## sdnea2025001

## Source details

**Canonical URL:** [sdnea2025001](https://www.imf.org/-/media/files/publications/sdn/2025/english/sdnea2025001.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/sdn/2025/english/sdnea2025001.pdf.md)
- [Structured JSON version](/-/media/files/publications/sdn/2025/english/sdnea2025001.pdf.json)

---

### Executive summary: core findings
- Banks’ income and expense tend to rise with inflation; gross exposures are large.
  - Income from interest-rate and non-traditional businesses, borrowing costs, and general business expenses all rise when inflation surges.
  - Interest-rate business exposures operate indirectly through policy rates; non-interest income and expense operate directly with inflation.
- Despite large gross exposures, most banks offset income and expense exposures and show little change in profitability as inflation shifts.
  - The return on assets (ROA) of the average bank appears largely unresponsive to either expected or unexpected inflation across a broad group of countries over the past three decades.
- Bank-level pockets of vulnerability exist.
  - Some banks have large simultaneous exposures to increases in inflation and policy rates observed in 2021–23.
  - About 3 percent of advanced economy banks and 6 percent of emerging market and developing economy banks—comprising 8 percent of assets in the average country—have interest-rate exposures at least as large as Silicon Valley Bank’s at the onset of its failure.
  - Accounting for non-interest exposures, 5 percent of advanced economy banks and 8 percent of emerging market and developing economy banks—comprising 9 percent of assets in the average country—could experience losses greater than 2 percent of assets.
  - Data up to 2022 show that 5 percent of banks experienced a decline in net income of at least 0.5 percent of assets.

### Conceptual distinctions and empirical strategy
- Key conceptual dimensions:
  - Gross versus net exposures: income and expense streams can each be highly exposed to inflation even if net profitability is not.
  - Direct versus indirect inflation effects: inflation can matter directly (through prices/wages/fees) or indirectly via policy-rate responses.
- Empirical operationalization:
  - Separate interest and non-interest lines using balance sheet and income statement data.
  - Distinguish expected and unexpected inflation and isolate inflation-driven policy-rate variation from inflation orthogonal to policy rates.
  - Use granular bank-level data combined with IMF macro projections across nearly three decades.
  - Regression approach: panel-fixed-effect specifications, bank and time fixed effects, block-bootstrapped country-level standard errors, regressions weighted by 1/(number of banks in each country-year).

### Empirical evidence on transmission channels
- Interest-rate business (indirect exposure):
  - A within-country two standard deviation rise in expected inflation for two years increases interest income margins by about 150 basis points and interest expense margins by about 100 basis points.
  - Average margins: interest income about 650 basis points; interest expense about 300 basis points.
  - Interest business exposure is primarily indirect—operating via policy-rate responses rather than directly to inflation orthogonal to policy rates.
  - Euro-area evidence supports that common monetary policy creates indirect exposure where inflation varies across countries.
- Non-interest business (direct exposure):
  - A one standard deviation move in expected (unexpected) inflation over two years is estimated to increase non-interest income by about 100 (80) basis points relative to total assets, and non-interest expense by about 80 (40) basis points relative to total assets.
  - Averages: non-interest income about 135 basis points; non-interest expense about 230 basis points.
  - Non-interest income is more sensitive to unexpected inflation; non-interest expense responds to both expected and unexpected inflation.
  - Non-traditional income (for example, asset management) shifts with unexpected inflation; fees and trading income do not play a significant role.
  - Non-wage components of non-interest expense (including rent) have meaningful exposures; wages do not systematically shift with inflation.

### Aggregate profitability and line-by-line effects
- Aggregate ROA largely insensitive to inflation:
  - A rise of 100 basis points in inflation for two consecutive years—either expected or unexpected—leads only to a little over a 1 basis point increase in ROA, compared with average ROA of nearly 100 basis points.
  - This relationship is not statistically significant at the 10 percent level.
- Line-by-line decomposition:
  - Net interest margins (NIMs) are sensitive to expected inflation (indirect channel) but not to unexpected inflation.
  - Net non-interest income matters mainly for unexpected inflation (direct channel).
  - Loan impairment charges: expected inflation appears to matter indirectly; unexpected inflation appears to matter directly and tends to reduce borrowers’ ability to repay.
  - Tax expense is the largest omitted component in the ROA breakdown and does not appear exposed to inflation or policy rates.

### Cross-country and banking-system patterns
- Coverage and sample:
  - Figures use data for 81 countries (28 advanced economies and 53 emerging market and developing economies).
  - Regression sample: annual bank-level data for more than 6,600 banks in 59 countries (28 advanced and 31 emerging market and developing economies) during 1995–2022; final panel: 5,496 banks in AEs and 1,149 banks in EMDEs (total 6,645 banks).
- Cross-country heterogeneity drivers:
  - Contracting conventions, regulatory frameworks, business models, competitive pressures, liability stickiness, reliance on foreign funding, maturity structures, and compliance with international regulatory standards.
- Typical patterns:
  - Gross exposures larger in interest-rate business than in non-interest business.
  - Income and expense pass-throughs within systems often similar, producing small net exposures on average.
  - Emerging market and developing economy banks tend to show positive net exposures in interest businesses due to deposit reliance and financial repression.

### Bank-level heterogeneity and recent experience (2021–23 illustrative scenario)
- Scenario numerical inputs:
  - Median increase in interest rates between 2021 and 2023: 4.5 percentage points.
  - Median unexpected inflation over the same period: 7 percent.
- Distributional outcomes (shifts in income as percent of assets):
  - 46 percent of advanced-economy banks and 55 percent of emerging market and developing economy banks experience shifts within –0.5 and 0.5 percent of assets.
  - About 5 percent of advanced economy banks and 8 percent of emerging market and developing economy banks experience losses larger than 2 percent of assets.
  - In the average country, these banks account for 9 percent of banking system assets.
- Sizable relative to capital-to-asset ratios:
  - 7.5 percent in advanced economies.
  - 10.8 percent in emerging market and developing economies.
- Systemic considerations:
  - Even limited losses at individual banks can prompt reassessment of the whole sector and lead to panic-fueled contagion; SVB is identified as an outlier with unusually large indirect interest-rate-business exposures.

