## Section 4 concludes.

## Source details

**Canonical URL:** [Section 4 concludes.](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026044-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026044-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026044-source-pdf.pdf.json)

---

### Data and Identification
- Sample and frequency
  - Sample spans January 1, 2019 to June 30, 2025.
  - Daily observations used for event-study analysis; aggregated to weekly frequency for VAR estimations.
  - Lag length p set to 13 weeks in VAR specifications.
- Data sources and variables
  - Stablecoin market capitalization: combined USDC and USDT from CoinGecko (log market capitalization used).
  - Financial market data: US Treasury yields, Nasdaq-100 index, Bloomberg Galaxy Crypto Index (BGCI), equity prices of payment providers/retailers/banks from Bloomberg; S&P 500 and broad dollar index from FRED.
  - High-frequency dataset includes stablecoin market capitalization, asset prices, and broader financial market indicators.
- Identification strategies overview
  - Primary: heteroskedasticity-based identification exploiting narrative-based stablecoin events in an event-study and SVAR-IV setting.
  - Complementary robustness: max-share identification (Uhlig 2004) within a VAR to recover the stablecoin shock orthogonal to other shocks.

### Narrative Identification (Event-study and SVAR-IV)
- Event selection and construction
  - Systematic week-by-week Google News search from January 2019 using keywords: ‘stablecoin’, ‘stablecoins’, ‘USD Coin’, ‘USDC’, ‘Tether’, and ‘USDT’.
  - Manual review and narrative judgment retained events originating within the stablecoin ecosystem and not driven by broader crypto or macro-financial developments.
  - Identified 50 distinct events.
  - Examples of retained events: GENIUS stablecoin bill introduced (February 4, 2025); Circle–Binance strategic partnership announcement (December 11, 2024); Fed Chair Powell remarks on overseeing stablecoins (June 21, 2024).
  - Excluded events originating outside the stablecoin ecosystem (examples: Silicon Valley Bank failure (March 10, 2023); FTX bankruptcy filing (November 11, 2022)).
- High-frequency market-capitalization measure
  - For each event date, change measured as ∆m_t = m_t − m_{t−1}, where m_t is the log market capitalization at end of day t.
  - Rationale: stablecoin prices are pegged, so market-capitalization fluctuations mainly reflect net issuance/redemption → proxy for changes in stablecoin demand.
  - Use daily changes as instruments for shocks to stablecoin demand (not as direct measures of underlying shock).
- Control-date selection and placebo diagnostics
  - For each event, a control date is randomly selected: less than two months away from the event date, but at least three days away from any event date.
  - Average absolute change in market capitalization:
    - Event dates: 0.42 percent
    - Control dates: 0.25 percent
  - Variance of event-date market responses is 2.18 times that of control series.
  - Brown-Forsythe test confirms the difference in variance is statistically significant with a p-value of less than 0.05.
- Event-study identification logic
  - Simultaneous equations model:
    - ∆m_t = β∆y_t + γF_t + ν_t
    - ∆y_t = α∆m_t + F_t + η_t
    - Key parameter of interest: α (causal effect of change in stablecoin market capitalization on outcome ∆y_t).
  - Identifying assumption: variance of idiosyncratic shock to stablecoin market capitalization (ν_t) is higher on event days than non-event days, while variances of common factor F_t and outcome idiosyncratic shock η_t are the same across event and control samples:
    - σ^2_{ν,S} > σ^2_{ν,~S}
    - σ^2_{F,S} = σ^2_{F,~S}
    - σ^2_{η,S} = σ^2_{η,~S}
  - Appendix B.1 contains algebraic derivations to estimate α using covariance matrices on event and control days.
- SVAR-IV implementation
  - Reduced-form VAR in levels at weekly frequency with baseline variables: log stablecoin market capitalization, 1-month U.S. Treasury yields, log BGCI, log S&P 500, log broad dollar index.
  - Reduced-form: y_t = α + B_1 y_{t−1} + ... + B_p y_{t−p} + u_t, with u_t covariance Σ.
  - Structural mapping: u_t = S ε_t, with ε_t mutually uncorrelated structural shocks and Ω = Var(ε_t).
  - Stablecoin shock denoted ε^d_t (first structural shock ε_{1,t}); first column s_1 of S captures impact of ε^d_t on u_t.
  - Use heteroskedasticity in market-capitalization changes on event days as an external instrument to estimate s_1 (details in Appendix B.2).
  - With estimated ŝ_1 and B̂_ℓ, compute impulse response functions and structural shock series.

### Max-Share Identification (Complementary VAR approach)
- Procedure
  - Estimate same reduced-form weekly VAR as narrative approach with variables: log stablecoin market cap, 1-month Treasury yields, log BGCI, log S&P 500, log broad dollar index.
  - Impose u_t = S ε_t and write S = S̃ Q where S̃ is Cholesky of Σ and Q is orthonormal rotation; identification reduces to finding Q.
  - For each non-stablecoin variable i, find q*_i that maximizes cumulative forecast-error variance (FEV) up to horizon H (baseline H = 13 weeks) for that variable:
    - q*_i = arg max_{q_i} Σ_{h=0}^H S_{i,i}(h) subject to q'_i q_i = 1.
  - Identify stablecoin shock column q*_1 by maximizing cumulative FEV of stablecoin market capitalization up to H subject to orthogonality to q*_i for all i in non-stablecoin set:
    - q*_1 = arg max_{q_1} Σ_{h=0}^H S_{1,1}(h) subject to q'_1 q_1 = 1 and q'_1 q*_i = 0 ∀ i ∈ J_{−1}.
  - Structural impact vector s_1 obtained as s_1 = S̃ q*_1.
  - Use s_1 and VAR coefficient matrices to compute impulse responses and structural shocks.

