specifications).*

### Appendix 3), suggesting diminished demand for loans as aging progresses.

### wpiea2024118-print-pdf - Appendix 3), suggesting diminished demand for loans as aging progresses.

### A. Effects of aging on banks' assets and liabilities
- Deposits-to-total funding:
  - Differential coefficient (β1 in formula 2): -0.02 to -0.03 (negative differential effect).
  - Table 2 reported coefficient: -0.04*** (expected sign: +) for one specification and -0.03** to -0.04*** (expected sign: +) for another.
  - Average deposits-to-total funding: 60%.
  - Interpretation: Banks in the most aged countries have lower deposit-to-total funding ratios compared to other advanced economies at earlier aging stages.
- Loans-to-assets:
  - Differential coefficient (β1 in formula 1): -0.02*** to -0.03*** (expected sign: -); another specification: -0.03(*) (expected sign: -).
  - Average loans-to-assets: 55%.
  - Economic magnitude: A decline in the share of active population to population 65+ of about 2-3 percentage points corresponds to a fairly notable effect on the share of loans in total assets; exemplified by an actual difference between Japan and Singapore of 15 percentage points (40 vs. 55 percent).
- Deposits-to-loans:
  - Aggregate differential effect: positive.
  - Table 2 reported: 0.05*** to 0.08*** (expected sign: +) for one metric; not significant for the other (expected sign: +).
  - Average deposits-to-loan ratio reported: 99%.
  - Differential impact observed on deposits-to-loan ratio: 5-8 percenage points.
- Projection based on demographics:
  - "Since the average aged countries will approximately be as 'old' in 20235 as the oldest advanced economies at present (Figure 4), we can expect a general drop of the loan to asset ratios and the deposits to loan ratios, respectively." (Preserve text as provided.)
- Other asset composition signals:
  - Aging is linked to a higher share ownership of securities and government securities.
  - Foreign claims have grown differentially for the oldest economies, potentially through indirect channels such as larger ownership of securities.

### B. Impact of aging on banks' risk taking
- Risk weight density:
  - Descriptive data: banks in aging advanced economies tend to reduce per-unit risk level ("risk weight density") of assets.
  - More pronounced when aging metric is ratio of working-age population to elderly.
  - Table 3 summary coefficients:
    - Risk weight density (avg: 48.7%): -0.015(*) to -0.02(*) (expected sign: -) for Metric 1; -0.03** to -0.04* (expected sign: -) for Metric 2.
  - Economic magnitude: differential impact from aging (metric 1) amounting to approximately 1.5 to 2 percentage points; contrasted with the youngest country the difference widens to approximately 3-4 percentage points. Many advanced economies exhibit risk weight density levels of 30-35 percent.
- Credit loss rates:
  - Descriptive: loan loss rates are notably higher in countries with aging populations, but this differential dissipates when Greece and Italy are excluded.
  - Table 3 summary coefficients:
    - Credit loss rates (avg 1.5%): 0 to 0.002** (expected sign: +) for Metric 1; not significant for Metric 2.
  - Overall: potential positive influence of aging on credit loss rates detected but not economically significant within current analysis scope.
- Return on Equity (RoE) and profitability:
  - Average RoE: 13.8%.
  - Table 3: RoE not significant for either metric (expected sign: -).
  - Descriptive: banks in aging societies have tended to be marginally lower on average in profitability and exhibited greater volatility; disparities in RoE and volatility diminish when Italy and Greece are excluded, implying macro-financial conditions matter.
  - Some evidence of higher loan rates for aging countries from additional regressions.
- Summary interpretation:
  - Econometric analysis of risk weight density, loan loss ratios, and RoE corroborates descriptive patterns: banks in aging societies hold portfolios with relatively lower risk weights, consistent with reduced lending and increased low-risk securities.
  - Tail risks affecting profitability metrics (higher volatility of RoEs) appear reduced when excluding countries that underwent financial crises (e.g., Greece and Italy).

### C. Impact of aging on banks' international exposure and asset allocation
- Foreign claims:
  - No pronounced differential impact related to banks' international expansion as measured by foreign claims relative to GDP, although significant in some econometric specifications.
  - Aggregate foreign-claim data may not capture subtler international engagements (e.g., securities markets) or heterogeneity across banks.
- Shift toward securities:
  - Evidence points to strategic adjustment by banks in aging nations toward greater focus on securities, particularly increased accumulation of government bonds.
  - Adjustment more pronounced in the three oldest economies; persists (less markedly) when broadened to the five oldest, with differential effect statistically significant.

