## gfsnea2025003

## Source details

**Canonical URL:** [gfsnea2025003](https://www.imf.org/-/media/files/publications/gfs-notes/2025/english/gfsnea2025003.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/gfs-notes/2025/english/gfsnea2025003.pdf.md)
- [Structured JSON version](/-/media/files/publications/gfs-notes/2025/english/gfsnea2025003.pdf.json)

---

### Overview
- Research question: Do banks transmit credit risk to the sovereign?
- Context: Revisits the sovereign–bank nexus in the post–global financial crisis reform environment, using the sudden collapse of the Silicon Valley Bank (SVB) in March 2023 as an exogenous adverse shock to the banking sector.
- Sample and period:
  - Daily data on credit default swap (CDS) spreads for both sovereigns and banks.
  - Sample period: 2015–23.
  - Sample composition: the United States and 31 other major advanced and emerging market economies.
  - Event dates for SVB collapse: March 9 and 10, 2023.
- Identification strategy: Use the SVB collapse—a banking-sector event driven mainly by internal risk factors—as an exogenous shock to identify causal transmission from banks to sovereigns.

### Transmission channels (three key channels)
- Direct exposure channel: banks’ holdings of government debt—fire sales to meet liquidity needs can raise sovereign yields and credit risk.
- Government safety-net channel: implicit and explicit guarantees—banking-sector stress may raise the need to activate guarantees, straining fiscal accounts and increasing sovereign credit risk.
- Indirect feedback loop (real-economy) channel: banking-sector stress can disrupt lending to corporates and households, undermining activity and government finances, raising sovereign credit risk.

### Data and empirical approach
- Primary measure: CDS spreads as measures of credit risk for banks and sovereigns.
- Aggregation: (asset-weighted) average of banks’ CDS spreads used for banking-sector credit risk.
- Identification: Short window around SVB collapse (2023:M2–M3, with event dates 2023:M3/9–10) to exploit a shock to banks’ credit risk that was largely unrelated to sovereign credit risk.
- Geographic scope: United States analyzed in detail; panel estimations across other major economies to assess cross-border transmission.
- Controls and global factors: daily changes in government bond yields, stock returns, public debt-to-GDP ratio; VIX; country–month–year interaction effects.

### Key empirical findings (summary)
- Baseline U.S. result over full sample (2015–23): changes in the (asset-weighted) average of banks’ CDS spreads in the United States are, on average, not significantly associated with changes in the US sovereign CDS spreads.
- SVB event effect (short window around March 9–10, 2023):
  - A 10 basis points increase in the (asset-weighed) average of banks’ one-year CDS spread implied an increase of 4 basis points in the US sovereign one-year CDS spread.
  - US panel estimate: a 10 basis point increase in the (asset-weighted) average of banks’ one-year CDS spreads implies an increase of about 4.5 basis points in the US sovereign one-year CDS spread (Annex Table 2.1, columns 1–3).
  - Effect on five-year CDS: a 10 basis points increase in banks’ five-year CDS spread raises the sovereign five-year CDS spread by about 1 basis point.
  - The effect was short-lived due to timely intervention by US authorities, which ensured that no direct losses were borne by US taxpayers.
- Cross-country/panel findings during the SVB episode:
  - Stress in the US banking sector spread to other major economies; rises in their banking-sector CDS spreads similarly impacted their sovereign CDS spreads.
  - Stronger bank-to-sovereign risk transmission occurred in economies with:
    - Higher public debt-to-GDP ratios.
    - Greater exposure of banking sectors to domestic sovereign debt.
    - Less well-capitalized banking systems.
  - Quantitative cross-country estimate: a 10 basis points increase in banks’ CDS spread implied, on average, an increase in sovereign CDS spread by about 3.5 basis points in the sample of countries (excluding the United States) during the SVB collapse.
  - Transmission is about three times larger for countries with public debt-to-GDP ratios above the sample median (median public debt-to-GDP ratio in the sample is 67 percent) and four times larger at the 75th percentile threshold.

