## CHAPTER 2 A NEw LOOk AT GLOBAL BANkING VuLNERABILITIES

## Source details

**Canonical URL:** [CHAPTER 2 A NEw LOOk AT GLOBAL BANkING VuLNERABILITIES](https://www.imf.org/-/media/files/publications/gfsr/2023/october/english/ch2.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/gfsr/2023/october/english/ch2.pdf.md)
- [Structured JSON version](/-/media/files/publications/gfsr/2023/october/english/ch2.pdf.json)

---

### Scope and methodology
- Purpose: fresh assessment of global banking vulnerabilities in an environment of still-elevated inflation and high interest rates using publicly available data and uniform methods across regions.
- Two complementary approaches:
  - Enhanced global stress test covering nearly 900 banks across 29 countries.
  - Key risk indicators (KRIs) for approximately 350 of the world’s largest publicly traded individual banks using short-term consensus analyst forecasts for real-time surveillance.
- Enhancements to the global stress test:
  - Expanded sample (nearly 900 banks versus 260 banks in the previous exercise).
  - Satellite econometric models linking macro scenarios to banks’ income sources: net interest income, valuation losses on bonds, and loan loss provisions.
  - Distinction between marked-to-market versus held-to-maturity securities.
  - Bank-by-bank betas for pass-through rates where possible.
  - New liquidity-to-solvency channel via illustrative reverse stress tests with hypothetical deposit run-off rates.
- Weak-bank identification criteria (conservative):
  - CET1 ratio falls below 7 percent (Basel minimum of 4.5 percent plus a capital conservation buffer of 2.5 percent), plus buffers for G-SIBs where applicable; OR
  - CET1 ratio at its lowest point over the stress test horizon (2023–25) represents a decrease of more than 5 percentage points from the stress test’s starting point of 2022, excluding banks with more than a 30 percent CET1 ratio.

### Key findings — baseline scenario (October 2023 World Economic Outlook baseline)
- Global banking system:
  - Remains broadly resilient under the baseline.
  - Many banks in advanced economies show significant capital losses driven largely by mark-to-market losses on securities holdings in a higher-for-longer interest rate environment, and by loan losses.
- Geographic concentration of stress:
  - United States and Europe flagged by KRIs as having the greatest levels of stress as of the end of March 2023.
  - In the United States, capital losses are concentrated in regional banks, consistent with the March 2023 turmoil.
  - Projections of KRIs to the end of December 2023 point to a substantial group of smaller banks at risk in the United States; elsewhere, risks concentrated in Asia, China, and Europe due to lower expected earnings and depressed price-to-book ratios.

### Key findings — adverse scenario (severe stagflation)
- Scenario features:
  - Global economy contracts by about 2 percent in the first year of the scenario (2023).
  - Peak global policy rate shock, over the baseline, is about 160 basis points.
  - Scenario derived from a structural macrofinancial model for 33 countries; more persistent inflation driven primarily by supply shocks and stronger monetary tightening.
  - Term premiums increase more in emerging market economies than in advanced economies; emerging markets experience steeper real GDP shocks.
- Outcomes:
  - Significant capital losses across a wide set of banks, including several systemically important institutions in China, Europe, and the United States.

### Liquidity-to-solvency channel and deposit runs
- Reverse stress-test approach used to illustrate liquidity-to-solvency interactions.
- Example simulation: deposit run-off of 25 percent at the end of 2023 used to assess additional capital impact.
- Central bank access assumption:
  - Banks pledge securities held to maturity with the central bank at penalty rates of 150 basis points above the adverse-scenario short-term rates when facilities are available (noted usual range between 100 and 300 basis points; could be zero in systemic stress scenarios).
- Key quantified results:
  - The liquidity-to-solvency feedback in the adverse scenario adds 10 basis points to the capital decline overall.
  - Under the assumption of access to central bank facilities at 150 basis points over policy rates, capital losses across regions at 25 percent deposit runs are at most 13 basis points (via annualized increases in funding costs).
  - At 25 percent runoffs about 40 banks lose more than 1 percentage point CET1 ratio; if facility cost doubles to 300 basis points, the number increases to 56.
  - If central bank facilities are not available, banks would need to sell HTM securities, realize marked-to-market losses, and deplete capital—leading to a much larger number of weak banks.

### Key Risk Indicators (KRIs) — framework and predictive value
- Objective: real-time monitor of forward-looking risks using consensus analyst forecasts and market-based pricing indicators for ~350 banks.
- Data and scope:
  - Data set encompasses more than 375 banks from 43 jurisdictions and includes 28 of the 30 G-SIBs.
  - Regional total-assets and bank counts examples:
    - Asia: Total assets: $18 trillion; Number of banks: 63
    - China: Total assets: $35 trillion; Number of banks: 18
    - Europe: Total assets: $30 trillion; Number of banks: 63
    - Latin America: Total assets: $2 trillion; Number of banks: 16
    - North America: Total assets: $29 trillion; Number of banks: 196
- Framework design:
  - Combines CAMELS supervisory framework with market-based metrics and IMF Financial Soundness Indicators.
  - Focuses on five observable risk dimensions: capital adequacy, asset quality, earnings, liquidity, and market-based sensitivity (management performance excluded).
  - Uses 12 key risk indicators in total.
- Monitoring methodology:
  - Two-stage process: threshold breaches per indicator (accounting for regional differences), then aggregation to identify vulnerability across dimensions.
- Validation:
  - KRIs have predictive power for bank stress and acute stress events (large stock declines, deposit outflows).
  - Historical validation: as early as Q4 2022, multiple KRIs flagged the three US regional banks and the Swiss G-SIB that failed during the March 2023 turmoil.
  - Regression result: coefficient of about –0.7 (highly statistically significant), suggesting every increase of one KRI flag is associated with a fall of 0.7 percentage point in the Tier 1 capital ratio in the GST adverse scenario.

