## 1. CIP Deviation Relative to US Dollar

## Source details

**Canonical URL:** [1. CIP Deviation Relative to US Dollar](https://www.imf.org/-/media/files/publications/gfsr/2025/october/english/ch2annex.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/gfsr/2025/october/english/ch2annex.pdf.md)
- [Structured JSON version](/-/media/files/publications/gfsr/2025/october/english/ch2annex.pdf.json)

---

### Data, scope, and definitions
- Sample period: January 1, 2015, to May 31, 2025.
- Data sources: Bloomberg Finance L.P.; Refinitiv; CLS; BIS; Treasury International Capital (TIC); IMF staff calculations.
- CIP deviation (panel 1): calculated using 5-year government rates for 10 emerging market economies and 12 advanced economies against the US dollar. A negative widening basis signals stress in dollar funding markets.
- Bid-ask spread (panel 2): calculated as (ask rate-bid rate)/mid rate (in percent). Sample includes 19 emerging market economies and 18 advanced economies.
- Excess exchange rate return (panel 3): defined as log(exchange rate at time t / exchange rate at time t-1) - log(forward rate at time t-1 / exchange rate at time t-1).
- Acronyms: AEs = advanced economies; EMs = emerging markets; GFC = global financial crisis; OIS = Overnight index swaps; CIP = covered interest parity; NBFI = non-bank financial institution; CCFR = cross-currency funding (gap) ratio; NIP = net foreign investment position.

### Transaction volumes: measurement and formulas
- Spot market gross inflows for currency c, country i, sector s in period t:
  - InFlows_c,i,s,t = ∑_{j≠i} ∑_{c'} max(FX_Flows_{c,c',i,j,s,t}, 0)  (equation 2.3.5)
  - OutFlows_c,i,s,t = ∑_{j≠i} ∑_{c'} min(FX_Flows_{c,c',i,j,s,t}, 0)  (equation 2.3.6)
  - FX_Flows_{c,c',i,j,s,t}: value of an FX spot transaction between sector s in country i and banking sector in country j; positive if sector s in country i receives currency c against payment in currency c’.
- Swap market gross inflows/outflows for currency c:
  - SwapInFlows_c,i,s,t = ∑_{j≠i} ∑_{c'} max(SWAP_Flows_{c,c',i,j,s,t}, 0)  (equation 2.3.7)
  - SwapOutFlows_c,i,s,t = ∑_{j≠i} ∑_{c'} min(SWAP_Flows_{c,c',i,j,s,t}, 0)  (equation 2.3.8)
  - NetSwapFlows_c,i,s,t = SwapInFlows_c,i,s,t − SwapOutFlows_c,i,s,t
- Outstanding open swap positions (accumulation of gross flows not yet matured as of t):
  - LongSwapPositions_{c,c',i,s,t} = ∑_{τ≤t; Maturity(τ)>t} SwapInFlows_{c,c',i,s,τ}  (equation 2.3.9)
  - ShortSwapPositions_{c,c',i,s,t} = ∑_{τ≤t; Maturity(τ)>t} SwapInFlows_{c,c',i,s,τ}  (equation 2.3.10)
  - Country-level aggregation: sum across sectors s (equations 2.3.11–2.3.12).
- Note: CLS membership mainly consists of large banks; every transaction flow from an institutional sector is recorded predominantly with banks.

### Stylized facts: USD spot and swap markets
- USD spot market:
  - US and non-US dealer banks are central market makers facilitating transactions among clients, including non-dealer banks.
  - Under normal conditions, US banks exhibit negative net USD flows (they act as providers of USD funding to other participants); non-US banks tend to exhibit positive net USD flows.
  - Non-banks domiciled outside the US are most often net buyers of USD; their trading patterns are highly sensitive to global macro-financial conditions.
  - US non-banks exhibit less distinctive patterns and can be net buyers or net sellers of USD at times.
- USD swap market (tenors ≥ 35 days):
  - Net demand for USD funding and hedging is driven mostly by non-banks; net trading flows for non-banks are mostly negative (they receive USD in the spot leg and sell USD in the forward leg).
  - US banks perform maturity transformation: borrowing dollars short term from other banks and lending to non-banks seeking longer-term hedges.
  - Aggregated flow data show US banks are net dollar borrowers in the FX swap market despite being the largest gross providers of USD, due to intermediation between short-term funding and longer-term hedging demand.
  - These aggregate flow patterns may differ when measured using outstanding positions (which capture full balance sheet exposures and reduce overrepresentation of positions rolled over at short maturities).