### Stylized facts on exposures and trends
- United States banks’ holdings of domestic sovereign debt:
  - About 8 percent of total assets in 2019.
  - Jumped to 12 percent in 2020–21.
  - Declined to about 10 percent in 2023.
- Emerging markets:
  - Average share of domestic sovereign debt in emerging market bank assets stood at about 17 percent in 2023.
  - More than 20 percent in several emerging markets.
- Advanced economies (excluding the United States): on average, banks’ exposure to domestic sovereign debt has fallen over the past decade, though material exposures remain in some cases (including indirect exposures via loans backed by government guarantees).

### The collapse of Silicon Valley Bank (SVB): facts and transmission
- Bank size and balance-sheet facts:
  - SVB was the 16th largest bank in the United States.
  - Assets at end-2022: "$209 billion in assets".
  - Deposits at end-2022: "$173 billion in deposits".
  - "More than half of SVB’s assets were government- and agency-issued mortgage-backed securities."
  - Depositor base: "venture capital and start-up firms comprised more than half of SVB’s deposits."
- Interest-rate dynamics and losses:
  - Unrealized losses on SVB’s held-to-maturity portfolio by end-2022: "$15.2 billion".
  - SVB sold available-for-sale securities on March 8, 2023, amounting to "$21 billion".
  - Loss incurred on sale: "$1.8 billion".
  - Planned capital raise announced: intended to raise "$2 billion" (footnote reports planned structure as "$2.25 billion" via "$1.25 billion in public offering of common shares, $0.5 billion in offering of mandatory convertible preferred stock, and $0.5 billion in private offering to its long-standing client, General Atlantic").
- Run dynamics and resolution:
  - On October 21, 2022, SVB stock price fell "about 24 percent"; the one-year EDF increased by "approximately 0.4 percentage points."
  - On March 9, 2023, deposit withdrawals totaled "about $42 billion—almost one-fourth of its total deposit base and 300 percent of its capital" and its stock price dropped "by 60 percent."
  - Management expected "$100 billion more the next day" in withdrawal requests.
  - California Department of Financial Protection and Innovation took possession and FDIC was appointed receiver on March 10, 2023.
  - EDF 1y on March 10, 2023 = "35 pct."
- Speed and novelty:
  - SVB collapse was "the fastest bank collapse in US history, occurring in less than 48 hours."
  - It was the second-largest US bank failure by dollar amount after Washington Mutual.
- Market impacts during March 9–10, 2023:
  - Asset-weighted average banks’ one-year CDS spreads increased by "5.4 basis points".
  - Asset-weighted average banks’ five-year CDS spreads increased by "15 basis points".
  - Banks’ EDF increased by "0.4 percentage points".
  - S&P 500 index dropped by "a cumulative 3 percent".
  - US banks stock index fell by "7 percent".
  - US regional banks stock index fell by "13 percent".
  - Four largest US banks lost "about $44.5 billion in market value between March 8 and 10, 2023."
  - Spread on US sovereign one-year CDS increased by "about 10 basis points during March 9 and 10, 2023."
  - Global spillovers: bank stock prices fell sharply in major economies despite limited direct international exposures to SVB.

### Policy response and immediate effects
- US authorities’ measures (March 12, 2023):
  - FDIC announced that "all deposits of SVB (as well as of Signature Bank) were to be fully protected."
  - Federal Reserve initiated an emergency funding program allowing banks to borrow using high-quality securities as collateral at full-face value.
  - Government announced losses to the Deposit Insurance Fund to support uninsured depositors would be recovered by a special assessment on banks and "no losses in the resolution of the banks were to be borne by taxpayers."
- Effect:
  - These actions "helped to alleviate investors’ concerns about systemic risk" and likely reduced expectations of taxpayer bailouts.
  - Broader US stock market recovered soon after; US sovereign CDS spread and CDS-implied EDF declined.

### Empirical framework and identification
- Regression specification:
  - Dependent variable: SOV_CDS_it_∆ = first difference of daily sovereign CDS spread for country i at time t.
  - Key independent variable: BANK_CDS_it_∆ = first difference of asset-weighted average of banks’ daily CDS spreads in country i at time t.
  - SVB dummy: equals 1 on March 9 and 10, 2023, and 0 otherwise.
  - Interaction term BANK_CDS_it_∆ × SVB isolates changes in banks’ CDS spreads during SVB collapse.
  - Controls (Z_k): daily changes in government bond yields, stock returns, public debt-to-GDP ratio; country–month–year interaction effects.
  - Global factors (G): capture common shocks such as daily changes in the VIX.
  - Fixed effects: α_i country fixed effects; error term e_it.
- Estimation details:
  - Equation estimated separately for United States and for panel excluding the United States.
  - Data frequency and sample: daily CDS spreads for "January 2015 to December 2023."
  - Standard errors: heteroscedasticity adjusted for US estimates; clustered at country and date levels for panel estimates.
  - Short two-day window (March 9–10, 2023) helps address simultaneity concerns.