### D. Impact of aging on bank stability
- Z-Score (stability metric):
  - Z-Score used to gauge default risk by comparing buffers (capitalization and profits) against earnings volatility; captures structural shifts rather than immediate tail risks.
  - Descriptive analysis depicts a differential reduction in Z-Scores for banks in aging countries.
  - Regression/econometric confirmation: "The Z-Score of banks in aging countries are found to be 2.5 to 3 percentage points lower than in median aged economies" (text preserved as provided; sentence cut off in source).

### Key statistics and coefficients (as reported)
- Deposits-to-total funding (avg: 60%): coefficients reported -0.04***; -0.03** to -0.04***.
- Loans-to-assets (avg: 55%): coefficients reported -0.02*** to -0.03***; -0.03(*).
- Deposits-to-loan (avg: 99%): coefficients reported 0.05*** to 0.08***; not significant in alternative metric.
- Risk weight density (avg: 48.7%): coefficients -0.015(*) to -0.02(*); -0.03** to -0.04*.
- Return on Equity (avg: 13.8%): not significant.
- Credit loss rates (avg 1.5%): coefficients 0 to 0.002**; not significant in alternative metric.
- Significance codes used: ***: 1%; **: 5%; *: 10%; (*) : 20%; t stats in parentheses (where reported).

*Source: IMF Working Paper "Aging Gracefully: Steering the Banking Sector through Demographic Shifts" — Appendix 3 content.*

### Appendix 3). Compared with the youngest economy, the differential effect would about 5-6 percentage points,

### wpiea2024118-print-pdf - Appendix 3). Compared with the youngest economy, the differential effect would about 5-6 percentage points,

### Z-Score, aging projections, and market-based indicators
- Compared with the youngest economy, the differential effect would about 5-6 percentage points, or one third of the average Z-Score metric across all advanced countries and over time.
- Extrapolation to the future: Z-Scores for the median aging country in 2035 would be 2.5-3 percentage points lower compared to 2022.
- Interpretation: Demographic shifts can influence foundational stability parameters within the banking sector, aligning with broader patterns of risk aversion and conservative financial management in aging societies.
- Market-based metrics show the most aged economies endured more adverse financial conditions during stress periods, reflected in:
  - Expected Default Frequencies (EDFs) (Figures 8).
  - Tail risk ratio “SRisk” (Figure 9).
  - Lower Price-to-Book ratios (Figure A2.7 in Appendix 2).

### Econometric summary (Table 4) — main coefficients and statistical significance
- Cut-off point for aging and young society dummy: 15th percentile (i.e. 5 out of 28 Advanced economies classified as “old” and “young”, respectively).
- Z-Score (avg: 18%)
  - Coefficient for Aging (Metric 1): -0.024** to -0.03*** (expected sign: -)
  - Coefficient for Aging (Metric 2): Not significant (expected sign: -)
- Price-to-book ratio (avg: 1.5)
  - Coefficient for Aging (Metric 1): -0.1(*) (expected sign: -)
  - Coefficient for Aging (Metric 2): Not significant (expected sign: -)
- EDF (avg: 0.68%)
  - Coefficient for Aging (Metric 1): 0.004* to 0.005** (expected sign: - or +)
  - Coefficient for Aging (Metric 2): Not significant (expected sign: - or +)
- SRisk to total assets (avg: 3.1%)
  - Coefficient for Aging (Metric 1): 0.016(*) to 0.04** (expected sign: - or +)
  - Coefficient for Aging (Metric 2): Not significant (expected sign: - or +)
- Significance codes: ***: 1%; **: 5%; *: 10%; (*) : 15%, t stats in parentheses.

### Key empirical patterns and drivers
- Aging (Metric 1) is associated with:
  - Decreases in average risk (Z-Score).
  - Increases in tail risk (EDF, SRisk).
- Aging (Metric 2: population growth) does not yet show a significant impact on most metrics, though graphical evidence suggests elevated stress-period vulnerabilities.
- Price-to-book ratios:
  - Aging has a negative contribution, but economic significance is fairly low: around 10 basis points for aging metric 1.
  - No significant differential impact for metric 2.
- Asset-liability drivers (Table A3.6):
  - More deposits relative to loans: contribute to tail risks and reduce price-to-book ratios, but lower EDFs.
  - Higher profitability (RoE): reduce EDFs but contribute to higher tail risk (SRisk).
  - Higher shares of foreign claims in total bank assets: not found to contribute to vulnerabilities across all four system-level risk metrics (data limitation: composition of foreign claims not controlled).
  - Higher allocation to government bonds and securities: appears to mitigate tail risks (SRisk).
  - Higher shares of loans: contribute to higher average bank risk (Z-Score) but lower tail risks and contribute positively to price-to-book ratios.