### Structural changes and vulnerability measures
- Dealer concentration (HHI):
  - HHI_{c,c',t} = ∑_{j, s, ī} ( |FX_Flows_{c,c',ī,j,s,t}| / ∑_{j,s,ī} |FX_Flows_{c,c',ī,j,s,t}| )^2  (equation 2.3.13)
  - ī denotes flows from/to country-sectors with dealer banks identified by the Federal Bank of New York primary dealers list (Canada, France, Germany, Japan, Switzerland, the UK, and the US).
- Share of NBFI flows:
  - ShareNBFI_{c,c',t} = (∑_{j,i} (|FX_Flows_{c,c',i,j,Fund,t}| + |FX_Flows_{c,c',i,j,OtherNBFI,t}|)) / (∑_{j,s,i} |FX_Flows_{c,c',i,j,s,t}|)  (equation 2.3.14)
  - Denominator excludes bank-to-bank flows.
- Currency mismatch (“hedging pressure”) for USD against currency c', sector s:
  - HedgingPressure_{USD,c',s,t} = ( (∑_i ShortSwapPositions_{USD,c',i,s,t} − ∑_i LongSwapPositions_{USD,c',i,s,t}) / (1/52 ∑_{k=0}^{51} ∑_{s,i} OI_{USD,c',i,s,t−k} ) )  (equation 2.3.15)
  - Example: funds are a counterparty in 63 percent of all outstanding interest in forwards for the EURUSD rate.
- Cross-currency funding (gap) ratio (CCFR) / USD FX mismatch:
  - USD_FX_Mismatch_{i,t} = (USD_Assets_{i,t} − USD_Liabilities_{i,t}) / USD_Assets_{i,t}  (equation 2.3.16)
  - USD_Assets_i,t and USD_Liabilities_i,t derived from BIS Locational Banking Statistics at quarterly frequency.
- Net foreign investment position (NIP):
  - NIP_{i,t} = (ForeignPositions_in_US_bonds_{i,t} − USPositions_in_Foreign_bonds_{i,t}) / (ForeignPositions_in_US_bonds_{i,t} + USPositions_in_Foreign_bonds_{i,t})  (equation 2.3.17)
  - Positive NIP indicates foreign holdings of US bonds exceed US holdings of foreign bonds for country i.
- CCFR and NIP are aggregated at the currency level using GDP-weighted averages.

### Empirical methodology for macrofinancial uncertainty shocks
- Uncertainty measures and shock definitions:
  - Economic Policy Uncertainty (EPU) — US index. Shock D_UNCRT_t: AR(1) residual for EPU exceeds two standard deviations above its mean.
  - Financial uncertainty — VIX index. Shock D_FINUNCRT_t: AR(1) residual of VIX exceeds two standard deviations above its mean.
  - Monetary policy uncertainty — MOVE index. Shock D_MOVE_t: AR(1) residual exceeds two standard deviations above its mean.
  - Robustness: alternative monetary policy uncertainty proxy D_MPU_t derived from US EPU index.
- Econometric specification (weekly frequency, January 1, 2015–May 31, 2025):
  - Δlog(InFlow_{c,i,s,t+h}) = α_{c,i,s}^h + λ_{c,i,t}^h + β_c^h · GShock_t + γ_c^h · Control_{c,i,s,t} + δ_c^h · Global_t^h + ν_{c,i,s,t}^h  (equation 2.4.1)
  - Controls: domestic country term spread (10-year minus 3-month), exchange rate for currency c; FX swap inflow models also control for 3-month CIP deviation.
  - Global_t controls: commodity price index, Chicago Financial Condition Index, US term spread (US 10-year minus 3-month), end-of-quarter dummies.
  - Fixed effects: α_{c,i,s} (country-sector) and λ_{c,i,t} (country-month-year). Standard errors clustered at the country level.
  - Models estimated for spot and swap transactions and across sectors; focus mainly on USD with CHF and JPY included for safe-haven comparisons.