### Market, liquidity, and run risk interactions
- Sharp interest-rate increases reduce market value of fixed-rate, long-dated assets; forced sales to meet withdrawals can realize losses and weaken capital.
- Banks with uninsured liabilities and large market exposure could face runs; withdrawals funded by asset sales at market prices can produce losses relative to bank equity.
- Illustrative central scenario (from cited empirical work): 20 percent withdrawals in vulnerable banks.
- Aggregated illustrative equity levels shown: equity = 3.5% (advanced economies) and equity = 0.1% (emerging market and developing economies) in plotted scenarios.

### Data, variables, and methodological details
- Primary data sources: Fitch Connect (bank financials, update from November 1, 2023); IMF World Economic Outlook (expected and realized inflation and real GDP growth); BIS, Haver Analytics, Bloomberg Finance L.P., IMF IFS (policy rates).
- Definitions:
  - Expected inflation: log change of CPI as predicted in the October forecast of the previous year.
  - Unexpected inflation = realized inflation − expected inflation.
  - Regressions use orthogonalization to separate policy-rate-driven effects from inflation orthogonal to policy rates (specifications (2.1) and (2.2) described).
- Estimation:
  - Unbalanced panel 1995–2022; regressions weighted by 1/(number of banks in each country-year).
  - Standard errors block-bootstrapped at the country level with 20,000 draws.
  - Supply-driven inflation proxied by a dummy where real GDP growth and inflation surprises have opposite signs.

### Key summary statistics (selected)
- Bank-level means and ranges:
  - ROA: Mean 0.5; Median 0.3; Standard Deviation 0.9; Min -3.6; Max 5.3.
  - NIM: Mean 2.7; Median 2.4; Standard Deviation 1.7; Min 0.2; Max 13.6.
  - Net non-interest income (scaled by total assets): Mean -1.3; Median -1.2; Standard Deviation 1.4; Min -32.8; Max 155.1.
  - Loan impairment charge (scaled by total assets): Mean 0.4; Median 0.2; Standard Deviation 1; Min -27.1; Max 75.4.
  - Deposits over liabilities: Mean 88.4; Median 94.5; Standard Deviation 15.2; Min 0; Max 100.
  - Equity over assets: Mean 8.9; Median 8; Standard Deviation 5.6; Min -14.5; Max 99.3.
  - Securities over assets: Mean 20.1; Median 18.6; Standard Deviation 13.7; Min 0; Max 99.8.
- Macroeconomic indicators:
  - Inflation shock: Mean 0.5; Median 0.0; Standard Deviation 4.4; Min -10.0; Max 45.2.
  - Realized inflation: Mean 4.6; Median 2.8; Standard Deviation 7.8; Min -1.4; Max 109.9.
  - Policy rate: Mean 5.4; Median 3.9; Standard Deviation 6.2; Min -0.8; Max 49.3.
  - Expected real GDP growth: Mean 3.4; Median 3.3; Standard Deviation 1.9; Min -4.1; Max 9.5.
  - Unexpected real GDP growth: Mean -0.5; Median -0.1; Standard Deviation 3.0; Min -13.2; Max 6.9.

### Policy implications and recommendations
- Monetary policy:
  - Across many countries, little evidence that banks are meaningfully exposed to monetary policy tightening aimed at tackling inflation spikes—supporting central banks’ pursuit of price stability.
  - Central banks should, however, be mindful that tightening might generate large losses for outlier banks and could create price–financial stability trade-offs if contagion occurs.
- Prudential and supervisory policy:
  - Strengthen prudential regulation and supervision to limit risks posed by outlier bank exposures.
  - Heighten requirements for risk-management governance and improve transparency of income- and expense-exposure profiles.
  - Use granular risk assessments to calibrate micro- and macroprudential capital requirements along key dimensions (interest-rate and non-interest exposures; direct versus indirect inflation effects).
- Containment of contagion risk:
  - Robust ex-ante measures and timely supervision are important because even limited losses at individual banks can prompt sector-wide panics and cross-border contagion.
  - Monitor specific banks with significant gross exposures and consider central bank funding facilities as mitigating factors.

*International Monetary Fund — Staff Discussion Note: Inflation and Bank Profits: Monetary Policy Trade-offs (sdnea2025001)*

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

### Executive Summary

### Key findings
- Banks’ income and expense tend to rise with inflation; gross exposures are large.
  - Income from interest-rate and non-traditional businesses, borrowing costs, and general business expenses all rise when inflation surges.
  - Interest-rate business exposures operate indirectly through policy rates; non-interest income and expense operate directly with inflation.
- Despite large gross exposures, most banks offset income and expense exposures and show little change in profitability as inflation shifts.
  - The return on assets (ROA) of the average bank appears largely unresponsive to either expected or unexpected inflation across a broad group of countries over the past three decades.
- There are pockets of vulnerability at the bank level.
  - Estimated bank-level exposures to simultaneous increases in inflation and policy rates seen in 2021–23 are large in both directions for some banks.
  - Some 3 percent of advanced economy banks and 6 percent of emerging market and developing economy banks—comprising 8 percent of assets in the average country—have interest-rate exposures at least as large as Silicon Valley Bank’s at the onset of its failure.
  - Accounting for non-interest exposures as well, 5 percent of advanced economy banks and 8 percent of emerging market and developing economy banks—comprising 9 percent of assets in the average country—could experience losses greater than 2 percent of assets.
  - Data up to 2022 show that 5 percent of banks experienced a decline in net income of at least 0.5 percent of assets.