### EDFs, SRisk, and stress-period behavior
- One-year EDFs (Merton-type model) findings:
  - During normal times: risk levels of banks in older and younger societies are comparably aligned (Figure 8).
  - During severe stress: banking systems in aging countries exhibit pronounced susceptibility to tail risks.
  - Excluding Italy and Greece from aging sample: coefficient for EDF effect decreases by one fifth, from 0.005 to 0.004, yet remains statistically significant (Table 4).
- SRisk findings:
  - Banks in aging countries exhibit a 1.5 percentage points higher SRisk to asset metric than in less aging economies (Table 4).
  - The outcome increases in economic significance when Italy and Greece are excluded from the “old” sample.
  - If pattern holds over time across countries, tail risks would substantially increase in many countries.

### Robustness checks and limitations
- Robustness controls:
  - Controlling for macro-financial conditions and bank characteristics (Tables A3.2, A3.4 and A3.6) confirms stability of findings.
  - Excluding Greece and Italy (countries with recent crises) does not significantly affect most results.
  - Controlling for increasing proportion of non-bank financial intermediaries in total financial system assets (FSB data since 2002) does not alter results (limited country coverage).
- Sensitivity checks:
  - Varying onset cut-off years for aging (2010, 2014, 2018) consistently support initial findings.
  - Excluding pandemic years (2020-2022) keeps findings resilient; some trends become more pronounced in recent years, suggesting potential non-linear effects of aging on bank stability risk.
  - Focusing on three most aged countries (vs top five) hints at stronger looming risks.
  - Using year-on-year population growth (instead of 10-year trend) for metric 2 revealed a more substantial influence of aging (not reported separately).
- Limitations:
  - Limited number of observations constrains ability to identify aging's influence on precise behavioral shifts within the banking sector.
  - Need for more granular data to dissect bank-level exposures (e.g., subprime exposures before the Great Financial Crisis, composition of government bond holdings during sovereign debt and recent banking disturbances).
  - Bank-level analysis would preclude assessment of aggregate risk metrics at national level.

### Conclusion and policy implications
- Main conclusions:
  - Demographic shifts can lead banks to reduce maturity transformation, face diminished loan demand, and shift investment preferences toward more conservative assets.
  - On average, bank stability tends to improve in aging societies (lower credit risk-taking), but tail risks emerge due to behavioral shifts and search-for-yield strategies.
- Policy implications and recommendations:
  - Early detection and diligent monitoring of demographic-induced risks are vital for financial surveillance (analogy: "boiling frog" syndrome, "Minsky moments").
  - Policymakers and regulators should refine strategies to manage financial stability risks related to tail risks as banks adjust business models.
  - Employ meaningful scenario analyses to guide bank strategies and public policies.
  - Rigorous supervisory analysis to prompt supervisory diligence, for example regarding banks venturing into riskier foreign markets or new products to offset low structural profitability at home.
  - Prudential measures could be considered, but should not further weaken banks’ solvency profiles; other policy instruments to backstop banks, including fiscal ones, should not be ruled out.
- Future research suggestions:
  - Explore implications of aging in emerging market economies on banking system stability.
  - Conduct detailed international bank-level analysis on effects of aging on banks’ business models and risk postures.