### Empirical results and quantitative findings
- Hedging pressure drivers:
  - Beyond term spread differentials, hedging pressure is closely linked to the USD net investment position (NIP).
  - Quantitative estimate: a one–percentage point change in NIP corresponds to a 1.2 percentage point change in hedging pressure.
- Behavioral patterns under macrofinancial uncertainty:
  - When macrofinancial uncertainty rises, nonresident NBFIs increase demand for safe-haven assets, including USD, EUR, and CHF; net spot purchases of EUR and CHF rise markedly.
  - Unconditional correlations between net spot flows of NBFIs to safe-haven currencies and uncertainty measures (VIX, EPU, MOVE) reported for 2015M1–2025M5.
- Sectoral flow dynamics:
  - Spot market 4-quarter moving-average USD net flows shown separately for US banks, US nonbank institutions, non-US banks, and non-US nonbank institutions (measures of uncertainty standardized).
  - Long-term swap (tenor ≥ 35 days) USD net flows by sector indicate non-banks drive net demand for USD hedging.
- Robustness:
  - Baseline estimates using D_MPU_t mirror main-text patterns (robustness figures referenced).

### Change in USD spot flows following monetary policy uncertainty shocks
- Methodology:
  - Shocks constructed by regressing the U.S. EPU index on alternative measures of monetary policy surprises following Gürkaynak, Sack, and Swanson (2005) and Nakamura and Steinsson (2018).
  - Panel-model estimates of weekly changes in USD inflows in the spot market for non-US financial and nonfinancial institutions across 15 jurisdictions.
  - Specification controls: Chicago Financial Conditions Index, commodity price index, U.S. term spread, domestic term spreads, the spot exchange rate; country-sector and country-time fixed effects.
  - Uncertainty shocks defined as dummy = 1 when AR(1) residual exceeds two standard deviations.
  - Some regressions use data from January 1, 2006–May 31, 2025.
- Key findings on USD and safe-haven currency flows:
  - Heightened global uncertainty increases demand for safe-haven currencies (CHF, JPY, EUR) by nonresidents.
  - VIX shock estimated to increase the volume of transactions in the spot market for JPY by about 40 percent.
  - Market size and absolute effects:
    - Average weekly USD spot transactions immediately preceding the COVID-19 outbreak: roughly USD 2.2 trillion.
    - A 21 percent rise in USD spot transactions during the crisis corresponded to an additional USD 0.46 trillion in trading volume.
    - Average weekly JPY spot transactions: about USD 0.6 trillion.
    - A 42 percent increase in JPY spot transactions amounted to USD 0.25 trillion.
  - Despite larger percentage changes in JPY turnover, the increase in USD inflows was larger in absolute terms.
- Sectoral differences:
  - NBFIs show the strongest increases in demand for safe-haven currencies after global uncertainty shocks.
    - Following VIX shocks, volume transacted by NBFIs in the JPY spot market increases by 55 percent, twice as much as the volume transacted by banks.
    - A MOVE shock increases the volume transacted by NBFIs in the CHF spot market by 47 percent; the effect for banks is below 20 percent.
  - NBFIs are more sensitive to large domestic financial uncertainty shocks; their swap-market response occurs with a one-period lag relative to the spot market.
- Macro-financial disconnect (US EPU vs VIX):
  - When US EPU is high while VIX is contained (EPU–VIX disconnect is wider), nonresident investors may reduce USD purchases; effect is nonlinear and increases with magnitude of the disconnect.
  - Institutions also decrease hedging activity in the swap market in response to the disconnect.
- Persistence and cross-country heterogeneity:
  - Impact of VIX shocks peaks at around four weeks and persists for up to 16 weeks; similar for US EPU shocks.
  - Emerging markets face stronger and more persistent impacts:
    - Treasury CIP deviation widens by 16 basis points on average.
    - Bid-ask spreads increase by 0.04 basis points on average.
    - Excess exchange rate return volatility rises by 0.2 percentage points on average.
  - CIP deviation in EMs interpreted as a signal of dollar funding stress when OIS rates unavailable; regressions control for expected default frequency of banking sector and are robust to sovereign CDS-based controls.
- Robustness and extensions:
  - Checks include alternative fixed-effect specifications, end-of-month effects, presence of FX swap lines, inclusion of FX implied volatility, LIBOR–OIS spread, USD broad index, and alternative clustering; results remain robust.
  - Pre-tariff-period sample (2015:M1–2024:M12) excluding unusually high EPU around April 2 tariff announcements yields consistent results.
  - Equation (2.5.1) estimated across weekly horizons h = 0,...,16 weeks for 11 currency pairs over 2006-01-01 to 2025-05-31.
  - Extensions test interactions with structural vulnerabilities (dealer concentration, currency mismatches, NBFI share) and policy measures (swap-line dummy, international reserves/GDP).
  - Standard errors computed using Newey–West HAC estimator.