### Conceptual distinctions and empirical strategy
- Key conceptual dimensions:
  - Gross versus net exposures: income and expense streams can each be highly exposed to inflation even if net profitability is not.
  - Direct versus indirect inflation effects: inflation can matter on its own (direct) or via its impact on policy rates (indirect).
- Operationalization in the empirical framework:
  - Separately analyze interest and non-interest lines of business using balance sheet and income statement data.
  - Distinguish expected and unexpected inflation and separate exposure through policy rates from direct exposure to inflation.
  - Use granular bank-level data combined with IMF macroeconomic projections over nearly three decades.

### Empirical evidence on channels
- Interest-rate business:
  - Exposed to inflation indirectly: interest income and interest expense rise only insofar as policy rates react to rising inflation.
  - Evidence within a large currency union (euro area) supports the interpretation that common monetary policy creates only indirect exposure where inflation varies across countries.
- Non-interest business:
  - Exposed to inflation directly: non-traditional banking services income, salaries, rent, and other operating costs increase with inflation independent of policy rate changes.
  - Direct exposure is especially relevant if inflation is supply-driven.

### Cross-country and banking-system patterns
- Exposures vary significantly across countries due to contracting conventions, regulatory frameworks, business models, and competitive pressures.
- Typically, gross exposures are larger in the interest-rate business than in the non-interest business.
- Banking-system level:
  - Gross interest-rate and non-interest exposures are generally offset, producing small net exposures on average.
  - Net non-interest exposures can be meaningful if inflation spikes unexpectedly.
  - In emerging market and developing economies, net exposures for interest-rate businesses tend to be, if anything, positive.

### Bank-level heterogeneity and recent experience
- Heterogeneity is large: some banks benefit from recent inflationary and high-interest-rate environments while others face large losses.
- The 2021–23 episode, including supply-driven inflation surprises and rapid policy tightening, created conditions where some banks’ estimated exposures are large enough to generate meaningful losses.
- Historical evidence and cross-sectional estimates indicate that even with strengthened regulation and supervision, outlier losses could trigger broader confidence effects and contagion.

### Policy implications and recommendations
- Monetary policy:
  - Across a wide range of countries, there is little evidence that banks are meaningfully exposed to monetary policy tightening aimed at tackling spikes in inflation—good news for central banks seeking price stability.
  - However, central banks should be mindful that tightening might generate large losses for outlier banks and could create price–financial stability trade-offs if contagion occurs.
- Prudential and supervisory policy:
  - Strengthen prudential regulation and supervision to limit the risk posed by outlier bank exposures.
  - Heighten requirements for risk-management governance and improve transparency of income- and expense-exposure profiles.
  - Use granular risk assessments to calibrate micro- and macroprudential capital requirements along the key dimensions identified (interest-rate and non-interest exposures, direct versus indirect inflation effects).
- Containment of contagion risk:
  - Even limited losses at individual banks can prompt reassessment of the entire sector and lead to panic-fueled contagion with systemic consequences, within and across borders; robust ex-ante measures and timely supervision are important to reduce this risk.

### Data and scope
- Data and coverage:
  - Combines balance sheet and income statement data for more than 6,600 banks operating in a broad cross section of advanced and emerging market and developing economies.
  - Empirical analysis spans nearly three decades and distinguishes income and expense exposures across lines of business.
- Empirical validation:
  - Results are consistent within a large currency union and with data up to 2022 that confirm predicted dispersion in bank performance and that some banks have already seen hits to profitability.

*International Monetary Fund — Staff Discussion Note: Inflation and Bank Profits: Monetary Policy Trade-offs (Executive Summary)*

### 1. Components of Bank Profits (1995–2022) 2. Line-of-Business Mix across Countries (1995–2022)

### sdnea2025001 - 1. Components of Bank Profits (1995–2022) 2. Line-of-Business Mix across Countries (1995–2022)

### Data, scope, and empirical strategy
- Data sources: Fitch Connect; IMF staff analysis; Bank for International Settlements; Bloomberg Finance L.P.; Haver Analytics (as noted in figure captions).
- Cross-sectional and time coverage as described in the chapter:
  - Figure panels use data that include 81 countries (28 advanced economies and 53 emerging market and developing economies).
  - The empirical analysis uses annual bank-level data for more than 6,600 banks in 59 countries (28 advanced and 31 emerging market and developing economies) during 1995–2022.
- Key identification features:
  - Distinguishes expected and unexpected inflation using IMF World Economic Outlook forecasts and differences of actual and forecasted inflation.
  - Separately estimates exposures to (a) inflation-driven variation in policy rates and (b) inflation variation orthogonal to policy rates to identify indirect (via policy) versus direct exposures.
  - Uses panel-fixed-effect specifications, country- and bank-level heterogeneity, and block-bootstrapped standard errors at the country level. Regressions are weighted by 1/(number of banks in each country-year).
  - Euro-area within-sample variation (common monetary policy but differing country inflation) is used to sharpen inference about indirect versus direct channels.

### Conceptual framework — how inflation can affect bank profitability
- Distinguishes gross exposures (individual income and expense lines) from net exposures (overall profitability), and interest versus non-interest businesses.
- Two main transmission channels:
  - Indirect channel: inflation → expected inflation → policy rate responses → changes in market, lending, and funding rates; relevant for the interest business where assets/liabilities reprice with policy/market rates.
  - Direct channel: inflation → changes in prices and wages for services/expenses and fee income; relevant for the non-interest business where prices/wages may respond directly to inflation.
- Importance of maturity composition and maturity mismatch:
  - Interest-earning assets and interest-bearing liabilities that reprice frequently or have short maturities are highly responsive to policy/market rates; long-term fixed nominal contracts are less responsive.
- Asset-quality channel is theoretically ambiguous:
  - Higher inflation with fixed policy rates → lower real rates → easier debt servicing for nominal-fixed borrowers (improves asset quality).
  - Higher inflation that drives policy tightening or raises costs faster than incomes → higher debt-servicing burdens and worse asset quality.
- Empirical exercise isolates supply-driven inflation episodes where unexpected inflation and unexpected GDP growth move in opposite directions.