### Empirical results (detailed highlights)
- United States:
  - Baseline relationship: "On average, there is no strong relationship between banks’ and sovereign short-term (one-year) CDS spreads over our sample period" (Annex Table 2.1).
  - Effect during SVB collapse:
    - A "10 basis point increase in the (asset-weighted) average of banks’ one-year CDS spreads implies an increase of about 4.5 basis points in the US sovereign one-year CDS spread" (Annex Table 2.1, columns 1–3).
    - SVB dummy indicates the short-term sovereign default risk was about 3 basis points higher during March 9 and 10, 2023.
  - Robustness: results hold with controls for government bond yields, S&P 500 return, VIX, month–year effects; robust to post-pandemic sample and to extending sample to 2010–2023.
  - Annex Table 2.2 (post-SVB, January 2015–December 2023; Observations: 2,128; R2: 0.079):
    - SVB collapse (March 9): 3.499** (standard error 1.543)
    - SVB collapse (March 10): 4.481*** (standard error 1.515)
    - SVB collapse (March 9-10): 2.815* (standard error 1.612)
    - ∆ Banks' CDS × SVB collapse (March 9-10): 0.433*** (standard error 0.161)
    - ∆ Sovereign CDS: -0.012 (standard error 0.078)
    - ∆ Banks' CDS: -0.015 (standard error 0.038)
- Panel (selected coefficients from Annex Table 2.1 and robustness):
  - ∆ Banks' CDS × SVB collapse = 0.379*** (column 1; (0.112))
  - SVB collapse coefficients: 3.026** (column 1; (1.458)), 2.914** (column 2; (1.477)), 2.902* (column 3; (1.481))
  - ∆ Banks' CDS examples: 0.018 (column 1; (0.046)), -0.016 (column 2; (0.037)), -0.022 (column 3; (0.043)), and 0.023*** (column 6; (0.008))
  - ∆ Banks' CDS-implied EDF × SVB collapse = 0.022*** (column 7; (0.001))
  - Sample periods reported: 2015-23, 2021-23, 2010-23.
  - Observations across columns: 2,317; 2,128; 2,010; 713; 3,266; 2,128; 2,128.
  - R2 values shown: 0.089; 0.078; 0.076; 0.088; 0.071; 0.092; 0.194.
- Robustness: results robust to restricting to January 2021–December 2023; to alternative SVB collapse dates; and to inclusion of stock market volatility and geopolitical risk measures.

### Cross-country patterns and conditional factors
- Across major economies (excluding the United States): changes in banking sector credit risk are not significantly associated with sovereign credit risk over the full sample, but during the SVB collapse transmission was strong and statistically significant.
- Transmission is stronger:
  - In economies with public debt-to-GDP above the sample median.
  - In emerging markets compared to advanced economies.
  - In economies where banks hold a higher share of domestic sovereign debt in total assets (above the sample median).
  - In economies with Tier 1 capital to assets ratio below the median.
- Interpretation: investors view high public debt, concentrated bank holdings of domestic sovereign debt, and weak bank capital adequacy as vulnerabilities that increase fiscal burden during banking shocks.

### Policy implications and recommendations
- Strengthen supervisory and crisis management frameworks:
  - Ensure proactive supervision and early intervention to maintain banking sector stability.
  - Ensure sufficient recovery and resolution planning to tackle failing banks swiftly and avoid systemic crises without undue strain on government finances.
  - Strengthen deposit insurance regimes to maintain depositor confidence given faster withdrawal speeds enabled by new technology.
- Contain fiscal vulnerabilities and build buffers:
  - Stronger fiscal positions can reduce perceived sovereign vulnerability and help prevent a doom loop.
- Reduce interdependency between banks and sovereigns:
  - Monitor concentration risks where banks’ asset portfolios are heavily tilted toward domestic sovereign debt.
  - Consider introducing capital surcharges on bank holdings of domestic sovereign bonds above certain thresholds (as noted in IMF 2022).