### Effect of an increase in VIX on CIP and FX market conditions
- Outcome variables analyzed over weekly horizons:
  - weekly 3-months OIS CIP deviation (Basis points),
  - excess exchange rate return volatility (Percentage points),
  - bid-ask spreads (Percent).
- High uncertainty shocks: dummy = 1 when AR(1) residual exceeds two standard deviations.
- Economies with reserve buffers about one standard deviation above the average:
  - Show markedly smaller CIP deviations and lower excess exchange rate return volatility after macro-financial uncertainty shocks.
- Robustness:
  - Fixed-effects variations, additional controls (end-of-month/end-of-quarter), inclusion of 3-months FX option-implied volatility and LIBOR–OIS spread tested.
  - Identification: Granular instrumental variables (GIV) methodology (Gabaix and Koijen 2023) used; residualization via principal component analysis also applied.
  - Pre-tariff robustness: analysis repeated excluding 2025 observations (sample up to 2024M12); results broadly consistent.
- Final robustness takeaway:
  - Results remain broadly consistent across specifications, additional controls, and identification strategies including GIV and residualization.

### Net yen cumulative spot flows by sector and nationality
- Data source: CLS Group; IMF staff calculations.
- Panels 1 and 2: cumulative flows beginning on November 1, 2024. Panels 3 and 4: cumulative flows starting from October 2019.
- Tariff-related event arrows: April 3, 2025 (Panels 1 and 2); COVID-19 market turmoil: March 9, 2020.
- Country codes: ISO 3-digits; EA (DE, FR, IT)=Germany, France, and Italy; EA Other = euro area except Germany, France, and Italy; ROW = Rest of the World.
- Key findings on yen flows:
  - Net purchases by foreign institutions have been steadily increasing and continued after the tariff announcement.
  - The rise occurred without notable shifts in the composition of buyers.
  - Compared to March 2020 COVID-19 turmoil, cumulative flows from non-euro countries into the euro are notably smaller during the tariff-related episode.
  - Composition of foreign buyers during the tariff-related episode has been more diversified.
  - Net flows into the yen since April 2 have significantly exceeded those observed during the COVID-19 period, with substantial inflows from the United Kingdom and major euro area economies.
- Displayed chart scales (example axis ticks as shown in figures):
  - -35,000; -25,000; -15,000; -5,000; 5,000; 15,000; 25,000; 35,000.
  - -400; -300; -200; -100; 0; 100; 200; 300; 400.
  - -150; 50; -35,000; -15,000; 5,000; 25,000.

### Settlement risk, PvP participation, and market resilience
- Two empirical approaches assess settlement risk (CLS/PvP) on excess FX returns and volatility:
  - (i) Difference-in-differences (DID) case study for Hungarian Forint (HUF) around accession to CLS on November 16, 2015.
  - (ii) Panel regression across expanded currency sample over January 1, 2000 to May 31, 2025.
- HUF DID case study:
  - Control currencies: CZK and PLN.
  - Pre-treatment correlations of exchange rate returns with HUF/USD over the one-year pre-event period: 0.8 for CZK/USD and 0.9 for PLN/USD.
  - Parallel trends test: interaction coefficient 훽3 = 8.3 × 10⁻⁸, p-value = 0.99 (not statistically significant).
  - Estimated effects of joining CLS (average daily DID impact):
    - Daily excess return declined by 11 basis points one month after entry.
    - Daily excess return declined by 10.7 basis points three months after entry.
    - Excess HUF/USD return volatility decreased by about 3.4 basis points in both one-month and three-month periods.
    - Volatility decline represents a 4 percent reduction relative to the average volatility in the one-year prior to accession.
  - Time windows: primary estimation October 16–December 15, 2015; robustness window August 16, 2015 to February 15, 2016.
- Panel analysis of PvP participation:
  - Sample period: January 1, 2000 to May 31, 2025. Currency sample: 26 currencies (20 settled through PvP, 6 not).
  - Treatment indicator: PvP dummy = 1 from date a currency joined a PvP arrangement onward.
  - Estimated average effects of PvP participation:
    - Significant reduction of 34.1 basis points on excess returns associated with PvP participation (both 2000-2025 and 2002–2015 subsamples).
    - Excess FX return volatility declines by about 3.1 basis points.
  - Regression controls: stock price returns and 3-month interest rate differential with the US; fixed effects: country-week-year, country, and time; standard errors clustered at the country level.
  - Statistical significance: effects significant at the 10 percent significance level or below.