### Evolution of flagged banks and monitoring-list sizes
- Time series behavior:
  - Spiked dramatically with the onset of COVID-19; fell sharply in 2021; climbed to another peak before March 2023 turmoil.
- Specific counts and assets:
  - 2023:Q2 — 85 banks with $26 trillion in total assets on the monitoring list (breaches in at least three KRI dimensions).
  - Projected 2023:Q3 — 80 banks with $21 trillion in assets.
  - Projected 2023:Q4 — 82 banks with $25 trillion in assets.
  - 2023:Q4 — 25 banks flagged on four or more risk dimensions, with $9 trillion in combined assets.
  - Historical peak in 2020: more than 200 banks on the monitoring list in Q1–Q2 2020.

### Number of weak banks, asset shares, and distribution
- Baseline scenario:
  - 55 banks with more than $5.5 trillion in assets see capital falling either below 7 percent or by more than 5 percentage points in 2023.
  - These 55 banks represent 4 percent of global bank assets.
- Adverse scenario:
  - 215 banks flagged as weak, accounting for 42 percent of global banking assets.
  - If criterion limited to banks with capital falling below 7 percent, the share would be 36 percent of global bank assets.
  - A fifth of G-SIB assets would be weak by the end of the stress-testing horizon in 2025.
- Geographic patterns:
  - Weak banks include many in Europe, some G-SIBs (including Credit Suisse), and subsidiaries; weak banks are spread across countries and sizes in Europe and concentrated in small banks in emerging markets and China.

### Drivers of capital changes (contributions to CET1 ratio change in 2023)
- Major components:
  - Starting CET1; Loan loss provisions (Loan loss); Net interest income (NII); Fee income (Fee); Valuation (Valuation); Operating expenses (Expense); Liquidity feedback (Liquidity feedback); Tax (Tax); Dividends (Dividends); Risk-weighted assets adjustments (RWA).
- 2023 adverse vs baseline:
  - Valuation losses contribute 1.7 percentage points to the decline in the CET1 ratio relative to the baseline.
  - Loan losses add nearly another 1 percentage point to the CET1 decline.
  - Loan losses dominate in the medium term in both baseline and adverse scenarios.
- Baseline:
  - Projected capital in the global banking system remains about 12.7 percent of risk-weighted assets in 2023 when the policy rate shock peaks and improves over the projection horizon.
- Adverse scenario path:
  - CET1 ratio troughs in 2024 and improves to 10.8 percent in 2025.

### Characteristics of weak banks and interest-margin vulnerabilities
- Common features of weak banks (baseline and adverse):
  - Less profitable (lower return on assets).
  - Net interest margins adversely affected by higher interest rates (weak banks have much lower NIM betas than nonweak banks).
  - High loan growth in the preceding two years.
  - Relatively low price-to-book ratios and very high market leverage.
- Additional distinguishing factors in the adverse scenario:
  - Lower net interest margins in 2022.
  - Weaker capitalization (lower book leverage ratios).
  - Higher share of bonds in total assets.
- Global interest-margin vulnerability:
  - More than 40 percent of banks stand to lose net interest income; concentrated especially in advanced economies outside the United States.
  - US banks exhibit exceptionally high interest income betas and emerge as particularly strong in the NIM analysis.
- Expense vs income sensitivity:
  - Banks with higher “expense betas” relative to “income betas” are at greater risk of losing net interest income when interest rates rise.
  - Expense betas are small at first but increase over time.

### Interest Income and Expense Betas; Interest Rate Margins
- Beta interpretation: a beta value of 0.5 means borrowing interest rates rise by 50 basis points when short-term rates rise by 100 basis points.
- Emerging markets: majority of banks have higher interest income betas than interest expense betas.
- Interest rate margins in 2022:
  - Emerging markets: in excess of 5 percent.
  - Advanced economies: about 2 percent.

### Vulnerabilities to bond valuation losses
- Portfolio facts:
  - Almost one-quarter of global bank assets are invested in securities, with about half in held-to-maturity securities.
  - Securities constitute from nearly 25 percent to about 30 percent of total bank assets in Brazil, China, India, Japan, Mexico, and the United States.
  - About 40 percent of banks’ positions are hedged on average.
- Baseline:
  - Marked-to-market bond portfolios generally suffer moderate valuation losses; capital ratios could fall significantly for only about 2 percent of banks.
- Adverse scenario:
  - Valuation losses significant; 11 percent of banks vulnerable to significant declines in capital from this channel (if part of bond exposures considered hedged).
  - If exposures are not considered hedged, about a quarter of banks would be deemed vulnerable.
- Drivers: higher share of HFT and AFS securities, longer durations, greater increases in yield curves.
- Advanced-economy exposure: a higher share of banks in advanced economies exposed to valuation losses; among advanced-economy outliers, German banks are among the most affected.

### Vulnerabilities to loan defaults
- Mechanisms:
  - Increases in interest rates and declines in economic growth drive loan defaults.
- Dynamics in the adverse scenario:
  - Advanced-economy banks: loan loss provisions rise initially due to increases in real interest rates; later unemployment effects dominate.
  - Emerging-market banks: overall economic growth matters more for loan performance.
- Loan composition:
  - Advanced economies: higher shares of mortgage and consumer loans.
  - Emerging markets: lend relatively more to firms.
- Implication: unemployment rate matters more for credit performance in advanced-economy banks; GDP growth matters more in emerging-market banks.