### Data and measurement notes (Annex highlights)
- Key data sources:
  - Banks’ domestic sovereign bond holdings and total assets from IMF, Monetary and Financial Statistics.
  - Banks’ and sovereign CDS spreads and CDS-implied EDF from Moody’s CreditEdge.
  - Government bond yield (10-year) from Bloomberg Finance L.P.
  - Public debt-to-GDP and macro variables from IMF, World Economic Outlook and other standard sources.
- Country sample (Annex Table 1.1) includes advanced and emerging economies such as Australia, Brazil, Canada, China, India, Japan, United Kingdom, United States, and others.

*Source: GLOBAL FINANCIAL STABILITY NOTES — Bank to Sovereign Risk Transmission: New Evidence (International Monetary Fund), Introduction and Annexes.*

### Introduction ...........................................................................................................

### Introduction

### Overview
- Research question: Do banks transmit credit risk to the sovereign?
- Context: Revisits the sovereign–bank nexus in the post–global financial crisis reform environment, using the sudden collapse of the Silicon Valley Bank (SVB) in March 2023 as an exogenous adverse shock to the banking sector.
- Sample and period:
  - Daily data on credit default swap (CDS) spreads for both sovereigns and banks.
  - Sample period: 2015–23.
  - Sample composition: the United States and 31 other major advanced and emerging market economies.
  - Event dates for SVB collapse: March 9 and 10, 2023.
- Identification strategy: Use the SVB collapse—a banking-sector event driven mainly by internal risk factors—as an exogenous shock to identify causal transmission from banks to sovereigns.

### Transmission channels (three key channels)
- Direct exposure channel: banks’ holdings of government debt. Adverse shocks can force banks to sell government bonds to meet liquidity needs, raising sovereign yields and credit risk.
- Government safety-net channel: implicit and explicit guarantees. Banking-sector stress may raise the actual or perceived need to activate guarantees, straining fiscal accounts and increasing sovereign credit risk.
- Indirect feedback loop (real-economy) channel: banking-sector stress can disrupt lending to corporates and households, undermining activity and government finances, raising sovereign credit risk.

### Data and empirical approach
- Primary measure: CDS spreads as measures of credit risk for banks and sovereigns.
- Aggregation: (asset-weighted) average of banks’ CDS spreads used for banking-sector credit risk.
- Identification: Short window around SVB collapse (2023:M2–M3, with event dates 2023:M3/9–10) to exploit a shock to banks’ credit risk that was largely unrelated to sovereign credit risk.
- Geographic scope: United States (epicenter) analyzed in detail; panel estimations across other major economies to assess cross-border transmission.

### Key empirical findings
- Baseline U.S. result over full sample (2015–23): Changes in the (asset-weighted) average of banks’ CDS spreads in the United States are, on average, not significantly associated with changes in the US sovereign CDS spreads.
- SVB event effect (short window around March 9–10, 2023):
  - A 10 basis points increase in the (asset-weighed) average of banks’ one-year CDS spread implied an increase of 4 basis points in the US sovereign one-year CDS spread.
  - The effect was short-lived due to timely intervention by the US government and the Federal Reserve, which ensured that no direct losses were borne by US taxpayers.
- Cross-country/panel findings during the SVB episode:
  - Stress in the US banking sector spread to other major economies; rises in their banking-sector CDS spreads similarly impacted their sovereign CDS spreads.
  - Stronger bank-to-sovereign risk transmission occurred in economies with:
    - Higher public debt-to-GDP ratios.
    - Greater exposure of banking sectors to domestic sovereign debt.
    - Less well-capitalized banking systems.
  - Interpretation: Investors view banking-sector stress as particularly economically costly for such countries, reflecting concerns about fiscal sustainability via potential government support and slower economic activity.