### Operational outages and liquidity impacts
- Two outages analyzed:
  - FX Matching outage on June 30, 2015.
  - EBS outage on July 25, 2023.
- Currencies analyzed vs USD: EUR, JPY, GBP, CHF, CAD, AUD, NZD, SEK, NOK.
- Market condition metrics:
  - Bid-ask spread sampled at 30-minute intervals (standardized within 181-day windows).
  - Realized illiquidity (Ranaldo and Santucci de Magistris 2022): ratio of realized absolute intraday return variation to volume (USD billions).
  - Price dispersion: coefficient of variation of volume-weighted spot transaction prices.
  - Sample windows: day of outage plus 90 days before and after.
- Implications for non-CLS currencies:
  - BIS 2022 Triennial Survey interbank shares:
    - 61% for the CLS currencies.
    - 66% for the non-CLS currencies (BRL, CNY, INR, PLN, RUB, TRY sample).
  - Estimated differential effects:
    - For currencies with interbank share equal to non-CLS currencies, increase in bid-ask spreads due to interdealer platform disruption is four times larger in spot and three times larger in forward markets compared with CLS currencies.
    - Increase in price impact of trading volume is estimated to be four times larger for non-CLS than for CLS currencies.

*International Monetary Fund | ch2annex — 1. CIP Deviation Relative to US Dollar (PDF chapter/section).*

### 1. CIP Deviation Relative to US Dollar

### 1. CIP Deviation Relative to US Dollar

### Data, scope, and definitions
- Sample period: January 1, 2015, to May 31, 2025.
- Data sources: Bloomberg Finance L.P.; Refinitiv; CLS; BIS; Treasury International Capital (TIC); IMF staff calculations.
- CIP deviation (panel 1): calculated using 5-year government rates for 10 emerging market economies and 12 advanced economies against the US dollar. A negative widening basis signals stress in dollar funding markets.
- Bid-ask spread (panel 2): calculated as (ask rate-bid rate)/mid rate (in percent). Sample includes 19 emerging market economies and 18 advanced economies. Wider spreads suggest reduced market liquidity.
- Excess exchange rate return (panel 3): defined as log(exchange rate at time t / exchange rate at time t-1) - log(forward rate at time t-1 / exchange rate at time t-1). Tariff announcement refers to April 2, 2025, declaration of new import tariff rates by the United States.
- Acronyms: AEs = advanced economies; EMs = emerging markets; GFC = global financial crisis; OIS = Overnight index swaps; CIP = covered interest parity; NBFI = non-bank financial institution; CCFR = cross-currency funding (gap) ratio; NIP = net foreign investment position.

### Transaction volumes: measurement and key formulas
- Spot market gross inflows for currency c, country i, sector s in period t:
  - InFlows_c,i,s,t = ∑_{j≠i} ∑_{c'} max(FX_Flows_{c,c',i,j,s,t}, 0)  (equation 2.3.5)
  - OutFlows_c,i,s,t = ∑_{j≠i} ∑_{c'} min(FX_Flows_{c,c',i,j,s,t}, 0)  (equation 2.3.6)
  - FX_Flows_{c,c',i,j,s,t}: value of an FX spot transaction between sector s in country i and banking sector in country j; positive if sector s in country i receives currency c against payment in currency c’.
- Swap market gross inflows/outflows for currency c:
  - SwapInFlows_c,i,s,t = ∑_{j≠i} ∑_{c'} max(SWAP_Flows_{c,c',i,j,s,t}, 0)  (equation 2.3.7)
  - SwapOutFlows_c,i,s,t = ∑_{j≠i} ∑_{c'} min(SWAP_Flows_{c,c',i,j,s,t}, 0)  (equation 2.3.8)
  - NetSwapFlows_c,i,s,t = SwapInFlows_c,i,s,t − SwapOutFlows_c,i,s,t
- Outstanding open swap positions (accumulation of gross flows not yet matured as of t):
  - LongSwapPositions_{c,c',i,s,t} = ∑_{τ≤t; Maturity(τ)>t} SwapInFlows_{c,c',i,s,τ}  (equation 2.3.9)
  - ShortSwapPositions_{c,c',i,s,t} = ∑_{τ≤t; Maturity(τ)>t} SwapInFlows_{c,c',i,s,τ}  (equation 2.3.10)
  - Country-level aggregation: sum across sectors s (equations 2.3.11–2.3.12).
- Note: CLS membership mainly consists of large banks; every transaction flow from an institutional sector is recorded predominantly with banks.