### Effects of Stablecoin Shocks on Financial Markets
- Estimation scope
  - Study both high-frequency (one-day) responses and dynamic weekly-frequency impulse responses.
  - Outcome variables in event-study regressions are in logs except U.S. Treasury yields, which are measured in basis points.
  - One-day windows used for event-study estimates.

### High-Frequency Causal Effects (Event-study results)
- Treasury yields (effects per 1 percent increase in stablecoin market capitalization)
  - 1-month Treasury yield: decrease of 0.423 basis points (coefficient: -0.423, standard error (0.233), significance: ∗ p<0.1).
  - 3-month Treasury yield: decrease of 0.498 basis points (coefficient: -0.498, standard error (0.174), significance: ∗∗∗ p<0.01).
  - 1-year Treasury yield: decrease of 0.328 basis points (coefficient: -0.328, standard error (0.284), statistically insignificant).
  - 10-year Treasury yield: decrease of 0.231 basis points (coefficient: -0.231, standard error (0.367), statistically insignificant).
  - Interpretation: Short-term yield declines are larger and statistically significant, consistent with stablecoin issuers’ reserves having large shares in short-term Treasury securities.
- Reserve composition notes (reported disclosures)
  - As of end-October 2025, 35 percent of USDC reserves consisted of U.S. Treasury securities maturing within three months; remainder consisted of cash and U.S. Treasury repurchase agreements.
  - As of most recent USDT reserve report in September 2025, 62 percent of USDT reserves were in the form of T-bills; USDT reserves include a more diverse range of assets including Bitcoin and commodities.
  - GENIUS Act stipulates future reserve holdings must be composed of highly liquid assets including Treasuries with a maturity of 93 days or less.
- Other financial and crypto market responses (per 1 percent increase in market capitalization)
  - Broad dollar index: weakens by 0.045 percent (coefficient: -0.045, standard error (0.021), significance: ∗∗ p<0.05).
    - Interpretation: net effect suggests lower U.S. Treasury yields provoke global portfolio rebalancing away from U.S. assets dominates potential strengthening through higher foreign demand for U.S. stablecoins.
  - Bloomberg Galaxy Crypto Index (BGCI): coefficient 0.239 (standard error (0.382), statistically insignificant).
  - S&P 500: coefficient 0.060 (standard error (0.060), statistically insignificant).
  - Nasdaq: coefficient 0.144 (standard error (0.092), statistically insignificant).
  - Insignificant BGCI response supports that identified stablecoin shocks are not capturing broader unbacked crypto-driven events.
- Equity responses of firms with potential stablecoin exposure (high-frequency)
  - Several payment providers show positive, statistically significant equity responses:
    - PayPal and Square: positive and statistically significant responses (exact coefficients and standard errors in Table 2 in source).
    - Adyen: positive and statistically significant response.
  - Crypto exchange:
    - Coinbase: largest sensitivity; a 1 percent increase in stablecoin market capitalization raises Coinbase equity price by 1.436 percent.
  - Major card networks and retailers:
    - Mastercard and Visa: no significant equity response.
    - Amazon and Walmart: no significant equity response.
  - Banking sector:
    - Nasdaq large banks index: statistically insignificant response.
    - Nasdaq community and small banks indexes: statistically insignificant responses.
    - No significant effects on equity prices of the ten largest U.S. banks or ten largest community banks (Tables C.1 and C.2 in source).

### 3.2 Dynamic Causal Effects of Stablecoin Shocks — Identification and Empirical Approach
- Narrative identification within an SVAR-IV using heteroskedasticity-based identification; shock scaled to correspond to a 1 percent increase in the stablecoin market capitalization at the peak of responses.
- Weak-instrument assessment: first-stage heteroskedasticity-robust F-statistic of 12.78, above the conventional critical value of 10.
- Statistical inference: residual-based moving block bootstrap standard errors from Jentsch and Lunsford (2019) with 10,000 replications; confidence intervals reported at 68 and 90 percent.
- Complementary identification: max-share approach (recovers structural disturbances that maximize forecast error variance of stablecoin market capitalization within the VAR) used as a robustness check.