### Stylized facts on exposures and trends
- Banks’ exposure to domestic sovereign debt and public debt:
  - United States banks’ holdings of domestic sovereign debt:
    - About 8 percent of total assets in 2019.
    - Jumped to 12 percent in 2020–21 as public borrowing increased for pandemic fiscal stimulus.
    - Declined to about 10 percent in 2023 (still higher than most other advanced economies).
  - Emerging markets:
    - Average share of domestic sovereign debt in emerging market bank assets stood at about 17 percent in 2023.
    - More than 20 percent in several emerging markets.
  - Advanced economies (excluding the United States): on average, banks’ exposure to domestic sovereign debt has fallen over the past decade, though material exposures remain in some cases (including indirect exposures via loans backed by government guarantees).

### Contributions and implications
- Contributions of this note:
  - Extends literature by focusing on a recent, exogenous banking-sector shock (SVB collapse) in the post-pandemic, post-reform environment.
  - Provides a global perspective (United States plus 31 other economies) allowing cross-country comparisons of risk transmission.
  - Uses the SVB collapse to achieve causal identification from banking-sector stress to sovereign risk.
  - Highlights that moderate shocks to the US banking sector can pose significant global contagion risks, even when direct exposure of foreign banks to failing US institutions is limited.
- Policy implications and considerations:
  - Findings suggest a continued need to monitor and possibly strengthen reforms to mitigate bank-to-sovereign risk transmission, especially in economies with higher public debt, larger bank holdings of domestic sovereign debt, and weaker bank capitalization.
  - Notes the potential tension: post–global financial crisis capital and liquidity rules that incentivize banks to hold larger amounts of sovereign bonds could strengthen the sovereign–bank nexus even as other crisis-recovery reforms weaken it.

*Source: GLOBAL FINANCIAL STABILITY NOTES — Bank to Sovereign Risk Transmission: New Evidence (International Monetary Fund), Introduction section.*

### Annex 1 for further information on the sample and data sources.

### gfsnea2025003 - Annex 1 for further information on the sample and data sources

### Correlation between bank and sovereign credit risk (2015–23)
- Correlation of banks’ and sovereigns’ CDS spreads "has more than doubled for the United States and emerging markets since the pandemic" (Figure 3, panel 1).
- Correlation between banks’ and sovereign CDS-implied EDF has "also increased significantly since the pandemic in both the United States and emerging markets" (Figure 3, panel 2).
- Bank–sovereign credit-risk correlation tends to be stronger around episodes of elevated financial market stress, proxied by the VIX, such as:
  - the onset of the pandemic, and
  - the Federal Reserve’s switch to monetary policy tightening in the first half of 2022 (Figure 3, panel 3).
- Note on metrics:
  - CDS spreads and CDS-implied expected default frequency (EDF, computed by Moody’s CreditEdge) are used as measures of credit risk.
  - Panels 1 and 2 coefficients are obtained from a simple linear regression of sovereign CDS on asset-weighted banks’ CDS; "All coefficients are statistically significant at conventional levels."
  - Panel 3 shows the 30-day rolling correlation between banks’ and sovereign one-year CDS spreads in the United States and the VIX.