### Key empirical findings on gross exposures (income and expense lines)
- Interest business:
  - A within-country two standard deviation rise in expected inflation for two years is estimated to increase interest income margins by about 150 basis points and interest expense margins by about 100 basis points.
  - These increases are large relative to average margins: average interest income margins of about 650 basis points and average interest expense margins of about 300 basis points.
  - Interest income and interest expense remain elevated up to four years after the initial rise in expected inflation.
  - Total exposure of the interest business to expected inflation is similar in magnitude to exposure to policy rates independent of inflation.
  - Direct exposure to expected inflation (inflation orthogonal to policy rates) in the interest business is minimal, implying the interest business is primarily indirectly exposed via policy-rate responses.
  - Euro-area within-sample evidence: despite country-level variation in inflation within the euro area, interest-business exposures require shifts in nominal policy rates to materialize—consistent with an indirect channel through policy rates.

- Non-interest business:
  - Non-interest income and non-interest expense are exposed to both expected and unexpected inflation.
  - Non-interest income is more sensitive to both expected and unexpected inflation than non-interest expense.
  - Some fee-based income streams and services can have prices that respond directly to inflation; key operating expenses (wages, rent, procurement) also respond and typically more than offset non-interest income in many systems.

### Interpretation: direct versus indirect exposures
- Exposures observed in regressions that include variation in policy rates driven by inflation but not in specifications isolating inflation orthogonal to policy rates are interpreted as indirect (policy-mediated).
- Exposures observed in both specifications (policy-driven and policy-orthogonal) are interpreted as direct exposures to inflation.
- Empirical evidence indicates:
  - Interest business exposures to expected inflation operate predominantly indirectly through policy-rate responses.
  - Non-interest business shows evidence of direct exposure to both expected and unexpected inflation.

### Implications for asset quality and macro linkages
- The net effect of inflation on asset quality is theoretically ambiguous and depends on:
  - Whether inflation is demand-driven (increases activity and incomes) or supply-driven (raises costs and depresses activity).
  - The monetary policy response: if policy tightens in response to inflation, higher interest rates can raise borrower debt-servicing costs.
- Empirical analysis accounts for these distinctions by isolating supply-driven inflation episodes and separating expected versus unexpected inflation.

### Methodological robustness and notes
- Results are robust to excluding the COVID-19 periods and to including unemployment as an additional macroeconomic control.
- Figures and regression specifications sum contemporaneous and lagged coefficients; statistical significance indicated at the 10 percent level with block-bootstrapped country-level standard errors.
- Regression weighting: 1/(number of banks in each country-year).

*Source: IMF staff analysis as presented in "INFLATION AND BANK PROFITS: A COMPLEX CONCEPTUAL RELATIONSHIP" and associated figures and notes from the chapter content provided.*

### Annex Tables 3.2 and 3.3 in Annex 3 present the full results behind this figure.

### Annex Tables 3.2 and 3.3 in Annex 3 present the full results behind this figure.

### Major findings
- A one standard deviation move in expected (unexpected) inflation over two years is estimated to increase:
  - non-interest income by about 100 (80) basis points relative to total assets, and
  - non-interest expense by about 80 (40) basis points relative to total assets.
- These changes are large relative to their respective averages of about 135 and 230 basis points.
- The effect of unexpected inflation is short-lived and subsides within the following year, while the effect of expected inflation lingers longer (see Annex Figure 4.1).
- Non-traditional sources of income (for example, asset management) shift with unexpected inflation, while fees and trading income do not appear to play a significant role (see Annex Figure 4.3).
- Non-wage components of non-interest expense (which include rent) have meaningful exposures to expected and unexpected inflation; wages do not appear to systematically shift with inflation (see Annex Figure 4.4).

### Interest business: indirect exposure and drivers
- Interest business exposures are largely indirect (operate through monetary policy reaction):
  - Total exposure to expected inflation is equally strong across demand- and supply-driven episodes.
  - Direct exposure to expected inflation is insignificant for both demand- and supply-driven episodes.
- Interpretation: interest business exposures matter insofar as monetary policy reacts to any source of inflation.
- Country-level features of interest-rate pass-through:
  - Gross income and expense exposures to policy rates are sizable and statistically significant in most countries.
  - Banking systems display gross exposures ranging from 0 to close to 1 (full pass-through of policy interest rates).
  - Income and expense exposures are very similar within banking systems, resulting in generally much smaller net exposures than gross exposures.
- Liability structure helps explain cross-country differences:
  - Systems with sticky liabilities tend to secure larger profit margins via funding costs lower than market rates.
  - Systems relying more on foreign funding see costs of funds that align more closely with market rates.
  - Emerging market and developing economy banks often rely more on depositors (less wholesale funding) and face more financial repression, contributing to greater positive exposures.

### Non-interest business: direct exposure and sources
- Non-interest business is directly exposed to both unexpected and expected inflation:
  - Non-interest income is exposed only to unexpected inflation, not to expected inflation.
  - Non-interest expense may be exposed to both expected and unexpected inflation.
  - Total and direct effects of unexpected inflation on non-interest business are similar.
- Source decomposition:
  - Direct exposure of the non-interest business is driven primarily by supply-side inflation, especially for unexpected inflation.
  - This is consistent with the less durable effect of unexpected inflation on the non-interest business.

### Aggregate profitability and line-by-line effects
- Overall bank returns are generally not exposed to inflation:
  - A rise of 100 basis points in inflation for two consecutive years—either expected or unexpected—leads only to a little over a 1 basis point increase in ROA, compared with the average ROA of nearly 100 basis points.
  - This relationship is not statistically significant at the 10 percent level.
  - Overall profitability also appears insensitive to variation in policy rates that is unrelated to inflation.
- Line-by-line decomposition (Figure 8):
  - Net interest margins (NIMs) are sensitive to expected inflation, but not to unexpected inflation; this effect is indirect.
  - Net non-interest income is directly relevant for unexpected inflation.
  - Loan impairment charges: on net, inflation reduces borrowers’ ability to repay, even after controlling for dynamics in output growth; expected inflation appears to matter indirectly (higher required payments), while unexpected inflation appears to matter directly (higher costs to borrowers than ability to repay).
- Tax expense is the largest omitted component in the ROA breakdown and does not itself appear exposed to either inflation or policy rates.