### Key Dynamic Effects of a 1 percent Peak Stablecoin Market-Capitalization Shock
- Short-term U.S. Treasury yields:
  - 1-month U.S. Treasury yield decreases by 1.9 basis points at the trough of the response.
  - Effects concentrated at the short end of the curve; limited broader yield-curve implications.
- Exchange rate:
  - Broad dollar index declines by 0.09 percent at the trough.
  - Interpreted as a modest depreciation of the U.S. dollar consistent with portfolio rebalancing in response to lower U.S. T-bill yields.
- Crypto markets:
  - Bloomberg Galaxy Crypto Index (BGCI) rises by 1.5 percent at its peak, indicating positive spillovers into broader crypto markets.
  - Mechanism: shocks increasing stablecoin demand may facilitate easier and lower cost access to (unbacked) crypto trading.
- Equity markets:
  - S&P 500: peak effect of 0.25 percent (statistically significant but economically small); response around 0.1 percent in the week of impact.
  - Heterogeneous equity responses across market segments: payment providers with established stablecoin-based infrastructure show positive and statistically significant equity gains; large banks, community banks, and major retailers show no meaningful responses.

### Robustness and Sensitivity Analyses
- Max-share vs narrative identification:
  - Impulse responses from the max-share method closely mirror narrative-identification results in sign, shape, and magnitude across all variables.
- Event selection robustness:
  - Excluding events coinciding with major macroeconomic announcements or FOMC decisions does not change results; variables exhibit extremely similar responses.
  - Restricting to regulatory announcements about stablecoins yields nearly identical impulse responses.
  - Limiting analysis to twenty events with the largest absolute change in market capitalization yields similar results.
- VAR specification and variable substitutions:
  - Augmenting the VAR with gold price or VIX yields very similar responses in sign, shape, and magnitude.
  - Replacing the 1-month Treasury yield with 3-month Treasury yields, the BGCI with the Bitcoin price, and the S&P 500 with the Nasdaq-100 produces responses consistent with baseline estimates.
- Lag and horizon sensitivity:
  - Baseline uses lags corresponding to a quarter of a year (13 weeks); alternative specification uses half a year of lags (26 weeks).
  - Using 26 weeks of lags produces more volatile responses overall; Treasury yield responses remain very similar or larger in magnitude.
  - Using a 26-week horizon for the max-share approach results in a more volatile stablecoin market-capitalization response but overall effects across the 1.5-year horizon remain consistent in sign.

### Conclusions and Policy Implications
- Main conclusions
  - Stablecoin demand shocks exert downward pressure on short-maturity Treasury yields; a 1 percent peak stablecoin market-capitalization shock lowers the 1-month U.S. Treasury yield by 1.9 basis points.
  - Stablecoin shocks generate spillovers beyond Treasury markets, affecting crypto and equity valuations and producing a modest U.S. dollar depreciation.
  - Equity responses are heterogeneous: payment providers benefit, while large banks, community banks, and major retailers do not show material negative effects.
- Policy relevance
  - Stablecoin demand represents an emerging financial-transmission channel.
  - Policies affecting stablecoin issuance—cited examples include recent regulations such as the GENIUS Act and MiCA—may have nontrivial consequences for Treasury markets, exchange rates, and broader asset prices.
  - As stablecoin use in payments and tokenized financial services expands, monitoring interactions with traditional financial markets is important for assessing potential financial stability implications.

### Key Statistics and Experimental Settings (reported exactly as in source)
- Shock scale: 1 percent increase in the stablecoin market capitalization at the peak of responses.
- First-stage heteroskedasticity-robust F-statistic: 12.78.
- 1-month U.S. Treasury yield trough response: decreases by 1.9 basis points.
- Broad dollar index trough response: declines by 0.09 percent.
- Bloomberg Galaxy Crypto Index peak response: rises by 1.5 percent.
- S&P 500 peak effect: 0.25 percent; week-of-impact response around 0.1 percent.
- Bootstrap replications: 10,000.
- Confidence intervals reported: 68 and 90 percent.
- VAR lag choices reported: baseline 13 weeks; alternative 26 weeks.
- Event subset used in sensitivity: twenty events with the largest absolute change in market capitalization.

*Source: wpiea2026044-source-pdf - Section 4 concludes.*

### Section 4 concludes.

### Section 4 concludes.

### Data and Identification
- Sample and frequency
  - Sample spans January 1, 2019 to June 30, 2025.
  - Daily observations used for event-study analysis; aggregated to weekly frequency for VAR estimations.
  - Lag length p set to 13 weeks in VAR specifications.
- Data sources and variables
  - Stablecoin market capitalization: combined USDC and USDT from CoinGecko (log market capitalization used).
  - Financial market data: US Treasury yields, Nasdaq-100 index, Bloomberg Galaxy Crypto Index (BGCI), equity prices of payment providers/retailers/banks from Bloomberg; S&P 500 and broad dollar index from FRED.
  - High-frequency dataset includes stablecoin market capitalization, asset prices, and broader financial market indicators.
- Identification strategies overview
  - Primary: heteroskedasticity-based identification exploiting narrative-based stablecoin events in an event-study and SVAR-IV setting.
  - Complementary robustness: max-share identification (Uhlig 2004) within a VAR to recover the stablecoin shock orthogonal to other shocks.