### Vulnerabilities to interactions between liquidity and solvency
- Liquidity buffers:
  - Most banks have enough liquid assets to sustain deposit outflows of 10 percent without having to repo or sell HTM bonds.
  - Banks in emerging markets can sustain slightly higher deposit outflows (20 percent) than those in advanced economies (5–10 percent) without needing to sell or pledge HTM securities.
- At outflows of 15–25 percent, an exponentially high share of banks would need to use HTM securities.
- Specific runoff implications:
  - For a 15 percent runoff: sale of HTM bonds generates moderate losses across regions when central bank facilities are not available.
  - For a 25 percent runoff: CET1 ratios would drop substantially in several banks, including Silicon Valley Bank and First Republic Bank, if central bank facilities are not available; losses multiply rapidly as outflows increase from 25 to 35 percent.
- If a bank runs out of eligible collateral, central banks assumed to extend emergency liquidity assistance by expanding accepted collateral types or providing unsecured loans at the same interest rates.
- Regulatory liquidity coverage ratio context: one-month runoff rates for deposits provided (insured retail demand deposits, 3–5 percent; less stable retail deposits, 10 percent; term deposits, 0 percent; small business deposits, 5–10 percent; other nonfinancial firms and sovereigns, 20 percent if insured and 40 percent if not).
- When central bank facilities are not available, additional capital losses are contained at a regional level up to 25 basis points with a 25 percent runoff rate; however, several banks could lose more than 150 basis points in additional Tier 1 capital across regions because of the impact of HTM sales on capital.

### Stress-test scenario severity and caveats
- Adverse scenario corresponds to 3½ standard deviations from historical means of global GDP growth.
- The scenario is severe and illustrative; other severity degrees could be chosen by supervisors.
- Caveats:
  - Simplifying assumptions due to limited publicly available bank-level data on duration and hedging.
  - Supervisors may have access to more detailed bank-level data to avoid some simplifying assumptions.
  - Sensitivity analyses for the Chinese banking system presented in Online Annex 2.1.

### Regional outcomes and differences
- Emerging markets (excluding China): particularly resilient in 2023; helped by high initial capital ratios, robust economic growth, and sizable net interest income.
- China:
  - Starts with one of the lowest capital levels among regions.
  - Experiences the highest decline in CET1 ratio of 3.9 percentage points in the adverse scenario and ends slightly above the 7 percent minimum.
  - Sensitivity: if the unemployment rate shock were halved in all three years, the share of Chinese bank assets considered weak in the adverse scenario would fall from about 62 to 55 percent.
- Euro area: relatively steep decline in capital ratio through the trough year, comparable to European Banking Authority (2023) stress tests, but ends with a relatively high level owing to a healthy starting point.
- United States: modest decline in the adverse scenario mainly due to gains in net interest income; settles at a level similar to average global levels.

### Using KRIs to monitor emerging vulnerabilities and KRI-driven policy implications
- KRI framework aims to identify banks meriting closer examination; designed to have a high level of type I errors given rarity of failures.
- Historical performance:
  - KRI framework flagged, in Q4 2022, the four banks that failed in March 2023 (Credit Suisse, Silicon Valley Bank, Signature Bank, First Republic) across different KRI dimensions.
- Policy implications and recommendations:
  - Sharpen analytical tools and closely monitor KRIs for real-time surveillance.
  - Make stress tests more stringent and granular, including coverage for smaller banks.
  - Make supervisory practices more intrusive and implement corrective actions in a more timely and effective manner.
  - Tighten prudential standards for capital held against interest rate risk.
  - Banks should prepare to access central bank facilities to mitigate potential capital losses from selling HTM securities under stress.
  - Enhance commercial banks’ preparedness to use eligible collateral and access central bank facilities; improve authority communication on availability and usage.
  - Strengthen regulations and crisis management frameworks, including deposit insurance and resolution authorities; progress on too-big-to-fail reforms and public sector liquidity backstops is needed.
  - Expand stress-test samples and methodologies, consider more severe but plausible adverse scenarios, and leverage more timely and granular data.

### Timing, congruence of GST and KRI frameworks, and historical run experience
- Liquidity stress began increasing as early as June 2022 (evidence: deposit outflows, higher loans-to-deposits, lower deposit shares).
- Samples overlap: 168 banks are in both the global stress test and KRI framework samples.
- GST–KRI congruence:
  - Among 47 banks flagged on two KRI dimensions as of 2022:Q4, Tier 1 ratio declines about 2 percentage points on average under the GST adverse scenario.
  - Among 12 banks flagged on four KRI dimensions, average Tier 1 impact about –4 percentage points.
  - Banks with three or four flagged KRIs: worst quintile had an average decline of Tier 1 capital ratio of more than 8 percentage points.
- Historical rapid online runs referenced: Northern Rock (2007), Landsbanki (2008), Continental Illinois (1984).
- Panel evidence: United States, 1930–33 — average of 67 failed banks (panel note).

*Source: IMF Global Financial Stability Report chapter content (October 2023).*

### Chapter 2 at a Glance

### Chapter 2 at a Glance

### Scope and methodology
- Purpose: fresh assessment of global banking vulnerabilities in an environment of still-elevated inflation and high interest rates using publicly available data and uniform methods across regions.
- Two complementary approaches:
  - Enhanced global stress test covering nearly 900 banks across 29 countries.
  - Key risk indicators (KRIs) for approximately 350 of the world’s largest publicly traded individual banks using short-term consensus analyst forecasts for real-time surveillance.
- Enhancements to the global stress test:
  - Expanded sample (nearly 900 banks versus 260 banks in the previous exercise).
  - Satellite econometric models linking macro scenarios to banks’ income sources: net interest income, valuation losses on bonds, and loan loss provisions.
  - Distinction between marked-to-market versus held-to-maturity securities.
  - Bank-by-bank betas for pass-through rates where possible.
  - New liquidity-to-solvency channel via illustrative reverse stress tests with hypothetical deposit run-off rates.
- Weak-bank identification criteria (conservative):
  - CET1 ratio falls below 7 percent (Basel minimum of 4.5 percent plus a capital conservation buffer of 2.5 percent), plus buffers for G-SIBs where applicable; OR
  - CET1 ratio at its lowest point over the stress test horizon (2023–25) represents a decrease of more than 5 percentage points from the stress test’s starting point of 2022, excluding banks with more than a 30 percent CET1 ratio.