### Cross-country heterogeneity in gross and net exposures
- Gross exposures vary significantly across countries, but net exposures tend to be limited because income and expense pass-throughs are often similar within systems.
- Interest business (country-level):
  - Income and expense gross exposures vary from 0 to close to 1 in pass-through.
  - Many emerging market and developing economy banking systems see rising incomes in response to higher rates.
- Non-interest business (country-level):
  - Direct exposures of non-interest income and expense to unexpected inflation differ across countries.
  - Net exposures for the non-interest business are more varied across banking systems than for the interest business.
  - Expenses often appear more sensitive than incomes in many emerging market and developing economy banking systems, possibly reflecting more prevalent price indexation.
- Factors explaining cross-country differences include liability stickiness, reliance on foreign funding, maturity structure of assets and liabilities, and compliance with international regulatory standards.

### Heterogeneity across banks
- Bank-level exposures:
  - Gross exposures are generally large at the bank level, but net exposures are minimal for most banks.
  - Emerging market and developing economy banks tend to have larger indirect exposures to interest rates—in both directions—than banks in advanced economies.
  - About 3 percent of banks in advanced economies exhibit unusually large net exposures (example noted: Silicon Valley Bank had unusually large exposures).
- Implication: differing gross exposures and differing degrees of offsetting across banks can lead to significant variation in net exposures at the bank level, which could matter for financial stability if losses concentrate on particular institutions.

### Statistical and methodological notes referenced
- Figures compare total effects of inflation (controlling for policy rates orthogonal to inflation) and direct effects (inflation orthogonal to policy rates); see specifications (2.1) and (2.2) in Annex 2.
- Bars in the figures are sums of contemporaneous and lagged coefficients; filled bars indicate statistical significance at the 10 percent level; unfilled bars indicate statistical insignificance.
- Standard errors are block-bootstrapped at the country-level.
- Regressions are weighted by 1/(number of banks in each country-year).
- Annex Tables 3.2 and 3.3 in Annex 3 present the full regression results behind the figure on non-interest income and expense; Annex Table 3.4 presents the full regression results behind the ROA and line-by-line analysis.

*Source: STAFF DISCUSSION NOTES — Inflation and Bank Profits: Monetary Policy Trade-offs (INTERNATIONAL MONETARY FUND). Annex Tables 3.2 and 3.3 in Annex 3 present the full results behind this figure.*

### 2. Non-interest Income and Expense Exposures

### 2. Non-interest Income and Expense Exposures

### Direct and indirect exposure patterns
- Direct exposures to unexpected inflation in the non-interest business are small for most banks, with more variation within emerging market and developing economy banks.
- Exposures appear symmetric for emerging market and developing economy banks, whereas advanced economy banks have largely positive exposure to unexpected inflation.
- Gross exposures are typically large and vary widely across countries and banks. The interest business is exposed indirectly to both demand- and supply-driven inflation; the non-interest business (fee-based income and expenses such as salaries and rent) is exposed directly to both unexpected and, to some extent, expected components of supply-driven inflation.
- Net exposures vary across banks driven by differences in:
  - bank business models and sophistication;
  - within-country competition;
  - regulatory frameworks (including tailoring of regulations based on size, systemic importance, and complexity);
  - bank size, reliance on deposits, and size of non-interest expense relative to assets.

### Recent shifts: illustrative impact of 2021–23 inflation and policy rate changes
- Methodology: combines estimates for changes in net interest income from country-specific changes in policy rates between 2021 and 2023 and changes in non-interest income from unexpected inflation over the same period, scaled by total bank assets, separately for advanced and emerging market and developing economy banks.
- Key numerical inputs for the illustrative scenario:
  - Median increase in interest rates over this period was 4.5 percentage points.
  - Median unexpected inflation was 7 percent.
- Distributional results for bank-level income shifts (percent of assets):
  - For most banks, net gains and losses appear small: 46 percent of advanced-economy banks and 55 percent of emerging market and developing economy banks experience shifts within –0.5 and 0.5 percent of assets.
  - About 5 percent of advanced economy banks and 8 percent of emerging market and developing economy banks experience losses larger than 2 percent of assets.
  - For the average country, these banks account for 9 percent of banking system assets.
- Direct exposures to inflation in the non-interest business appear larger and more skewed—particularly for advanced economy banks—than those for the interest business.
- These shifts are sizable relative to average capital-to-asset ratios:
  - 7.5 percent in advanced economies.
  - 10.8 percent in emerging market and developing economies.
- Implication: challenges to near-term profitability could persist in the medium term if rates remain higher than usual or inflation resurges.

### Concentration and systemic risk considerations
- Meaningful losses, even if concentrated at individual banks, could lead to broader panics through information-based contagion and impairment of market values that increase run risk.
- Silicon Valley Bank (SVB) is identified as an outlier with large indirect interest-rate-business exposures via policy rates, particularly relative to other banks in advanced economies.
- Even well-hedged banking systems contain banks with significant gross exposures, calling for careful monitoring of transmission of policy rates to bank borrowers and lenders.