### Narrative Identification (Event-study and SVAR-IV)
- Event selection and construction
  - Systematic week-by-week Google News search from January 2019 using keywords: ‘stablecoin’, ‘stablecoins’, ‘USD Coin’, ‘USDC’, ‘Tether’, and ‘USDT’.
  - Manual review and narrative judgment retained events originating within the stablecoin ecosystem and not driven by broader crypto or macro-financial developments.
  - Identified 50 distinct events.
  - Examples of retained events: GENIUS stablecoin bill introduced (February 4, 2025); Circle–Binance strategic partnership announcement (December 11, 2024); Fed Chair Powell remarks on overseeing stablecoins (June 21, 2024).
  - Excluded events originating outside the stablecoin ecosystem (examples: Silicon Valley Bank failure (March 10, 2023); FTX bankruptcy filing (November 11, 2022)).
- High-frequency market-capitalization measure
  - For each event date, change measured as ∆m_t = m_t − m_{t−1}, where m_t is the log market capitalization at end of day t.
  - Rationale: stablecoin prices are pegged, so market-capitalization fluctuations mainly reflect net issuance/redemption → proxy for changes in stablecoin demand.
  - Use daily changes as instruments for shocks to stablecoin demand (not as direct measures of underlying shock).
- Control-date selection and placebo diagnostics
  - For each event, a control date is randomly selected: less than two months away from the event date, but at least three days away from any event date.
  - Average absolute change in market capitalization:
    - Event dates: 0.42 percent
    - Control dates: 0.25 percent
  - Variance of event-date market responses is 2.18 times that of control series.
  - Brown-Forsythe test confirms the difference in variance is statistically significant with a p-value of less than 0.05.
- Event-study identification logic
  - Simultaneous equations model:
    - ∆m_t = β∆y_t + γF_t + ν_t
    - ∆y_t = α∆m_t + F_t + η_t
    - Key parameter of interest: α (causal effect of change in stablecoin market capitalization on outcome ∆y_t).
  - Identifying assumption: variance of idiosyncratic shock to stablecoin market capitalization (ν_t) is higher on event days than non-event days, while variances of common factor F_t and outcome idiosyncratic shock η_t are the same across event and control samples:
    - σ^2_{ν,S} > σ^2_{ν,~S}
    - σ^2_{F,S} = σ^2_{F,~S}
    - σ^2_{η,S} = σ^2_{η,~S}
  - Appendix B.1 contains algebraic derivations to estimate α using covariance matrices on event and control days.
- SVAR-IV implementation
  - Reduced-form VAR in levels at weekly frequency with baseline variables: log stablecoin market capitalization, 1-month U.S. Treasury yields, log BGCI, log S&P 500, log broad dollar index.
  - Reduced-form: y_t = α + B_1 y_{t−1} + ... + B_p y_{t−p} + u_t, with u_t covariance Σ.
  - Structural mapping: u_t = S ε_t, with ε_t mutually uncorrelated structural shocks and Ω = Var(ε_t).
  - Stablecoin shock denoted ε^d_t (first structural shock ε_{1,t}); first column s_1 of S captures impact of ε^d_t on u_t.
  - Use heteroskedasticity in market-capitalization changes on event days as an external instrument to estimate s_1 (details in Appendix B.2).
  - With estimated ŝ_1 and B̂_ℓ, compute impulse response functions and structural shock series.

### Max-Share Identification (Complementary VAR approach)
- Procedure
  - Estimate same reduced-form weekly VAR as narrative approach with variables: log stablecoin market cap, 1-month Treasury yields, log BGCI, log S&P 500, log broad dollar index.
  - Impose u_t = S ε_t and write S = S̃ Q where S̃ is Cholesky of Σ and Q is orthonormal rotation; identification reduces to finding Q.
  - For each non-stablecoin variable i, find q*_i that maximizes cumulative forecast-error variance (FEV) up to horizon H (baseline H = 13 weeks) for that variable:
    - q*_i = arg max_{q_i} Σ_{h=0}^H S_{i,i}(h) subject to q'_i q_i = 1.
  - Identify stablecoin shock column q*_1 by maximizing cumulative FEV of stablecoin market capitalization up to H subject to orthogonality to q*_i for all i in non-stablecoin set:
    - q*_1 = arg max_{q_1} Σ_{h=0}^H S_{1,1}(h) subject to q'_1 q_1 = 1 and q'_1 q*_i = 0 ∀ i ∈ J_{−1}.
  - Structural impact vector s_1 obtained as s_1 = S̃ q*_1.
  - Use s_1 and VAR coefficient matrices to compute impulse responses and structural shocks.

### Effects of Stablecoin Shocks on Financial Markets
- Estimation scope
  - Study both high-frequency (one-day) responses and dynamic weekly-frequency impulse responses.
  - Outcome variables in event-study regressions are in logs except U.S. Treasury yields, which are measured in basis points.
  - One-day windows used for event-study estimates.