### The collapse of Silicon Valley Bank (SVB): facts and transmission
- SVB overview and balance-sheet facts:
  - SVB was the 16th largest bank in the United States.
  - Assets at end-2022: "$209 billion in assets".
  - Deposits at end-2022: "$173 billion in deposits".
  - "More than half of SVB’s assets were government- and agency-issued mortgage-backed securities."
  - Depositor base: "venture capital and start-up firms comprised more than half of SVB’s deposits."
- Interest-rate dynamics and losses:
  - Unrealized losses on SVB’s held-to-maturity portfolio by end-2022: "$15.2 billion".
  - To strengthen its balance sheet, SVB sold available-for-sale securities on March 8, 2023, amounting to "$21 billion".
  - Loss incurred on sale: "$1.8 billion".
  - Planned capital raise announced: intended to raise "$2 billion" (text also reports planned structure in a footnote as "$2.25 billion" via "$1.25 billion in public offering of common shares, $0.5 billion in offering of mandatory convertible preferred stock, and $0.5 billion in private offering to its long-standing client, General Atlantic").
- Run dynamics and resolution:
  - On October 21, 2022, SVB stock price fell "about 24 percent" after weaker-than-expected third-quarter earnings per share; the one-year EDF increased by "approximately 0.4 percentage points."
  - On March 9, 2023, deposit withdrawals totaled "about $42 billion—almost one-fourth of its total deposit base and 300 percent of its capital" and its stock price dropped "by 60 percent."
  - Management expected "$100 billion more the next day" in withdrawal requests.
  - California Department of Financial Protection and Innovation took possession and FDIC was appointed receiver on March 10, 2023.
  - EDF 1y on March 10, 2023 = "35 pct." (from Figure 4 note).
- Speed and novelty:
  - SVB collapse was "the fastest bank collapse in US history, occurring in less than 48 hours."
  - It was the second-largest US bank failure by dollar amount after Washington Mutual.
- Aftermath and market impact:
  - US banking sector credit risk measures during March 9–10, 2023:
    - Asset-weighted average banks’ short-term (one-year) CDS spreads increased by "5.4 basis points".
    - Asset-weighted average banks’ long-term (five-year) CDS spreads increased by "15 basis points".
    - Banks’ EDF increased by "0.4 percentage points".
  - Equity market moves during the turmoil (March 9–10, 2023):
    - S&P 500 index dropped by "a cumulative 3 percent".
    - US banks stock index fell by "7 percent".
    - US regional banks stock index fell by "13 percent".
    - The four largest US banks—JPMorgan Chase, Bank of America, Wells Fargo, and Citigroup—together lost "about $44.5 billion in market value between March 8 and 10, 2023."
  - Sovereign-market reaction:
    - Spread on US sovereign one-year CDS increased by "about 10 basis points during March 9 and 10, 2023."
    - This occurred alongside falling US Treasury yields (flight to quality) and elevated investor risk aversion.
  - Global spillovers:
    - Bank stock prices fell sharply in major economies; international bank exposures to SVB were limited but stress spread through multiple channels.

### Policy response
- US authorities’ measures (March 12, 2023):
  - FDIC announced that "all deposits of SVB (as well as of Signature Bank) were to be fully protected."
  - Federal Reserve initiated an emergency funding program allowing banks to borrow using high-quality securities as collateral at full-face value.
  - Government announced losses to the Deposit Insurance Fund to support uninsured depositors would be recovered by a special assessment on banks and "no losses in the resolution of the banks were to be borne by taxpayers."
- Effect:
  - These actions "helped to alleviate investors’ concerns about systemic risk" and likely reduced expectations of taxpayer bailouts.
  - Broader US stock market recovered soon after; US sovereign CDS spread and CDS-implied EDF declined (Annex Figure 1.1).

### Empirical framework and identification
- Regression specification (equation 1):
  - Dependent variable: SOV_CDS_it_∆ = first difference of daily sovereign CDS spread for country i at time t.
  - Key independent variable: BANK_CDS_it_∆ = first difference of asset-weighted average of banks’ daily CDS spreads in country i at time t.
  - SVB dummy: takes value 1 on dates March 9 and 10, 2023, and 0 otherwise.
  - Identification: interaction term BANK_CDS_it_∆ × SVB isolates change in banks’ CDS spreads during SVB collapse to capture a shock to banks’ CDS spreads "that is unrelated to sovereign risk."
  - Controls (Z_k): daily changes in government bond yields, stock returns, public debt-to-GDP ratio; country–month–year interaction effects to control for lower-frequency macro factors (economic growth, inflation).
  - Global factors (G): capture common shocks such as daily changes in financial uncertainty proxied by the VIX.
  - Fixed effects: α_i country fixed effects; error term e_it.
- Estimation details:
  - Equation 1 estimated separately for:
    - United States (i ∈ United States), and
    - panel of major advanced and emerging market economies excluding the United States (i ∉ United States).
  - Data frequency and sample: daily CDS spreads for "January 2015 to December 2023."
  - Standard errors:
    - For US estimates: heteroscedasticity adjusted.
    - For panel estimates: clustered at the country and date levels.
  - Short two-day window (March 9–10, 2023) helps address simultaneity concerns.