### Key findings — baseline scenario (October 2023 World Economic Outlook baseline)
- Global banking system:
  - Remains broadly resilient under the baseline.
  - Many banks in advanced economies show significant capital losses driven largely by mark-to-market losses on securities holdings in a higher-for-longer interest rate environment, and by loan losses.
- Geographic concentration of stress:
  - United States and Europe currently flagged by KRIs as having the greatest levels of stress as of the end of March 2023.
  - In the United States, capital losses are concentrated in regional banks, consistent with the March 2023 turmoil.
  - Projections of KRIs to the end of December 2023 using analyst forecasts point to a substantial group of smaller banks at risk in the United States; elsewhere, risks concentrated in Asia, China, and Europe due to lower expected earnings and depressed price-to-book ratios.

### Key findings — adverse scenario (severe stagflation)
- Adverse scenario outcomes:
  - Significant capital losses identified across a wide set of banks, including several systemically important institutions in China, Europe, and the United States.
  - Stress test adverse scenario assumes global economy contracts by about 2 percent in the first year of the scenario (2023).
  - Peak global policy rate shock, over the baseline, is about 160 basis points.
  - Adverse scenario derived from a structural macrofinancial model for 33 countries; features more persistent inflation driven primarily by supply shocks and stronger monetary tightening.
  - In the adverse scenario, term premiums increase more in emerging market economies than in advanced economies; emerging markets experience steeper real GDP shocks.

### Liquidity-to-solvency channel and deposit runs
- Reverse stress-test approach used to illustrate liquidity-to-solvency interactions because depositor behavior is hard to pin down with available data.
- Example simulation: deposit run-off of 25 percent at the end of 2023 used to assess additional capital impact.
- Assumption on central bank access:
  - Banks would need to pledge securities held to maturity with the central bank at penalty rates of 150 basis points above the adverse-scenario short-term rates, under the assumption that central bank facilities are available.
  - The penalty rate usually ranges from 100 to 300 basis points above policy rates and could sometimes be zero in certain systemic stress scenarios.

### Key Risk Indicators (KRIs)
- Purpose: real-time monitor of forward-looking risks using consensus analyst forecasts on future bank balance sheet, valuation, and profitability metrics for ~350 banks.
- KRIs have predictive power for bank stress and acute stress events (large stock declines, deposit outflows).
- Historical validation: as early as Q4 2022, multiple KRIs flagged the three US regional banks and the Swiss G-SIB that failed during the March 2023 turmoil.
- Concordance: banks flagged as outliers on multiple KRIs are more likely to experience large capital losses under the stress test’s adverse scenario.

### Observations on depositor and investor behavior
- Recent episodes show depositor behavior and investor sentiment can amplify bank stress:
  - Investors rapidly penalize banks with low price-to-book ratios and low profitability even when regulatory capital and liquidity appear adequate.
  - A group of weak banks, even if not individually systemic, can pose financial stability risks.

### Policy recommendations
- Sharpen analytical tools for risk assessments and closely monitor relevant market metrics (KRIs) for real-time surveillance.
- Make stress tests more stringent and granular, including coverage for smaller banks.
- Make supervisory practices more intrusive and implement corrective actions in a more timely and effective manner.
- Tighten prudential standards for capital held against interest rate risk.
- Banks should prepare to access central bank facilities to substantially mitigate potential capital losses from selling held-to-maturity securities under stress.

*Italic: Source — Chapter 2 at a Glance, ch2 - Chapter 2 at a Glance (PDF).*

### 1. Global Real GDP

### ch2 - 1. Global Real GDP

### Stress-test scenarios and assumptions
- Adverse scenario assumes bank runs at the end of 2023, generating liquidity-to-solvency interaction channels; banks pledge held-to-maturity securities with the central bank after selling available-for-sale and held-for-trading portfolios.
- When central bank facilities are available, banks can pledge securities at a moderate penalty rate taken as 150 basis points above policy rate (noted usual range between 100 and 300 basis points) for a year.
- Panels 3 and 4 show the maximum difference between adverse and baseline over the three-year stress-testing horizon.
- The liquidity-to-solvency interaction feedback is quantified as the maximum difference over a three-year horizon in basis points for short-term interest rates and in percent for real GDP adverse deviations.

### Overall global results
- Baseline: projected capital in the global banking system remains about 12.7 percent of risk-weighted assets in 2023 when the policy rate shock peaks and improves over the projection horizon.
- Adverse scenario: CET1 ratio troughs in 2024 and improves to 10.8 percent in 2025.
- In 2023 under the adverse scenario:
  - Valuation losses contribute 1.7 percentage points to the decline in the CET1 ratio relative to the baseline.
  - Loan losses add nearly another 1 percentage point to the CET1 decline.
- The liquidity-to-solvency feedback in the adverse scenario adds 10 basis points to the capital decline overall.
- Loan losses dominate in the medium term in both baseline and adverse scenarios.

### Regional outcomes and differences
- Emerging markets (excluding China) are particularly resilient in 2023; over the medium term they are helped by high initial capital ratios, robust economic growth, and sizable net interest income.
- China:
  - Starts with one of the lowest capital levels among regions.
  - Experiences the highest decline in CET1 ratio of 3.9 percentage points in the adverse scenario and ends slightly above the 7 percent minimum.
  - Sensitivity analysis: if the unemployment rate shock were halved in all three years, the share of Chinese bank assets considered weak in the adverse scenario would fall from about 62 to 55 percent.
- Euro area: experiences a relatively steep decline in capital ratio through the trough year, comparable to European Banking Authority (2023) stress tests, but ends with a relatively high level owing to a healthy starting point.
- United States: modest decline in the adverse scenario mainly due to gains in net interest income; settles at a level similar to average global levels.