### High-Frequency Causal Effects (Event-study results)
- Treasury yields (effects per 1 percent increase in stablecoin market capitalization)
  - 1-month Treasury yield: decrease of 0.423 basis points (coefficient: -0.423, standard error (0.233), significance: ∗ p<0.1).
  - 3-month Treasury yield: decrease of 0.498 basis points (coefficient: -0.498, standard error (0.174), significance: ∗∗∗ p<0.01).
  - 1-year Treasury yield: decrease of 0.328 basis points (coefficient: -0.328, standard error (0.284), statistically insignificant).
  - 10-year Treasury yield: decrease of 0.231 basis points (coefficient: -0.231, standard error (0.367), statistically insignificant).
  - Interpretation: Short-term yield declines are larger and statistically significant, consistent with stablecoin issuers’ reserves having large shares in short-term Treasury securities.
- Reserve composition notes (reported disclosures)
  - As of end-October 2025, 35 percent of USDC reserves consisted of U.S. Treasury securities maturing within three months; remainder consisted of cash and U.S. Treasury repurchase agreements.
  - As of most recent USDT reserve report in September 2025, 62 percent of USDT reserves were in the form of T-bills; USDT reserves include a more diverse range of assets including Bitcoin and commodities.
  - GENIUS Act stipulates future reserve holdings must be composed of highly liquid assets including Treasuries with a maturity of 93 days or less.
- Other financial and crypto market responses (per 1 percent increase in market capitalization)
  - Broad dollar index: weakens by 0.045 percent (coefficient: -0.045, standard error (0.021), significance: ∗∗ p<0.05).
    - Interpretation: net effect suggests lower U.S. Treasury yields provoke global portfolio rebalancing away from U.S. assets dominates potential strengthening through higher foreign demand for U.S. stablecoins.
  - Bloomberg Galaxy Crypto Index (BGCI): coefficient 0.239 (standard error (0.382), statistically insignificant).
  - S&P 500: coefficient 0.060 (standard error (0.060), statistically insignificant).
  - Nasdaq: coefficient 0.144 (standard error (0.092), statistically insignificant).
  - Insignificant BGCI response supports that identified stablecoin shocks are not capturing broader unbacked crypto-driven events.
- Equity responses of firms with potential stablecoin exposure (high-frequency)
  - Several payment providers show positive, statistically significant equity responses:
    - PayPal and Square: positive and statistically significant responses (exact coefficients and standard errors in Table 2 in source).
    - Adyen: positive and statistically significant response.
  - Crypto exchange:
    - Coinbase: largest sensitivity; a 1 percent increase in stablecoin market capitalization raises Coinbase equity price by 1.436 percent.
  - Major card networks and retailers:
    - Mastercard and Visa: no significant equity response.
    - Amazon and Walmart: no significant equity response.
  - Banking sector:
    - Nasdaq large banks index: statistically insignificant response.
    - Nasdaq community and small banks indexes: statistically insignificant responses.
    - No significant effects on equity prices of the ten largest U.S. banks or ten largest community banks (Tables C.1 and C.2 in source).

*Source: wpiea2026044-source-pdf - Section 4 concludes.*

### 3.2    Dynamic Causal Effects of Stablecoin Shocks

### 3.2    Dynamic Causal Effects of Stablecoin Shocks

### Identification and empirical approach
- Narrative identification within an SVAR-IV using heteroskedasticity-based identification; shock scaled to correspond to a 1 percent increase in the stablecoin market capitalization at the peak of responses.
- Weak-instrument assessment: first-stage heteroskedasticity-robust F-statistic of 12.78, above the conventional critical value of 10.
- Statistical inference: residual-based moving block bootstrap standard errors from Jentsch and Lunsford (2019) with 10,000 replications; confidence intervals reported at 68 and 90 percent.
- Complementary identification: max-share approach (recovers structural disturbances that maximize forecast error variance of stablecoin market capitalization within the VAR) used as a robustness check.

### Key dynamic effects of a 1 percent peak stablecoin market-capitalization shock
- Short-term U.S. Treasury yields:
  - 1-month U.S. Treasury yield decreases by 1.9 basis points at the trough of the response.
  - Effects concentrated at the short end of the curve; limited broader yield-curve implications.
- Exchange rate:
  - Broad dollar index declines by 0.09 percent at the trough.
  - Interpreted as a modest depreciation of the U.S. dollar consistent with portfolio rebalancing in response to lower U.S. T-bill yields.
- Crypto markets:
  - Bloomberg Galaxy Crypto Index (BGCI) rises by 1.5 percent at its peak, indicating positive spillovers into broader crypto markets.
  - Mechanism: shocks increasing stablecoin demand may facilitate easier and lower cost access to (unbacked) crypto trading.
- Equity markets:
  - S&P 500: peak effect of 0.25 percent (statistically significant but economically small); response around 0.1 percent in the week of impact.
  - Heterogeneous equity responses across market segments: payment providers with established stablecoin-based infrastructure show positive and statistically significant equity gains; large banks, community banks, and major retailers show no meaningful responses.