### Data and sample notes (Annex I and Table A1.2 highlights)
- Sample composition:
  - 28 economies included in study (Table A1.2 lists economies and aging metrics for 2022, 2012, 2023, 2013, 2022, 2012).
  - Aging group (5 most aged economies) identified for Metric 1 and Metric 2; examples given earlier include Finland, France, Greece, Italy, Japan (Metric 1) and Finland, Greece, Italy, Japan, Spain (Metric 2).
- Key variable sources and availability (Table A1.1):
  - Risk weight density: From 2000 FitchConnect.
  - Loan impairments to total loans: From 2000 FitchConnect.
  - Return on equity: From 2000 FitchConnect.
  - Price-to-book ratio: From 1990 FitchConnect.
  - Expected default frequency of banks: From 2000, 22 countries Moody’s.
  - Z Scores: From 2000 WorldBank.
  - SRisk: From 2000, 22 countries NY Stern University.
  - Aging Metric 1: Active population to population above 65 — From 1990, 240 economies United Nations.
  - Aging Metric 2: Relative change of absolute population during the past 10 years and year-on-year — From 1990, 236 economies United Nations.
- Other covariates and sources include: Real GDP growth (IMF WEO), Uncertainty index (Home - World Uncertainty Index), Bank credit to GDP (BIS), Policy rates (BIS), Bank concentration (The Global Economy), Liquidity metric, Foreign claims, Capital adequacy ratio, Leverage ratio (FitchConnect where noted).

*Source: IMF Working Paper "Aging Gracefully: Steering the Banking Sector through Demographic Shifts" (selected excerpts from Appendix 3 and related sections).*

### Annex III. Supplementary regression results

### Annex III. Supplementary regression results

### Regression results for aging metric 1 (Active population to population above 65) — formula 1 specification (Table A3.1)
- Cut-off year for aging metric: 2010 (except Price-to-book ratio cut-off year 2014).
- Agingdummy (훽1) definition: 1 for five countries with lowest ratio of active population to population above 65 in 2022 (Finland, France, Greece, Italy, and Japan), otherwise 0.
- Estimated 훽1 coefficients (significance):
  - Deposits to total funding: -0.09***  
  - Loans to assets: -0.04***  
  - Deposits to loans: 0.07*  
  - Risk weight density: -0.05***  
  - Return on Equity: Not sig  
  - Loan loss rate: 0.004***  
  - ZScore: -0.28***  
  - Price-to-book ratio: -0.2**  
  - EDF: 0.0033**  
  - SRisk to total assets (cut-off year 2014): 0.015(*)  
- Observations and panel size:
  - Observations: 644 (for most columns), 585 (Return on Equity), 504 (ZScore), 561 (Price-to-book ratio), 506 (EDF), 474 (SRisk to total assets)
  - Number of countries: typically 28; exceptions: Return on Equity: 8; ZScore: 22; Price-to-book ratio: 20; EDF: 22; SRisk to total assets: 23
- Goodness of fit (selected):
  - R2 adjusted: 0.44 (Deposits to total funding), 0.004 (Loans to assets), 0.16 (Deposits to loans), 0.37 (Risk weight density), 0.11 (Return on Equity), 0.1 (Loan loss rate), 0.08 (ZScore), 0.43 (Price-to-book ratio), 0.14 (EDF), 0.14 (SRisk to total assets)
  - R2 within: 0.48, 0.08, 0.23, 0.42, 0.18, 0.17, 0.16, 0.49, 0.21, 0.22 (corresponding order)
- Fixed effects: Year Fixed Effects Y; Country Fixed Effects Y.
- Significance codes: ***: 1%; **: 5%; *: 10%; (*): 20%.