### Weak banks: counts, assets, and distribution
- Under the baseline scenario:
  - 55 banks with more than $5.5 trillion in assets see capital falling either below 7 percent or by more than 5 percentage points in 2023.
  - These 55 banks represent 4 percent of global bank assets.
- Under the adverse scenario:
  - 215 banks are flagged as weak, accounting for 42 percent of global banking assets.
  - If the criterion is limited to banks with capital falling below 7 percent, the share would be 36 percent of global bank assets.
  - A fifth of G-SIB assets would be weak by the end of the stress-testing horizon in 2025.
- Weak banks include many in Europe, some G-SIBs (including Credit Suisse), and subsidiaries; weak banks are spread across countries and sizes in Europe and concentrated in small banks in emerging markets and China.

### Drivers of capital changes (contributions to CET1 ratio change in 2023)
- Major contributing components:
  - Starting CET1
  - Loan loss provisions (Loan loss)
  - Net interest income (NII)
  - Fee income (Fee)
  - Valuation (Valuation)
  - Operating expenses (Expense), assumed constant as a share of risk-weighted assets
  - Liquidity feedback (Liquidity feedback)
  - Tax (Tax)
  - Dividends (Dividends)
  - Risk-weighted assets adjustments (RWA)
- In 2023 baseline relative to starting point, net interest income improvements help increase capital on average; valuation losses and loan loss provisions are the dominant detractors in the adverse scenario.
- Regionally in 2023 adverse vs baseline:
  - Advanced economies and China: valuation losses dominate the decline, with loan losses the second-biggest contributor.
  - United States and other advanced economies: net interest income helps modestly in 2023.
  - Other regions: net interest income mildly hurts in 2023.

### Liquidity-to-solvency interactions and deposit runs
- A 25 percent deposit run would cause CET1 ratios of several banks in advanced economies to decline by almost one additional percentage point owing to higher expenses related to the use of central bank deposit facilities.
- The feedback from liquidity to solvency adds only 10 basis points to the capital decline overall in the adverse scenario, highlighting the relatively small cost of accessing central bank facilities during bank runs when such facilities are available.
- If central bank facilities were not available in deposit runs, banks would need to sell held-to-maturity securities, realize marked-to-market losses, and deplete capital—leading to a much larger number of weak banks.

### Characteristics of weak banks
- Common features of weak banks (baseline and adverse):
  - Less profitable (lower return on assets).
  - Net interest margins adversely affected by higher interest rates (weak banks have much lower NIM betas than nonweak banks).
  - High loan growth in the preceding two years.
  - Relatively low price-to-book ratios and very high market leverage.
- Additional distinguishing factors for weak banks in the adverse scenario:
  - Lower net interest margins in 2022.
  - Weaker capitalization (lower book leverage ratios).
  - Higher share of bonds in total assets.
- Global result on interest-margin vulnerability:
  - More than 40 percent of banks stand to lose net interest income; this is concentrated especially in advanced economies outside the United States.
  - US banks exhibit exceptionally high interest income betas and emerge as particularly strong in the NIM analysis.
- Expense vs income sensitivity:
  - Banks with higher “expense betas” relative to “income betas” (below the 45-degree line in Figure 2.7) are at greater risk of losing net interest income when interest rates rise.
  - Expense betas are small at first but increase over time, possibly because depositors seek higher returns within the same bank from other financial instruments.

*Sources: IMF, World Economic Outlook; Vitek 2018; and IMF staff calculations.*

### CHAPTER 2 A NEw LOOk AT GLOBAL BANkING VuLNERABILITIES

### CHAPTER 2 A NEw LOOk AT GLOBAL BANkING VuLNERABILITIES

### Interest Income and Expense Betas; Interest Rate Margins
- A beta value of 0.5 means borrowing interest rates rise by 50 basis points when short-term rates rise by 100 basis points (figure note).
- In emerging markets, the majority of banks have higher interest income betas than interest expense betas.
- Interest rate margins in emerging markets were in excess of 5 percent in 2022, compared with about 2 percent in advanced economies; the higher margins help emerging markets absorb losses.

### Vulnerabilities to Bond Valuation Losses
- Almost one-quarter of global bank assets are invested in securities, with about half in held-to-maturity securities.
- Securities constitute from nearly 25 percent to about 30 percent of total bank assets in Brazil, China, India, Japan, Mexico, and the United States.
- Banks in emerging markets tend to:
  - have higher exposures to securities than those in advanced economies, and
  - keep more of their securities as held-to-maturity (HTM) at book value rather than marked to market (held for trading (HFT) and available for sale (AFS)).
- About 40 percent of banks’ positions are hedged on average.
- Baseline scenario:
  - Banks’ marked-to-market bond portfolios generally suffer moderate valuation losses.
  - Overall, capital ratios could fall significantly for only about 2 percent of banks (deemed vulnerable to this channel).
- Adverse scenario (part of bond exposures considered hedged):
  - Valuation losses are significant; 11 percent of banks are vulnerable to significant declines in capital from this channel.
  - If exposures are not considered hedged, about a quarter of banks would be deemed vulnerable.
- Drivers of vulnerability: higher share of HFT and AFS securities, longer durations, greater increases in yield curves.
- A higher share of banks in advanced economies than in emerging markets is exposed to valuation losses due to exposure, duration, and interest rate shocks.
- Among advanced-economy outliers, German banks are among the most affected (relatively longer durations and large policy rate shocks in the stagflationary scenario).

### Vulnerabilities to Loan Defaults
- Increases in interest rates and declines in economic growth drive loan defaults.
- In the adverse scenario:
  - For banks in advanced economies, loan loss provisions rise initially because of increases in real interest rates; later, unemployment effects dominate as the interest rate shock wanes.
  - For banks in emerging markets, overall economic growth matters more for loan performance.
- Loan composition differences:
  - Banks in advanced economies tend to have higher shares of mortgage and consumer loans.
  - Banks in emerging markets lend relatively more to firms.
- Implication: Unemployment rate matters more for credit performance in advanced-economy banks, whereas GDP growth matters more in emerging-market banks.