### Robustness and sensitivity analyses
- Max-share vs narrative identification:
  - Impulse responses from the max-share method closely mirror narrative-identification results in sign, shape, and magnitude across all variables.
- Event selection robustness:
  - Excluding events coinciding with major macroeconomic announcements or FOMC decisions does not change results; variables exhibit extremely similar responses.
  - Restricting to regulatory announcements about stablecoins yields nearly identical impulse responses.
  - Limiting analysis to twenty events with the largest absolute change in market capitalization yields similar results.
- VAR specification and variable substitutions:
  - Augmenting the VAR with gold price or VIX yields very similar responses in sign, shape, and magnitude.
  - Replacing the 1-month Treasury yield with 3-month Treasury yields, the BGCI with the Bitcoin price, and the S&P 500 with the Nasdaq-100 produces responses consistent with baseline estimates.
- Lag and horizon sensitivity:
  - Baseline uses lags corresponding to a quarter of a year (13 weeks); alternative specification uses half a year of lags (26 weeks).
  - Using 26 weeks of lags produces more volatile responses overall; Treasury yield responses remain very similar or larger in magnitude.
  - Using a 26-week horizon for the max-share approach results in a more volatile stablecoin market-capitalization response but overall effects across the 1.5-year horizon remain consistent in sign.

### Conclusions and policy implications
- Three main conclusions:
  - Stablecoin demand shocks exert downward pressure on short-maturity Treasury yields; a 1 percent peak stablecoin market-capitalization shock lowers the 1-month U.S. Treasury yield by 1.9 basis points.
  - Stablecoin shocks generate spillovers beyond Treasury markets, affecting crypto and equity valuations and producing a modest U.S. dollar depreciation.
  - Equity responses are heterogeneous: payment providers benefit, while large banks, community banks, and major retailers do not show material negative effects.
- Policy relevance:
  - Stablecoin demand represents an emerging financial-transmission channel.
  - Policies affecting stablecoin issuance—cited examples include recent regulations such as the GENIUS Act and MiCA—may have nontrivial consequences for Treasury markets, exchange rates, and broader asset prices.
  - As stablecoin use in payments and tokenized financial services expands, monitoring interactions with traditional financial markets is important for assessing potential financial stability implications.

### Key statistics and experimental settings (reported exactly as in source)
- Shock scale: 1 percent increase in the stablecoin market capitalization at the peak of responses.
- First-stage heteroskedasticity-robust F-statistic: 12.78.
- 1-month U.S. Treasury yield trough response: decreases by 1.9 basis points.
- Broad dollar index trough response: declines by 0.09 percent.
- Bloomberg Galaxy Crypto Index peak response: rises by 1.5 percent.
- S&P 500 peak effect: 0.25 percent; week-of-impact response around 0.1 percent.
- Bootstrap replications: 10,000.
- Confidence intervals reported: 68 and 90 percent.
- VAR lag choices reported: baseline 13 weeks; alternative 26 weeks.
- Event subset used in sensitivity: twenty events with the largest absolute change in market capitalization.

*Source: 3.2 Dynamic Causal Effects of Stablecoin Shocks (wpiea2026044-source-pdf).*

### References

### References

### Key bibliographic citations
- Ahmed, Rashad and I ̃naki Aldasoro, “Stablecoins and safe asset prices,” Technical Report, Bank for International Settlements 2025.
- , James A Clouse, Fabio Natalucci, Alessandro Rebucci, and Geyue Sun, “Stablecoins: A Revolutionary Payment Technology with Financial Risks,” Technical Report, National Bureau of Economic Research 2025.
- Aldasoro, Inaki, Giulio Cornelli, Massimo Ferrari Minesso, Leonardo Gambacorta, and Maurizio Michael Habib, “Stablecoins, money market funds and monetary policy,” Economics Letters, 2025, 247, 112203.
- Auer, Raphael, Ulf Lewrick, and Jan Paulick, “DeFiying gravity? An empirical analysis of cross-border Bitcoin, Ether and stablecoin flows,” Technical Report, Bank for International Settlements 2025.
- Barbon, Andrea, Jean Barthélemy, and Benoît Nguyen, “DeFi-ying the Fed? Monetary Policy Transmission to Stablecoin Rates,” Monetary Policy Transmission to Stablecoin Rates (December 22, 2023), 2023.
- Barsky, Robert B. and Eric R. Sims, “News shocks and business cycles,” Journal of Monetary Economics, 2011, 58(3), 273–289.
- Barthélemy, Jean, Paul Gardin, and Benoit Nguyen, “Stablecoins and short-term funding markets,” Journal of International Money and Finance, 2025, p. 103469.
- Boer, Lukas, Lukas Menkhoff, and Malte Rieth, “The multifaceted impact of US trade policy on financial markets,” Journal of Applied Econometrics, 2023, 38(3), 388–406.
- Cerutti, Eugenio, Melih Firat, and Martina Hengge, “Global Cross-Border Payments: A $1 Quadrillion Evolving Market?,” Technical Report, IMF Working Paper 2025/120. 2025.
- Cesa-Bianchi, Ambrogio and Marco Pinchetti, “Crypto Shocks,” Technical Report, Bank of England 2025.
- Gorton, Gary B, Elizabeth C Klee, Chase P Ross, Sharon Y Ross, and Alexandros P Vardoulakis, “Leverage and stablecoin pegs,” Journal of Financial and Quantitative Analysis, 2022, pp. 1–65.
- Hébert, Benjamin and Jesse Schreger, “The Costs of Sovereign Default: Evidence from Argentina,” American Economic Review, October 2017, 107(10), 3119–45.
- Jentsch, Carsten and Kurt G. Lunsford, “The Dynamic Effects of Personal and Corporate Income Tax Changes in the United States: Comment,” American Economic Review, July 2019, 109(7), 2655–78.
- Kim, Sang Rae, “Macro-Financial Impact of Stablecoin’s Demand for Treasuries,” Available at SSRN 5259528, 2025.
- , “What Happens When Stablecoins Are Introduced? Evidence from Short-Term Funding Market,” 2025.
- Känzig, Diego R., “The Macroeconomic Effects of Oil Supply News: Evidence from OPEC Announcements,” American Economic Review, April 2021, 111(4), 1092–1125.
- Montiel Olea, José Luis, James H. Stock, and Mark W. Watson, “Inference in Structural Vector Autoregressions identified with an external instrument,” Journal of Econometrics, 2021, 225(1), 74–87. Themed Issue: Vector Autoregressions.
- Nigrinis, Andrew, “The Lending Impact of Stablecoin-Induced Deposit Outflows,” 2025.
- Rigobon, Roberto, “Identification Through Heteroskedasticity,” The Review of Economics and Statistics, 11 2003, 85(4), 777–792.
- and Brian Sack, “The impact of monetary policy on asset prices,” Journal of Monetary Economics, 2004, 51(8), 1553–1575.
- Stock, James H. and Mark W. Watson, “Identification and Estimation of Dynamic Causal Effects in Macroeconomics Using External Instruments,” The Economic Journal, 05 2018, 128(610), 917–948.
- Tsyrennikov, Viktor, “Stablecoins’ impact on community bank deposits,” 2025.
- Uhlig, Harald, “Do technology shocks lead to a fall in total hours worked?,” Journal of the European Economic Association, 2004, 2(2-3), 361–371.
- Wright, Jonathan H, “What does monetary policy do to long-term interest rates at the zero lower bound?,” The Economic Journal, 2012, 122(564), F447–F466.