### Interactions between market and liquidity risk (Box 1)
- A sharp increase in interest rates can reduce market value of fixed-rate, long-dated bank assets; accounting rules may allow holding at book value, but forced sales to meet withdrawals can realize losses and weaken capital.
- Banks with uninsured liabilities and large market exposure could face runs; withdrawals that cannot be met by cash must be funded by asset sales at market prices, generating losses compared with bank equity.
- Empirical and scenario-based work:
  - Copestake, Kirti, and Liu (2023) estimate exposures for more than 1,200 banks across 47 countries using indicative losses on market value of securities and assuming larger losses on other asset classes such as loans.
  - Central scenario considered: 20 percent withdrawals in vulnerable banks.
  - Aggregate results show higher exposure in advanced economies than in emerging market and developing economies, reflecting greater use of wholesale funding and tighter cash and capital buffers.
  - Two illustrative aggregated metrics shown in Figure 1.1: equity = 3.5% (advanced economies) and equity = 0.1% (emerging market and developing economies) in the plotted scenarios.

### Conclusions and policy implications
- Conceptual summary:
  - Gross exposures can differ substantially from net exposures.
  - Both direct effects of inflation and indirect effects via policy rates matter for bank profitability.
  - Asset quality can deteriorate through direct and indirect channels, affecting borrowers and creditors.
- Empirical summary:
  - Aggregate bank profitability generally does not appear to shift significantly with inflation, reflecting active matching of income and expense exposures, particularly in the interest business.
  - Net exposures in the non-interest business are more significant in both directions across countries, though gross income and expense exposures are generally smaller than for the interest business.
- Policy recommendations:
  - Strengthen prudential regulation and supervision ex ante to minimize trade-offs between price and financial stability.
  - Heighten requirements for risk management governance within banks to support active management of gross exposures to inflation in both interest and non-interest businesses.
  - Improve transparency through better public reporting and granular auditing.
  - Use ongoing, granular risk assessments that account for differences in exposure across bank businesses, expected and unexpected inflation, and direct and indirect effects to calibrate micro- and macroprudential capital requirements.
  - Monitor specific banks particularly exposed to inflation and policy rates carefully; consider availability of central bank funding facilities to replace lost deposits as a mitigating factor.

*Source: IMF staff discussion notes — "Inflation and Bank Profits: Monetary Policy Trade-offs", 2. Non-interest Income and Expense Exposures.*

### Annex 1. Data

### sdnea2025001 - Annex 1. Data

### Data sources and sample construction
- Primary combined dataset components:
  - (1) bank balance sheets and income statements (annual, bank-level, unconsolidated except for the United States where consolidated data are used);
  - (2) macroeconomic variables; and
  - (3) policy rates.
- Bank-level financial data source: Fitch Connect.
  - Update from November 1, 2023.
- Bank sample inclusion criteria:
  - Only banks whose main business is one of: “co-operative banks”, “commercial banks”, “retail banks”, “consumer banks” and “bank holding companies”.
  - For the US, only bank holding companies are included.
  - Observations retained only if banks satisfy:
    - positive total assets, average earning assets, and total securities;
    - greater total liabilities than total deposits;
    - larger total assets than average earning assets;
    - total equity ratios larger than -15 percent of total assets;
    - total assets exceeding $100 million.
  - Banks must have at least five years of data and country years must have at least five banks.
  - For non-US countries, one of: local generally accepted accounting practices, regulatory requirements, or International Financial Reporting Standards is used; US uses regulatory accounting standards.
- Macroeconomic variables:
  - Expected and realized inflation and expected and realized real GDP growth obtained from the IMF World Economic Outlook (WEO) Database.
  - Expected inflation: the log change of the consumer price index (CPI) as predicted in the October forecast of the previous year (i.e., forecast for t + 1 in October publication of the WEO in year t).
  - Realized inflation: log differences in CPI.
  - Unexpected inflation = realized inflation − expected inflation.
  - Expected real GDP growth: log difference in GDP at constant prices taking the previous year’s October forecast.
  - Realized real GDP growth: log change in GDP at constant prices.
  - Unexpected real GDP growth = realized real GDP growth − expected real GDP growth.
  - CPI and GDP data: for inflation in t and for GDP, data confirmed in t + 2 are used because CPI data can undergo large revisions.
- Policy rates:
  - Sources: Bank for International Settlements; Haver Analytics; Bloomberg Finance L.P.; and IMF International Financial Statistics.
  - Annual policy-rate values are obtained by averaging quarterly values.
- Coverage and final regression sample:
  - Most variables available for 81 countries (28 advanced economies [AEs] and 53 emerging market and developing economies [EMDEs]).
  - Regression sample confined to 59 countries (28 AEs and 31 EMDEs) because of policy-rate availability.
  - Final annual panel: 5,496 banks in AEs and 1,149 banks in EMDEs (total 6,645 banks).
  - Unbalanced coverage from 1995 to 2022.
  - Coverage before 2000 is available for 37 countries.

### Key summary statistics (Annex Table 1.1)
- Bank statistics (means, medians, standard deviations, min, max):
  - Return on assets (ROA): Mean 0.5; Median 0.3; Standard Deviation 0.9; Min -3.6; Max 5.3.
  - Net interest margin (NIM): Mean 2.7; Median 2.4; Standard Deviation 1.7; Min 0.2; Max 13.6.
  - Net non-interest income (scaled by total assets): Mean -1.3; Median -1.2; Standard Deviation 1.4; Min -32.8; Max 155.1.
  - Loan impairment charge (scaled by total assets): Mean 0.4; Median 0.2; Standard Deviation 1; Min -27.1; Max 75.4.
  - Deposits over liabilities: Mean 88.4; Median 94.5; Standard Deviation 15.2; Min 0; Max 100.
  - Equity over assets: Mean 8.9; Median 8; Standard Deviation 5.6; Min -14.5; Max 99.3.
  - Securities over assets: Mean 20.1; Median 18.6; Standard Deviation 13.7; Min 0; Max 99.8.
- General economic indicators (means, medians, standard deviations, min, max):
  - Inflation shock: Mean 0.5; Median 0.0; Standard Deviation 4.4; Min -10.0; Max 45.2.
  - Realized inflation: Mean 4.6; Median 2.8; Standard Deviation 7.8; Min -1.4; Max 109.9.
  - Policy rate: Mean 5.4; Median 3.9; Standard Deviation 6.2; Min -0.8; Max 49.3.
  - Expected real GDP growth: Mean 3.4; Median 3.3; Standard Deviation 1.9; Min -4.1; Max 9.5.
  - Unexpected real GDP growth: Mean -0.5; Median -0.1; Standard Deviation 3.0; Min -13.2; Max 6.9.
- Note: Net non-interest income and loan impairment charge are scaled by total assets.