### Empirical results (United States)
- Baseline relationship:
  - "On average, there is no strong relationship between banks’ and sovereign short-term (one-year) CDS spreads over our sample period" (Annex Table 2.1).
- Effect during banking distress (SVB collapse):
  - Change in banking-sector CDS spread has a "statistically significant effect on the sovereign CDS spread in times of banking sector distress (proxied by the SVB collapse)."
  - Magnitudes:
    - A "10 basis point increase in the (asset-weighted) average of banks’ one-year CDS spreads implies an increase of about 4.5 basis points in the US sovereign one-year CDS spread" (Annex Table 2.1, columns 1–3).
    - The SVB dummy is "positive and statistically significant," suggesting that "the short-term sovereign default risk was about 3 basis points higher during March 9 and 10, 2023, than otherwise."
  - Robustness:
    - Results robust to inclusion of daily change in government bond yields, S&P 500 return, VIX, and month–year interaction effects.
    - Robust to restricting sample to post-pandemic years and to extending sample to 2010–2023.
- Long-term spreads:
  - Considering five-year CDS spreads:
    - A "10 basis points increase in banks’ five-year CDS spread raises the sovereign five-year CDS spread by about 1 basis point."
    - The effect is quantitatively smaller than for short-term spreads.
- Additional notes:
  - Results for one-year CDS-implied EDF are reported as consistent with strong transmission; specific magnitudes are provided in Annex Table 2.1 (referenced).

*Source: Annex 1 of "Bank to Sovereign Risk Transmission: New Evidence" (GLOBAL FINANCIAL STABILITY NOTES), IMF.*

### 0.1   percentage point. The results are also robust to restricting the sample period to the post-pandemic

### gfsnea2025003 - 0.1   percentage point. The results are also robust to restricting the sample period to the post-pandemic

### Main findings
- An adverse shock to the US banking sector could quickly spill over to the sovereign, raising the risk of a doom loop.
- The primary transmission channel during the SVB collapse appears to be investors’ perceived government support for banks (the “safety net” channel), rather than fire sales of government bonds (the exposure channel).
- Soon after the US government announced on March 12, 2023, that there would be no public bailout for the collapsed SVB and Signature Bank, the short-term US sovereign CDS spread declined and the strong link between changes in banks’ and the sovereign CDS spread weakened.
- The SVB episode shows that transmission via the safety net channel can occur even when bailouts or support have not been formally announced or conferred by the government.

### Quantitative estimates and event evidence
- Using the SVB collapse as an exogenous shock:
  - A 10 basis points increase in banks’ CDS spread implied, on average, an increase in sovereign CDS spread by about 3.5 basis points in the sample of countries (excluding the United States) during the SVB collapse.
  - The transmission effect is about three times larger for countries with public debt-to-GDP ratios above the sample median (median public debt-to-GDP ratio in the sample is 67 percent).
  - Considering the 75th percentile threshold for high public debt-to-GDP, the transmission of banks’ credit risk to the sovereign is four times as large as the baseline effect.
- Panel estimation for the United States (selected coefficients from Annex Table 2.1):
  - ∆ Banks' CDS × SVB collapse = 0.379*** (columns 1; standard error (0.112))
  - SVB collapse coefficient examples: 3.026** (column 1; (1.458)), 2.914** (column 2; (1.477)), 2.902* (column 3; (1.481))
  - ∆ Banks' CDS coefficients: examples include 0.018 (column 1; (0.046)), -0.016 (column 2; (0.037)), -0.022 (column 3; (0.043)), and 0.023*** (column 6; (0.008))
  - ∆ Banks' CDS-implied EDF × SVB collapse = 0.022*** (column 7; (0.001))
- Sample periods and observations reported in Annex Table 2.1:
  - Sample periods include 2015-23, 2021-23, 2010-23.
  - Observations: 2,317; 2,128; 2,010; 713; 3,266; 2,128; 2,128 across columns.
  - R2 values shown: 0.089; 0.078; 0.076; 0.088; 0.071; 0.092; 0.194.
- Robustness:
  - Results are robust to restricting the sample period to the post-pandemic years (January 2021 to December 2023).
  - Results are robust to alternative dates for the SVB collapse (e.g., treating March 9 or March 10, 2023, individually rather than both).
  - Results remain largely robust when including measures of stock market volatility and geopolitical risk.