### Appendix A — Data
H3: Data sources and coverage
- Stablecoin market cap: USDT + USDC market capitalization, CoinGecko, Jan 1, 2019–Jun 30, 2025, Log.
- U.S. Treasury yields: 1m, 3m, 6m, 1y, 2y, and 10y Treasury yields, Bloomberg, Jan 1, 2019–Jun 30, 2025, Bps.
- Bloomberg Galaxy Crypto Index: Benchmark index of major crypto assets traded in USD, Bloomberg, Jan 1, 2019–Jun 30, 2025, Log.
- S&P 500: Broad U.S. equity index of about 500 large public companies, FRED, Jan 1, 2019–Jun 30, 2025, Log.
- Nasdaq-100: Tracks the 100 largest publicly traded non-financial firms listed on Nasdaq, Bloomberg, Jan 1, 2019–Jun 30, 2025, Log.
- Nominal Broad Dollar Index: Trade-weighted value of the U.S. dollar vs. major trading partners, FRED, Jan 1, 2019–Jun 30, 2025, Log.
- VIX: 30-day expected volatility of the U.S. stock market, FRED, Jan 1, 2019–Jun 30, 2025, Log.
- Gold price: Daily spot price of gold in USD per troy ounce, FRED, Jan 1, 2019–Jun 30, 2025, Log.
- Bitcoin price: Daily USD price of Bitcoin, Chainalysis, Jan 1, 2019–Jun 30, 2025, Log.
- Firm and bank equity prices: Daily equity price indexes for firms and banks, Bloomberg, Jan 1, 2019–Jun 30, 2025, Log.

H3: Stablecoin news events (selected entries and metrics)
- Notes: The market capitalization change is calculated as the daily change in the combined market capitalization of USDC and USDT on event days.
- Examples from Table A.2 (Date — Description — Market Cap Change (%)):
  - 4/1/2019 — Launch of TrueGBP (British stablecoin) — 0.317
  - 4/25/2019 — Bitfinex covers $850M loss with Tether funds — -0.704
  - 5/21/2019 — Tether reveals that its reserves include Bitcoin and other assets — -0.173
  - 8/27/2020 — USDC upgrade allows for ’gasless sends’ — 1.368
  - 1/4/2021 — US federal regulator says banks can conduct payments using stablecoins — 1.893
  - 2/5/2021 — Tether freezes $1.7 Million in response to hack — 1.474
  - 4/11/2024 — Circle enables USDC transfers for BlackRock’s first tokenized fund — 0.143
  - 1/21/2025 — Circle acquires Hashnote — 1.621
  - 2/4/2025 — US Senator Hagerty introduces GENIUS stablecoin bill — 0.803
  - 6/17/2025 — Senate passes GENIUS Act — -0.067

### Appendix B — Methodology
H3: Event-study methodology (B.1)
- Estimation target: coefficient α from simultaneous equations (Equations 2–3) using covariance matrices of ∆y_t and ∆m_t on event days (S) and control days (˜S).
- Covariance matrix definition: Ω_j = [ Var_j(∆y_t)   Cov_j(∆y_t,∆m_t) ; Cov_j(∆y_t,∆m_t)   Var_j(∆m_t) ] .(16)
- Difference in covariance matrices: ∆Ω = Ω_S − Ω_˜S = (σ_ν,S − σ_ν,˜S)/(1−αβ)^2 [ α^2  α ; α  1 ].(17)
- Identification formula example: α = ∆Ω_1,2 / ∆Ω_2,2 = (Cov_S(∆y_t,∆m_t) − Cov_˜S(∆y_t,∆m_t)) / (Var_S(∆m_t) − Var_˜S(∆m_t)).(18)
- Estimation approach: Implement instrumental variables (IV) regression from Rigobon and Sack (2004) using ∆m and instrument ∆˜m; the IV coefficient corresponds to the sample analogue of α.