### Robustness and conditional results for aging metric 1 (Table A3.2)
- Default cut-off year for aging: 2011 in this exercise.
- Two "old" sample definitions shown:
  - Metric 1 (Finland, France, Greece, Italy, and Japan).
  - Metric 1 (Finland, France and Japan).
- Controls noted: with macro / bank controls we control for real GDP growth, general economic uncertainty, policy rates, the level of bank credit to GDP, bank leverage, and profitability.
- Key coefficient 훽1 results (without vs with macro/bank controls; t-statistics in parentheses; R2 adj reported):
  - Deposits to total funding (avg: 60%)
    - without controls: -0.09*** (-6.86); R2 adj: 0.43
    - with controls: -0.04*** (-3.73); R2 adj: 0.60
    - alternative old sample: -0.08*** (-5.13); R2 adj: 0.43 and -0.04*** (-3.24); R2 adj: 0.60
  - Loans to assets (avg: 55%)
    - without controls: -0.04*** (-3.31); R2 adj: 0.004
    - with controls: -0.02* (-1.53); R2 adj: 0.13
    - alternative old sample: -0.07*** (-5.72); R2 adj: 0.04 and -0.03** (-1.78); R2 adj: 0.12
  - Deposits to loans (avg: 99%)
    - without controls: Not sign (post 2018: 0.07* (1.8)); R2 adj: 0.16
    - with controls: 0.05*** (2.28); R2 adj: 0.52
    - alternative old sample: 0.08** (2.25); R2 adj: 0.16 and 0.08*** (3.23); R2 adj: 0.6
  - Risk weight density (avg: 48.7%)
    - without controls: -0.05*** (-3.42); R2 adj: 0.42
    - with controls: -0.015(*) (-1.07); R2 adj: 0.61
    - alternative: -0.07*** (-3.99); R2 adj: 0.36 and -0.02(*) (-1.1); R2 adj: 0.59
  - Return on Equity (RoE) (avg: 13.8%): Not significant in main sample; 0.05** (2.08); R2 adj: 0.13 in one alternative
  - Loan loss rate (avg: 1.5%)
    - without controls: 0.004*** (3.30); R2 adj: 0.1
    - with controls: 0.002** (1.94); R2 adj: 0.36
    - alternative: -0.002*** (-1.67); R2 adj: 0.12; Not significant in another variant
  - ZScore (avg: 18%)
    - without controls: -0.028*** (-3.05); R2 adj: 0.08
    - with controls: -0.003*** (-3.39); R2 adj: 0.34
    - alternative: -0.021** (-2.07); R2 adj: 0.09 and -0.024** (-2.36); R2 adj: 0.37
  - Price-to-book ratio (avg: 1.5)
    - without controls: -0.2** (-2.12); R2 adj: 0.44
    - with controls: -0.09(*) (-1.30); R2 adj: 0.66
    - alternative: -0.13(*) (-1.16); R2 adj: 0.41 and -0.1(*) (-1.15); R2 adj: 0.65
  - EDF (avg: 0.68%)
    - without controls: 0.0034** (2.15); R2 adj: 0.14
    - with controls: 0.005** (2.23); R2 adj: 0.19
    - alternative: 0.0023(*) (1.32); R2 adj: 0.11 and 0.004* (1.62); R2 adj: 0.16
  - SRisk to total assets (avg: 3.1%)
    - Generally Not significant; selected post-2014 coefficients: 0.015(*) (1.5) and 0.016(*) (1.31) with R2 adj: 0.14 and 0.25 respectively; other entries: 0.014(*) (1.48); R2 adj: 0.15 and 0.04** (2.27); R2 adj: 0.27
- Significance codes: ***: 1%; **: 5%; *: 10%; (*): 20%.

### Regression results for aging metric 2 (Population growth during the past 10 years) — time and country dummies only (Table A3.3)
- Cut-off year for aging metric: 2014 for Deposits to total funding; 2010 for most other metrics.
- Agingdummy definition: 1 for five countries with lowest 10 year population growth in 2022 (Finland, Greece, Italy, Japan, Spain), otherwise 0.
- Estimated 훽1 coefficients (significance):
  - Deposits to total funding: Not sign
  - Loans to assets: -0.03(*)  
  - Deposits to loans: 0.05*  
  - Risk weight density: -0.05*  
  - Return on Equity: Not sig  
  - Loan loss rate: Not sig  
  - ZScore: Not sig  
  - Price-to-book ratio: Not sig  
  - EDF: Not sig  
  - SRisk to total assets: Not sig
- Observations: 644 (most), 585 (Return on Equity), 504 (ZScore), 561 (Price-to-book ratio), 506 (EDF), 474 (SRisk).
- Number of countries: mostly 28; ZScore: 22; Price-to-book ratio: 20; EDF: 22; SRisk: 23.
- R2 adjusted and R2 within reported (selected):
  - R2 adjusted: 0.39 (Deposits to total funding), -0.01 (Loans to assets), 0.16 (Deposits to loans), 0.36 (Risk weight density), 0.11 (Return on Equity), 0.08 (Loan loss rate), 0.06 (ZScore), 0.43 (Price-to-book ratio), 0.13 (EDF), 0.14 (SRisk)
  - R2 within: 0.44, 0.07, 0.23, 0.41, 0.18, 0.15, 0.15, 0.48, 0.20, 0.22
- Fixed effects: Year Fixed Effects Y; Country Fixed Effects Y.
- Significance codes: ***: 1%; **: 5%; *: 10%; (*): 20%.