### Vulnerabilities to Interactions between Liquidity and Solvency
- Most banks have enough liquid assets to sustain deposit outflows of 10 percent without having to repo or sell HTM bonds.
- Banks in emerging markets can sustain slightly higher deposit outflows (20 percent) than those in advanced economies (5–10 percent) without needing to sell or pledge HTM securities.
- At outflows of 15–25 percent, an exponentially high share of banks would need to use HTM securities to address liquidity needs.
- For a deposit runoff rate of 15 percent:
  - The sale of HTM bonds would generate moderate losses across regions when central bank facilities are not available.
- At 25 percent runoff:
  - Common Equity Tier 1 (CET1) ratios would drop substantially in several banks, including Silicon Valley Bank and First Republic Bank, if central bank facilities are not available.
  - Losses multiply rapidly as deposit outflows increase from 25 to 35 percent.
- Central bank facilities mitigate losses noticeably:
  - Under the assumption that banks can access central bank facilities and pledge HTM securities at 150 basis points over policy rates, capital losses across regions at 25 percent deposit runs are at most 13 basis points (occurring via annualized increases in funding costs).
  - Results vary across banks; at 25 percent runoffs about 40 banks lose more than 1 percentage point CET1 ratio or more.
  - If the cost of facilities doubles to 300 basis points, the number of banks losing more than 1 percentage point CET1 increases to 56.
- If a bank runs out of eligible collateral, the scenario assumes central banks extend emergency liquidity assistance by expanding accepted collateral types or, if needed, providing unsecured loans at the same interest rates.
- Regulatory liquidity coverage ratio (for context in runoff assumptions) uses one-month runoff rates for deposits: insured retail demand deposits, 3–5 percent; less stable retail deposits, 10 percent; term deposits, 0 percent (except maturing contracts); small business deposits, 5–10 percent; other nonfinancial firms and sovereigns, 20 percent if insured and 40 percent if not.
- When central bank facilities are not available, additional capital losses are contained at a regional level up to 25 basis points with a 25 percent runoff rate; however, several banks could lose more than 150 basis points in additional Tier 1 capital across regions because of the impact of HTM sales on capital.

### Stress-Test Scenario Severity and Caveats
- The adverse scenario coordinated through the global dynamic stochastic general equilibrium model corresponds to 3½ standard deviations from historical means of global GDP growth.
- The scenario is severe and meant to be illustrative; other degrees of severity could be chosen by supervisors.
- Caveats:
  - Simplifying assumptions are used because publicly available bank-level data on duration and hedging are limited.
  - Supervisors may have access to more detailed bank-level data to avoid some simplifying assumptions.
  - Sensitivity analyses for the Chinese banking system (which lacks historical precedent) are presented in Online Annex 2.1.

### Using Key Risk Indicators (KRIs) to Monitor Emerging Vulnerabilities
- Objective: develop a forward-looking tool for monitoring vulnerabilities in publicly traded individual banks using financial and aggregate consensus analyst forecasts and market-based pricing indicators.
- Data and scope:
  - An extensive new data set encompasses more than 375 banks from 43 jurisdictions.
  - The data set includes 28 of the 30 G-SIBs as identified by the Financial Stability Board.
  - Regional data coverage examples:
    - Asia: Total assets: $18 trillion; Number of banks: 63
    - China: Total assets: $35 trillion; Number of banks: 18
    - Europe: Total assets: $30 trillion; Number of banks: 63
    - Latin America: Total assets: $2 trillion; Number of banks: 16
    - North America: Total assets: $29 trillion; Number of banks: 196
- Framework design:
  - Combines CAMELS supervisory framework with market-based metrics and IMF Financial Soundness Indicators.
  - Focuses on five observable risk dimensions: capital adequacy, asset quality, earnings, liquidity, and market-based sensitivity (management performance excluded due to limited direct measurement).
  - Uses 12 key risk indicators in total to measure these five dimensions.
  - Incorporates consensus analyst forecasts (third and fourth quarters of 2023, and second quarter of 2023 where needed) and market pricing indicators to capture forward-looking elements.
- Monitoring methodology:
  - Two-stage process:
    1. For each risk indicator, banks’ values are highlighted if they exceed calibrated thresholds (thresholds account for regional structural differences).
    2. Banks are identified as potentially vulnerable within a risk dimension if one or more indicators in that dimension are outliers; banks are flagged if they are vulnerable across a majority of dimensions.
- Validation:
  - The framework includes econometric analysis showing predictive power for previous stress events and anticipation of potential capital shortfalls revealed by the global stress test.

*Source: CHAPTER 2 A NEw LOOk AT GLOBAL BANkING VuLNERABILITIES (IMF, October 2023).*

### 1. Number of Banks and Size Distribution

### ch2 - 1. Number of Banks and Size Distribution

### Key Risk Indicator (KRI) framework and predictive value
- The KRI framework aggregates 12 indicators across five CAMELS and market dimensions: Capital adequacy, Asset quality, Management (proxied by dividend forecasts), Earnings, Liquidity, and Market metrics.
- Historical data from the first quarter of 2018 to the second quarter of 2023, aggregate consensus analysts forecasts for the second quarter of 2023 if actual data were not available, and aggregate consensus analysts forecasts for the third and fourth quarters of 2023 are used to determine expectations for bank performance and potential risk.
- Two econometric results:
  - The indicators have predictive power in forecasting bank stress events (see Online Annex 2.5).
  - The number of KRIs flagged is a quantitatively meaningful and statistically significant predictor of capital losses in the global stress tests adverse scenario.
- Limitation: The consensus of analyst forecasts is made under varying, unrevealed macro assumptions, reducing congruence with the global stress test; KRIs and stress tests should be used complementarily.
- Note: Price-to-book and market leverage metrics for Q3 2023 and Q4 2023 used market data as of September 8, 2023.