H3: SVAR methodology (B.2)
- Model for stablecoin market capitalization movements: ∆m_t = ε_1,t + Σ_{i≠1} ε_i,t + ξ_t, where ξ_t ~ N(0, σ_ξ^2) captures measurement error.(19)
- Heteroskedasticity assumptions for identification:
  - σ^2_{ε1,S} > σ^2_{ε1,˜S}.(20)
  - σ^2_{εi,S} = σ^2_{εi,˜S} for i = 2,...,n.(21)
  - σ^2_{ξ,S} = σ^2_{ξ,˜S}.(22)
- Control sample ˜S: dates at least three days away from any event date.
- Structural impact vector formula: s_1 = (E_S[∆m_t u_t] − E_˜S[∆m_t u_t]) / (E_S[(∆m_t)^2] − E_˜S[(∆m_t)^2]).(23)
- Sample analogue: ˆs_1 = (U′∆m_S − U′∆m_˜S) / (∆m′_S ∆m_S − ∆m′_˜S ∆m_˜S).(24)
- Data frequency: ∆m_t is aggregated weekly from daily responses.

H3: Shock series and identification (B.3)
- Figure B.1 presents weekly stablecoin shock series from two complementary methods: heteroskedasticity-based identification in a SVAR-IV and the max-share approach.
- Notes: The figure displays the weekly shock series labeled "Stablecoin Shock" with narrative identification and "Max-Share".

### Appendix C — Additional event-study results
H3: High-frequency equity index responses (C.1)
- Table C.1: Event-study one-day window estimates using heteroskedasticity-based identification (heteroskedasticity-robust standard errors).
  - Nasdaq: coefficient 0.060 (standard error 0.116), Observations 1447.
  - Large Banks: coefficient -0.031 (standard error 0.120), Observations 1447.
  - Comm. Banks: coefficient -0.028 (standard error 0.120), Observations 1447.
- Significance levels reported: ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01.

H3: Financial institution equity responses (C.2)
- Table C.2: One-day window event-study (heteroskedasticity-based identification) for top non-community and community banks by U.S. market capitalization; heteroskedasticity-robust standard errors.
- Selected coefficients (Stablecoin Market Cap) and standard errors:
  - J.P. Morgan Chase (1): -0.010 (0.110), Observations 1404.
  - Bank of America (2): 0.094 (0.120), Observations 1447.
  - Wells Fargo (3): 0.085 (0.140), Observations 1404.
  - Citi (4): 0.041 (0.154), Observations 1226.
  - PNC (5): 0.018 (0.133), Observations 1226.
  - [Additional banks listed with their coefficients, standard errors, and observation counts in the table.]

### Appendix D — Additional impulse responses
H3: Alternative event selection (D.1)
- Figures D.1–D.3: Impulse responses to a stablecoin shock using heteroskedasticity-based identification in a SVAR-IV with different event selection rules:
  - Excluding economic release dates (Figure D.1).
  - Regulatory events only (Figure D.2).
  - Strongest events only (Figure D.3).
- Shock scaling: Each shock is scaled to correspond to a 1 percent increase in the stablecoin market capitalization at the peak of responses.
- Confidence intervals: Dark and light shaded areas correspond to the 68 and 90 percent confidence intervals, respectively.
- Standard errors: Residual-based moving block bootstrap standard errors from Jentsch and Lunsford (2019) with 10,000 replications.

H3: Alternative specifications and robustness (D.2–D.4)
- Figures D.4–D.8: Robustness checks adding alternative variables:
  - Adding gold prices (Figure D.4).
  - Adding VIX (Figure D.5).
  - Using 3-month Treasury yields (Figure D.6).
  - Using Bitcoin price (Figure D.7).
  - Using Nasdaq-100 (Figure D.8).
- Figures D.9: Varying lag order (26 lags).
- Figures D.10: Varying horizon length (horizon length of 26 weeks) and comparison with the max-share approach.
- Common details across figures:
  - Shock scaled to a 1 percent peak increase in stablecoin market capitalization.
  - Solid lines: point estimates.
  - Dark/light shading: 68 and 90 percent confidence intervals.
  - Standard errors: residual-based moving block bootstrap with 10,000 replications.
  - Outcome variables shown include: Stablecoin Market Cap, 1M Treasury Yield (or 3M where noted), Bloomberg Galaxy Crypto Index, S&P 500, Broad Dollar Index, Gold Price, VIX, Bitcoin Price, NASDAQ/Nasdaq-100 as indicated per figure.

*Stablecoin Shocks — Working Paper No. WP/2026/044*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026044-source-pdf.pdf_