### Robustness and conditional results for aging metric 2 (Table A3.4)
- Default cut-off year for aging: 2011 in robustness exercises.
- Two "old" sample definitions shown:
  - Metric 2 (Finland, Greece, Italy, Japan, Spain).
  - Metric 2 (Finland, Japan, Spain).
- Controls: with macro / bank controls include real GDP growth, general economic uncertainty, policy rates, the level of bank credit to GDP, bank leverage, and profitability.
- Key 훽1 coefficient results (without vs with controls; t-statistics in parentheses; R2 adj reported):
  - Deposits to total funding (avg: 60%)
    - Not significant without controls
    - with controls: -0.04*** (-1.73); R2 adj: 0.58
    - alternative old sample: Not significant; with controls: -0.03** (-1.52); R2 adj: 0.59
  - Loans to assets (avg: 55%)
    - without controls: -0.03(*) (-1.45); R2 adj: -0.013
    - with controls: Not significant (R2 adj: -0.013)
    - alternative old sample: -0.03(*) (-1.52); R2 adj: -0.011 and -0.03(*) (-1.21); R2 adj: 0.13
  - Deposits to loans (avg: 99%)
    - without controls: 0.05* (1.56); R2 adj: 0.16
    - with controls: Not significant
    - alternative: Not significant
  - Risk weight density (avg: 48.7%)
    - without controls: -0.05(*) (-1.73); R2 adj: 0.36
    - with controls: -0.04(*) (-1.30); R2 adj: 0.52
    - alternative: -0.05* (-1.74); R2 adj: 0.35 and -0.04(*) (-1.36); R2 adj: 0.5
  - Return on Equity (RoE) (avg: 13.8%): Not significant in all variants
  - Credit loss rate (avg: 1.5%): Not significant in all variants
  - Solvency/systemic risk metrics (ZScore, Price-to-book ratio, EDF, SRisk to total assets): Not significant in these robustness results
- Significance codes: ***: 1%; **: 5%; *: 10%; (*): 20%.

### Differential impacts on balance sheet structure (Table A3.5)
- Agingdummy (훽1) as oldest three economies, controlling for macro and bank effects:
  - Securities to Total Assets: Not sign
  - Government securities to total assets: Not sign
  - Foreign claims: +** (positive at 5%)
- Agingdummy (훽1) as oldest five economies, controlling for macro and bank effects:
  - Securities to Total Assets: +*** (positive at 1%)
  - Government securities to total assets: Not sign
  - Foreign claims: Not sign
- Note: 1/ Not significant if Italy and Greece are removed from the sample.
- Aging dummy definition: set to 1 for five countries with lowest ratio of active population to population above 65 in 2022, otherwise 0.
- Significance codes: ***: 1%; **: 5%; *: 10%; (*) : 20%.

### Marginal differential impacts (interaction effects) for aging metric 1 (Table A3.6)
- Specification: add one-by-one the interaction terms shown to specification 2 (section III), controlling for macro and bank characteristics. Aging dummy set to 1 for five countries with lowest ratio of active population to population above 65 in 2022.
- Dependent variables shown: ZScore, Price-to-book ratio, EDF, SRisk to total assets.
- Interaction term effects (significance as shown):
  - Agingdummy x loans to assets: ZScore +(*); Price-to-book ratio +**; EDF Not sign; SRisk to total assets -***
  - Agingdummy x deposits to loans: ZScore Not sign; Price-to-book ratio -***; EDF -***; SRisk to total assets +***
  - Agingdummy x risk weight density: ZScore - (*) ; Price-to-book ratio Not sign; EDF Not sign; SRisk to total assets -***
  - Agingdummy x return on equity: ZScore Not sign; Price-to-book ratio Not sign; EDF -***; SRisk to total assets +**
  - Agingdummy x foreign claims: ZScore Not sign; Price-to-book ratio Not sign; EDF -*; SRisk to total assets Not sign
  - Agingdummy x share of securities: ZScore Not sign; Price-to-book ratio Not sign; EDF - (*) ; SRisk to total assets Not sign
  - Agingdummy x share of government securities: ZScore Not sign; Price-to-book ratio Not sign; EDF -***; SRisk to total assets -(*)
- Note: The result is comparable for dropping Greece and Italy from the “old” sample.
- Significance codes: ***: 1%; **: 5%; *: 10%; (*) : 20%.

*Source: Annex III. Supplementary regression results, IMF Working Paper "Aging Gracefully: Steering the Banking Sector through Demographic Shifts".*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024118-print-pdf.pdf_