### Cross-country patterns and conditional factors
- On average across major economies (excluding the United States), changes in banking sector credit risk are not significantly associated with changes in sovereign credit risk over the full sample period; however, during the SVB collapse the transmission was strong and statistically significant.
- Transmission of bank credit risk to sovereigns is stronger:
  - In economies with higher public debt-to-GDP (above the sample median).
  - In emerging markets compared to advanced economies, on average.
  - In economies where banks hold a higher share of domestic sovereign debt in total assets (above the sample median).
  - In economies with less well-capitalized banking systems (Tier 1 capital to assets ratio below the median).
- Interpretation: investors view high public debt, concentrated banking holdings of domestic sovereign debt, and weak bank capital adequacy as vulnerabilities that increase fiscal burden during banking-sector shocks.

### Policy implications and recommendations
- Strengthen supervisory and crisis management frameworks:
  - Ensure proactive supervision and early intervention to maintain banking sector stability.
  - Ensure sufficient recovery and resolution planning to tackle failing banks swiftly and avoid systemic crises without undue strain on government finances.
  - Strengthen deposit insurance regimes to maintain depositor confidence given faster withdrawal speeds enabled by new technology.
- Contain fiscal vulnerabilities and build buffers:
  - Stronger fiscal positions can reduce perceived sovereign vulnerability and help prevent a doom loop in the face of banking-sector shocks.
- Reduce interdependency between banks and sovereigns:
  - Monitor concentration risks where banks’ asset portfolios are heavily tilted toward domestic sovereign debt.
  - Consider introducing capital surcharges on bank holdings of domestic sovereign bonds above certain thresholds (as noted in IMF 2022).

### Data and measurement notes (Annex 1 highlights)
- Key financial variables and sources used include:
  - Banks’ domestic sovereign bond holdings and total assets from IMF, Monetary and Financial Statistics.
  - Banks’ and sovereign CDS spreads and CDS-implied EDF from Moody’s CreditEdge.
  - Government bond yield (10-year) from Bloomberg Finance L.P.
  - Public debt-to-GDP and macro variables from IMF, World Economic Outlook and other standard sources.
- Country sample listed in Annex Table 1.1 includes advanced and emerging economies such as Australia, Brazil, Canada, China, India, Japan, United Kingdom, United States, and others.

*International Monetary Fund — GLOBAL FINANCIAL STABILITY NOTES: Bank to Sovereign Risk Transmission: New Evidence (excerpts from Annex and main text).*

### Annex Table 2.2. Estimation Results for US Bank to Sovereign Risk Transmission:  Post–SVB

### Annex Table 2.2. Estimation Results for US Bank to Sovereign Risk Transmission: Post–SVB Collapse, 2015–23

### Main estimation: effects on ∆ Sovereign CDS (one year)
- Dependent variable: change in the U.S. sovereign one-year credit default swap (CDS) spread (basis points).
- Sample period: January 2015 to December 2023.
- Observations: 2,128
- R2: 0.079

### Key coefficient estimates (Columns (1) and (2))
- ∆ Sovereign CDS: -0.012 (standard error 0.078)
- ∆ Banks' CDS: -0.015 (standard error 0.038)

### SVB collapse date dummies and immediate effects
- SVB collapse (March 9): 3.499** (standard error 1.543)
- SVB collapse (March 10): 4.481*** (standard error 1.515)
- Post-SVB collapse (March 13): 0.167 (standard error 1.606)
- Post-SVB collapse (March 14): -2.267 (standard error 1.636)
- SVB collapse (March 9-10): 2.815* (standard error 1.612)

### Interaction effects: banks' CDS × SVB collapse / post-collapse
- ∆ Banks' CDS × SVB collapse (March 9-10): 0.433*** (standard error 0.161)
- ∆ Banks' CDS × Post-collapse (March 13): 0.020 (standard error 0.194)
- ∆ Banks' CDS × Post-collapse (March 14): -1.180 (standard error 0.852)

### Controls included
- ∆ VIX (log): 0.300 (standard error 0.573)
- ∆ Government bond yield (10 year): 0.241 (standard error 1.029)
- Month × year effects: Yes
- All equations include a constant (not shown).
- Robust estandard errors are reported in parentheses.

### Statistical significance notation
- *, **, and *** indicate statistical significance at the 10, 5 and 1 percent levels, respectively.

*Source: Authors' estimates.*

---


_Source: https://www.imf.org/-/media/files/publications/gfs-notes/2025/english/gfsnea2025003.pdf_