### Key variables and data sources (Annex Table 1.2)
- Bank statistics:
  - Return on assets (ROA): Fitch Connect.
  - Net interest margin (NIM): Fitch Connect.
  - Deposits over liabilities, Equity over assets, Securities over assets: calculated using Fitch Connect.
- General economic indicators:
  - Expected and unexpected inflation: calculated using World Economic Outlook.
  - Policy rates: BIS; Haver Analytics; Bloomberg Finance N.A.; and IMF, International Financial Statistics.
  - Expected and unexpected real GDP growth: calculated using World Economic Outlook.

### Empirical specifications (Annex 2)
- Sequential approach separates direct and indirect (through policy rates) effects of inflation using two main specifications.
- Specification (2.1) — captures the total effect of inflation:
  - Includes interest rates orthogonalized with respect to inflation; controls only for the component of interest rates unrelated to inflation.
  - General form (variables as defined in the source):
    - Y_{b,i,t} = α1 π_{i,t}^f + α2 π_{i,t-1}^f + σ1 π_{i,t}^s + σ2 π_{i,t-1}^s + β i_{i,t}^{orth} + φ GDP^f_{i,t} + χ GDP^s_{i,t} + η X_{b,t} + γ_b + γ_t + ε_{b,i,t}
  - The orthogonalized component of interest rates is obtained from a regression of the policy rate on expected and unexpected inflation, expected and unexpected real GDP growth, bank controls, and bank/time fixed effects.
- Specification (2.2) — captures the direct effect of inflation:
  - Controls for all variation in interest rates and focuses on inflation that is orthogonal to interest rates.
  - General form (variables as defined in the source):
    - Y_{b,i,t} = λ1 π_{i,t}^{orth,f} + λ2 π_{i,t-1}^{orth,f} + μ1 π_{i,t}^{orth,s} + μ2 π_{i,t-1}^{orth,s} + σ i_{i,t} + φ GDP^f_{i,t} + χ GDP^s_{i,t} + η X_{b,t} + γ_b + γ_t + ε_{b,i,t}
  - Orthogonalized components of expected and unexpected inflation (and their lags) are obtained by regressing each on policy-independent controls and fixed effects; residuals represent inflation orthogonal to policy rates.
- Definitions of variables used in regressions:
  - Y_{b,i,t}: ROA, non-interest-related subcomponents, loan impairment charges; NIM and splits of interest income and expense margins.
  - π_{i,t}^f: expected inflation.
  - π_{i,t}^s: unexpected inflation (realized − expected).
  - GDP^f_{i,t}, GDP^s_{i,t}: expected and unexpected real GDP growth.
  - X_{b,t}: bank-level controls — deposits over liabilities, equity over assets, securities over assets.
  - γ_b and γ_t: bank and time fixed effects, respectively.

### Estimation details and identification strategies
- All regressions use an unbalanced panel of 6,645 banks from 59 countries for 1995–2022.
- Standard errors:
  - Block-bootstrapped at the country level to account for generated regressors.
  - 20,000 draws are used for bootstraps.
- Separating supply- from demand-driven inflation:
  - A dummy indicating that real GDP growth and inflation surprises are of opposite signs is used to proxy for supply-driven inflation at time t.
  - The dummy is interacted with key macroeconomic variables (expected and unexpected inflation, their lags, expected and unexpected GDP growth) in both specifications to allow differential effects under supply-driven inflation.

### Annex overview: results, figures, and robustness
- Annex 3: Detailed empirical results include:
  - Annex Table 3.1. Taylor Rule Estimates (modified Taylor rule regressions on an unbalanced panel of 59 countries from 1995 to 2022; dependent variable is the nominal policy rate; standard errors clustered at time and country level).
  - Annex Table 3.2. Gross Measures of Bank Profitability (Total Effect of Inflation) — reports results for specification (2.1) using 6,645 banks in 59 countries from 1995 to 2022; dependent variables are in basis points; standard errors by clustering and country-level bootstraps (20,000 draws).
  - Annex Table 3.3. Gross Measures of Bank Profitability (Direct Effect of Inflation) — reports results for specification (2.2) with analogous reporting conventions to Table 3.2.
- Annex 4: Additional figures and decomposition analyses:
  - Annex Figure 4.1. Local Projections for Income and Expenses — local projections of interest income margin, interest expense margin, non-interest income, and non-interest expense to expected, unexpected inflation, and policy rates independent of inflation using specification (2.1); regressions weighted by 1/(number of banks in each country year); shaded area shows confidence interval at the 10 percent significance level; standard errors block-bootstrapped at the country level.
  - Annex Figure 4.2. Figure 4 Repeated within Euro Area Countries — compares total and direct effects of inflation for 17 euro area countries; controls for euro area expected and unexpected real GDP growth and excludes time fixed effects; bars are sums of contemporaneous and lagged coefficients; filled bars indicate statistical significance at the 10 percent level.
  - Annex Figure 4.3. Breakdown of Non-Interest Income — exposure of net fees and commissions, other non-interest income, and trading revenues to expected and unexpected inflation when controlling for policy-rate component orthogonal to inflation (specification (2.1)); bars show sums of contemporaneous and lagged coefficients; trading revenues shown separately due to lower coverage.
  - Annex Figure 4.4. Breakdown of Non-Interest Expense — exposure of non-interest expense components to expected and unexpected inflation controlling for policy-rate component orthogonal to inflation (specification (2.1)); bars show sums of contemporaneous and lagged coefficients.
  - Annex Figure 4.5. Quantile Regressions for Return on Assets — coefficients of expected inflation, unexpected inflation, and policy rate independent of inflation from quantile regressions on the 10th, 25th, 50th, 75th, and 90th percentiles using specification (2.1); bars show sums of contemporaneous and lagged coefficients.
  - Annex Figure 4.6. Banking System Equity Return and Inflation (1870–2016) — nominal banking sector equity returns with annual CPI inflation (source: Baron, Verner, and Xiong 2021).
  - Annex Figure 4.7. Total and Direct Exposure of Loan Impairments — compares total and direct effects of inflation on loan impairment charges (specifications (2.1) and (2.2)); bars are sums of contemporaneous and lagged coefficients.