### Evolution of flagged banks and monitoring-list sizes
- The number of banks flagged as vulnerable in three or more of the five risk dimensions:
  - Spiked dramatically with the onset of the COVID-19 pandemic.
  - Fell sharply in 2021.
  - Climbed to another peak just before the March 2023 bank turmoil.
- Specific monitoring-list counts and asset totals:
  - In the second quarter of 2023: 85 banks with $26 trillion in total assets were on the KRI monitoring list due to breaches in at least three KRI risk dimensions.
  - Aggregate consensus analyst forecasts project:
    - Third quarter of 2023: 80 banks with $21 trillion in assets (decline driven primarily by an improvement in liquidity and earnings).
    - Fourth quarter of 2023: 82 banks with $25 trillion in assets (driven by weaker earnings).
  - In the fourth quarter of 2023: 25 banks flagged as vulnerable on four or more risk dimensions, with $9 trillion in combined assets.
- Historical peak during COVID-19:
  - The period from the first to second quarter of 2020 shows the largest concentration of vulnerable banks, with more than 200 banks in the monitoring list.
- The KRI framework flagged, in the fourth quarter of 2022, the four banks that ultimately failed in March 2023:
  - Credit Suisse, Silicon Valley Bank, and Signature Bank breached the threshold in the market KRI dimension (with additional breaches: earnings and liquidity for Signature Bank; capital adequacy and earnings for Silicon Valley Bank; capital adequacy and liquidity for Signature Bank).
  - First Republic breached thresholds in capital adequacy, asset quality, and earnings dimensions.

### Risk characteristics distinguishing monitoring-list banks
- Comparison of flagged (monitoring list) versus non-flagged banks using standardized z-scores:
  - Monitoring-list banks score significantly worse across nearly all categories except nonperforming loan ratio, coverage ratio, and quarterly deposit growth.
  - Key differentiators: low price-to-book ratios, low return on equity, and stretched net loan-to-deposit ratios.
- Sector and balance-sheet features by bank size:
  - Small banks: total assets of $10 billion or less — low profitability, stretched net loan-to-deposit ratios, and low price-to-book ratios.
  - Medium-sized banks: total assets between $10 and $100 billion — profitability struggles, stretched net loan-to-deposit ratios, low price-to-book ratios, and high market leverage.
  - Large banks: total assets of more than $100 billion — low profitability, low price-to-book ratios, and high market leverage; consensus forecasts call for profitability to decline by the end of the year due to net interest margin compression and rising provision expenses.

### Regional distribution and structural highlights (four-or-more KRI flags and monitoring-list shares)
- Europe:
  - Flagged banks include some of the largest banks in Europe, with estimated combined total assets of more than $8 trillion by the end of the year.
  - Forecast-based KRIs show European banks are expected to comprise 30 percent of the monitoring list on a total asset basis by the fourth quarter of 2023.
  - Key vulnerabilities: low ratios of equity to total assets, low profitability, low price-to-book ratios, and higher dependency on noncore deposit funding.
- Asia (excluding China-specific note below):
  - Fourth-quarter forecast group of flagged banks has combined total assets of more than $1 trillion.
  - Asia is expected to comprise 10 percent of monitoring-list assets in the fourth quarter of 2023.
  - Vulnerabilities: low equity-to-assets ratios, pressures on profitability from rising funding costs and lower fee income, elevated market leverage for a handful of banks, low price-to-book ratios, and expected deterioration in profitability from lower net interest income, higher noninterest expenses, and higher provision expenses.
- China:
  - Fourth-quarter projections call for the number of flagged banks to increase to $4.6 trillion in assets by the fourth quarter, representing 31 percent of monitoring-list total assets.
  - Flagged banks characterized by lower capital ratios, low profitability, and low price-to-book ratios; consensus forecasts call for lower profitability from compression of net interest margins due to declines in the prime rate for loans and lower noninterest income.
- Latin America:
  - Fourth-quarter flagged banks include a few large banks in several countries with estimated combined total assets of more than $900 billion.
  - Consensus forecasts expect profitability to improve by the third quarter of 2023, but equity-to-total-assets and price-to-book ratios are expected to remain low.
  - Projected share of monitoring-list assets in the fourth quarter: 4 percent.
- North America:
  - Flagged banks include a few large banks and many US regional banks, with estimated combined total assets of more than $6 trillion.
  - Projected share of monitoring-list assets in the fourth quarter: 25 percent.
  - Market-driven indicators such as market leverage and changes in forecasted dividend per share signal stress for banks with high concentrations of commercial real estate in total loans.

### Temporal and cross-sectional volatility (heat map findings)
- Table 2.2 heat map (KRI Global Volatility Heat Map) highlights:
  - Three main observations:
    - Peak concentration of vulnerable banks in Q1–Q2 2020 related to the COVID-19 pandemic (more than 200 banks on monitoring list).
    - Gradual run-up in the number of vulnerable banks in early 2022, mainly in Europe, corresponding to the invasion of Ukraine in Q1 2022.
    - Capital adequacy, earnings, and market KRIs capture an increasing number of banks ahead of the banking turmoil in Q1 2023.
- The heat map counts are based on historical data from Q1 2018 to Q2 2023, aggregate consensus data for Q2 2023 if actual data were not available, and aggregate consensus forecast data for Q3 and Q4 2023.

*Sources: Bloomberg Finance L.P.; Visible Alpha; and IMF staff calculations.*

### 2023. The indicators suggest that liquidity stress, a key

### ch2 - 2023. The indicators suggest that liquidity stress, a key

### Timing and evidence of rising liquidity stress
- Liquidity stress began increasing as early as June 2022.
- Evidence: growing number of banks experiencing deposit outflows, higher ratios of loans to deposits, and lower shares of deposits in total liabilities.