*Source: sdnea2025001 - Annex 1. Data (Staff Discussion Note: Inflation and Bank Profits: Monetary Policy Trade-offs), IMF.*

### Annex Figure 4.8. Total and Direct Exposure of Interest and Non-Interest Expense to Inflation by Source

### Annex Figure 4.8. Total and Direct Exposure of Interest and Non-Interest Expense to Inflation by Source

### Interest Expense Exposures to Inflation
- The figure compares the total effects of inflation, which control for policy rates orthogonal to inflation, and direct effects of inflation, which focus on inflation orthogonal to policy rates. See specification (2.1) and (2.2) defined in Annex 2.
- Each bar shows the effect of inflation driven by demand or supply, in which inflation is interacted with a dummy that indicates the presence of supply-driven inflation at t.
- Bars are sums of contemporaneous and lagged coefficients.
- Filled bars indicate statistically significant coefficients at the 10 percent level, while unfilled bars indicate statistically insignificant coefficients.
- Regressions are weighted by 1/(number of banks in each country year).
- Data sources: Bank for International Settlements; Bloomberg Finance L.P.; Fitch Connect; Haver Analytics; and IMF staff analysis.

### Non-Interest Expense Exposures to Inflation
- Panel structure and notation mirror the interest expense specifications: total effects (controlling for policy rates) and direct effects (inflation orthogonal to policy rates), using specification (2.1) and (2.2) in Annex 2.
- Bars show demand-driven and supply-driven inflation effects; filled bars denote statistical significance at the 10 percent level.
- Regressions are weighted by 1/(number of banks in each country year).
- Data sources: Bank for International Settlements; Bloomberg Finance L.P.; Fitch Connect; Haver Analytics; and IMF staff analysis.

### Cross-Country Heterogeneity (Annex 5 highlights)
- Panel and sample details:
  - Both specifications include country interactions with an unbalanced panel sample of 59 countries from 1995 to 2022.
  - The scatter plot depicts the coefficient estimated for each country. The regressions are weighted by 1/(number of banks in each country and year).
  - Data labels use International Organization for Standardization country codes.
- Interest Expense (Annex Figure 5.1, Panel 1):
  - Panel 1 shows the coefficient of expected inflation (x-axis) and policy rate (y-axis) on interest expense, using specification (2.1) for the former and specification (2.2) for the latter, capturing the total effect of each, respectively.
  - The correlation between the x-axis and y-axis is 0.31 and significant at the 1.7 percent significance level.
  - Japan was omitted for readability.
- Interest Income (Annex Figure 5.1, Panel 2):
  - Panel 2 shows the coefficient of inflation expectation (x-axis) and policy rates (y-axis) on interest income, using specification (2.1) for the former and specification (2.2) for the latter, capturing the total effect of each, respectively.
  - The correlation between the x-axis and y-axis is 0.23 and significant at the 8 percent level.
  - Japan, Germany, and Uruguay were omitted for readability.

### Interest Expense Pass-Through and Competition (Annex Figure 5.2)
- Definitions and methodology:
  - Interest expense pass-through is defined as the coefficient of the policy rate on interest expense using specification (2.2) defined in Annex 2, but including country interactions using the same unbalanced panel of 59 countries from 1995 to 2022.
  - The regressions are weighted by 1/(number of banks in each country and year).
- Panel 1:
  - Shows interest expense pass-through and composition of foreign liabilities.
  - Banks’ foreign liabilities are shown as percent of GDP on an ultimate risk basis.
- Panel 2:
  - Shows interest expense pass-through and implied liability spreads at the country level.
  - Liability spread is calculated by subtracting interest expense margin from the policy rate.
- Omitted outliers for readability: Luxembourg, Ireland, and Japan.
- Data sources: Bank for International Settlements; Bloomberg Finance L.P.; Fitch Connect; Haver Analytics; and IMF staff analysis.

### Net Negative Exposure to Unexpected Inflation and Bank Characteristics (Annex Figure 5.3)
- Metric definitions:
  - Net negative exposure is the difference between bank-specific income and expense pass-throughs for non-interest business on unexpected inflation, based on specification (2.2) defined in Annex 2.
  - The range is defined as 1.5 times the interquartile range of the bank-level data for each country.
- Panels and correlations:
  - Panel 1 (Variation in Net Negative Exposure and Bank Size) shows a correlation of 0.5 with a 0 p value.
  - Panel 2 (Variation in Net Negative Exposure and Deposit Reliance) shows a correlation of 0.26 with a 0.025 p value.
  - Panel 3 (Variation in Net Negative Exposure and Non-interest Expenses) shows a correlation of 0.3 with a 0.02 p value.
- Regressions and data labeling:
  - The figure shows the within-country variation or range in net negative exposure against variation or range in bank size, deposit reliance, and size of non-interest expenses.
  - Data labels use International Organization for Standardization country codes.
- Data sources: Bank for International Settlements; Bloomberg Finance L.P.; Fitch Connect; Haver Analytics; and IMF staff analysis.

*Inflation and Bank Profits: Monetary Policy Trade-offs — Staff Discussion Note No. SDN/2025/001*

---


_Source: https://www.imf.org/-/media/files/publications/sdn/2025/english/sdnea2025001.pdf_