### Congruence of Global Stress Test (GST) and Key Risk Indicators (KRI) frameworks
- Samples overlap: 168 banks are in both the global stress test and KRI framework samples.
- Relationship between number of KRIs flagged and Tier 1 / CET1 impacts under the GST adverse scenario:
  - Among 47 banks flagged as vulnerable on two KRI dimensions as of 2022:Q4, the Tier 1 ratio declines about 2 percentage points, on average, under the GST adverse scenario.
  - Among 12 banks flagged as vulnerable on four KRI dimensions, the average Tier 1 impact increases almost –4 percentage points.
  - Among banks with three or four flagged KRIs, the worst quintile had an average decline of Tier 1 capital ratio of more than 8 percentage points.
- Cross-bank regression result:
  - Regression coefficient of about –0.7 (highly statistically significant), suggesting every increase of one flag among the KRI dimensions is associated with a fall of 0.7 percentage point in the Tier 1 capital ratio in the GST adverse scenario.
- Figure reference: Average Impact on CET1 Ratios under GST Adverse Scenario for Banks by Number of KRIs Flagged, 2022:Q4 (percentage points).

### Key analytical insights
- GST finds banks struggling to stay solvent under stagflationary scenarios with many banks suffering significant capital losses driven largely by mark-to-market losses on securities holdings in a higher-for-longer interest rate environment.
- In the United States, losses are concentrated in smaller, regional banks (consistent with March 2023 observations).
- KRI flags identify banks expected to be weak for various reasons:
  - Lower expected capital (some Chinese banks).
  - Further declines in price-to-book ratios (some European banks).
  - Declining liquidity (some banks in advanced economies).
- The KRI framework is designed to identify banks meriting closer examination and will by construction have a high level of type I errors given the rarity of actual bank failures.

### Policy recommendations
- Strengthen banking sector resilience urgently by:
  - Enhancing banks’ capital levels to ensure all banks maintain adequate capital ratios under stress scenarios.
  - Reinvigorating supervision and risk assessments, including through enhanced stress testing.
  - Timely and consistent implementation of international standards.
  - Strengthening regulations and crisis management frameworks.

### Enhancing risk assessments
- Expand the sample of banks subjected to stress tests and enhance methodologies, for example by incorporating interactions between funding and solvency and deposit stability.
- Consider making adverse scenarios more severe while keeping narratives plausible to uncover vulnerabilities.
- Leverage more timely and granular data to improve accuracy and comprehensiveness of supervisory risk assessments, while narrowing gaps in data coverage and granularity.
- Monitor market metrics closely, given markets can shift rapidly from a balance sheet view to a mark-to-market view of risks; be particularly cautious about banks with persistent price-to-book ratios below 1.
- Note: net interest income can appear systemically insensitive to interest rates while individual banks remain vulnerable.

### Sharpening supervision and regulation
- Supervisory capacity issues:
  - Financial Sector Assessment Program (FSAP) assessments indicate that more than half of jurisdictions still do not have independent bank supervisors with a clear mandate to effect financial stability, sound internal governance, or resources appropriate to responsibilities.
- Liquidity and interest rate risk supervision gaps:
  - Nearly one-fifth of jurisdictions have weak supervisory and regulatory practices with respect to liquidity (assessments based on 47 Basel Core Principles assessments conducted between 2012 and 2019).
  - Several jurisdictions fail to address liquidity needs in foreign currency, define liquid assets appropriately, or impose requirements on a consolidated level.
  - Several jurisdictions do not require banks to maintain capital against interest rate risk in the banking book; more than a quarter have material deficiencies in monitoring and controlling this risk (Dordevic and others 2021).
- Recommendation: implement prudential rules ensuring banks hold appropriate capital against interest rate risk and guard against hidden losses that could materialize abruptly in liquidity shocks.
- Full, timely, and consistent implementation of Basel III elements remains important; deviations and delays by some major jurisdictions risk regulatory fragmentation.
- Reminder from Basel Core Principles review: although proportionality is needed, all segments of the banking sector should be subject to rigorous prudential standards compatible with the Basel framework.

### Fortifying crisis management frameworks
- Interactions between solvency and liquidity in the absence of central bank liquidity facilities could lead to distress among a considerable number of banks under adverse shocks.
- Enhance commercial banks’ preparedness to use eligible collateral and access central bank facilities; improve authority communication on availability and usage (acceptable collateral, haircuts).
- Institutional arrangements for emergency liquidity provision vary widely; all banks should be required to periodically test access to central bank instruments.
- Central banks should set up emergency liquidity assistance frameworks in normal times and abide by principles concerning collateralization, conditions, and state guarantees.
- Too-big-to-fail reform agenda: progress needed on effective public sector liquidity backstops among resolution authorities and deposit insurers, preparedness to operationalize resolution options, and clarity on deposit insurance roles in a digital context where deposit runs can accelerate.
- FSAP assessments highlight deposit insurers often face significant weaknesses in funding arrangements and backstop arrangements for funding liquidity.
- Authorities should plan resolution strategies broadly, recognizing failures of relatively small banks can have systemic implications and cross-border contagion effects.

### Bank run experience and recent episodes
- Recent (March 2023) bank runs in Switzerland and the United States were unusually large and fast, facilitated by rapid online withdrawals and rapid information spread via social media and digital channels.
- Historical precedents of rapid online runs:
  - Northern Rock (UK), 2007: the bank lost almost 60 percent of its retail deposits in 2007, including 20 percent over just five days (between September 13 and 17).
  - Landsbanki (Iceland), 2008: UK internet banking branch suffered a rapid run amid broader Icelandic banking crisis.
  - Continental Illinois rescue (1984) resulted from a “high-speed electronic bank run.”
- Systemic crises can involve massive total deposit outflows; note from panel data:
  - United States, 1930–33: an average of 67 failed banks (panel note).
- Case-study figure references: Scale and Speed of Deposit Outflows; Total Outflows as a Share of Banking Deposits in Systemic Crises.

*Source: IMF Global Financial Stability Report chapter content (October 2023).*

---


_Source: https://www.imf.org/-/media/files/publications/gfsr/2023/october/english/ch2.pdf_