### Stylized facts: USD spot and swap markets
- USD spot market:
  - US and non-US dealer banks are central market makers facilitating transactions among clients, including non-dealer banks.
  - Under normal conditions, US banks exhibit negative net USD flows (they act as providers of USD funding to other participants); non-US banks tend to exhibit positive net USD flows.
  - Non-banks domiciled outside the US are most often net buyers of USD; their trading patterns are highly sensitive to global macro-financial conditions.
  - US non-banks exhibit less distinctive patterns and can be net buyers or net sellers of USD at times.
- USD swap market (tenors ≥ 35 days):
  - Net demand for USD funding and hedging is driven mostly by non-banks; net trading flows for non-banks are mostly negative (they receive USD in the spot leg and sell USD in the forward leg).
  - US banks perform maturity transformation: borrowing dollars short term from other banks and lending to non-banks seeking longer-term hedges.
  - Aggregated flow data show US banks are net dollar borrowers in the FX swap market despite being the largest gross providers of USD, due to intermediation between short-term funding and longer-term hedging demand.
  - These aggregate flow patterns may differ when measured using outstanding positions (which capture full balance sheet exposures and reduce overrepresentation of positions rolled over at short maturities).

### Structural changes and vulnerabilities: constructed measures
- Dealer concentration (Herfindahl-Hirschman Index for currency pair c,c' at time t):
  - HHI_{c,c',t} = ∑_{j, s, ī} ( |FX_Flows_{c,c',ī,j,s,t}| / ∑_{j,s,ī} |FX_Flows_{c,c',ī,j,s,t}| )^2  (equation 2.3.13)
  - ī denotes flows from/to country-sectors with dealer banks identified by the Federal Bank of New York primary dealers list (Canada, France, Germany, Japan, Switzerland, the UK, and the US).
  - Higher HHI indicates greater dealer concentration and fewer dominant dealers.
- Share of NBFI flows (Funds and Other NBFIs) for currency pair c,c' at time t:
  - ShareNBFI_{c,c',t} = (∑_{j,i} (|FX_Flows_{c,c',i,j,Fund,t}| + |FX_Flows_{c,c',i,j,OtherNBFI,t}|)) / (∑_{j,s,i} |FX_Flows_{c,c',i,j,s,t}|)  (equation 2.3.14)
  - Denominator excludes bank-to-bank flows to compare end-customer flows directly.
  - Rationale: greater NBFI participation may destabilize markets because these institutions adjust cross-border positions rapidly in response to shocks.
- Currency mismatch (“hedging pressure”) for USD against currency c', sector s at time t:
  - HedgingPressure_{USD,c',s,t} = ( (∑_i ShortSwapPositions_{USD,c',i,s,t} − ∑_i LongSwapPositions_{USD,c',i,s,t}) / (1/52 ∑_{k=0}^{51} ∑_{s,i} OI_{USD,c',i,s,t−k} ) )  (equation 2.3.15)
  - Denominator: one-year moving average of outstanding interest in USD (sum of short and long FX swap outstanding positions across sectors), used to smooth large fluctuations.
  - Relevance: particularly meaningful for NBFIs (investment funds account for a large share of forward market outstanding interest; example: funds are a counterparty in 63 percent of all outstanding interest in forwards for the EURUSD rate).
- Cross-currency funding (gap) ratio (CCFR) for country i (USD FX mismatch):
  - USD_FX_Mismatch_{i,t} = (USD_Assets_{i,t} − USD_Liabilities_{i,t}) / USD_Assets_{i,t}  (equation 2.3.16)
  - USD_Assets_i,t and USD_Liabilities_i,t derived from BIS Locational Banking Statistics at quarterly frequency.
  - Used as proxy for reliance on “synthetic” funding via FX swaps.
- Net foreign investment position (NIP) for country i:
  - NIP_{i,t} = (ForeignPositions_in_US_bonds_{i,t} − USPositions_in_Foreign_bonds_{i,t}) / (ForeignPositions_in_US_bonds_{i,t} + USPositions_in_Foreign_bonds_{i,t})  (equation 2.3.17)
  - Positive NIP indicates foreign holdings of US bonds exceed US holdings of foreign bonds for country i, implying greater potential hedging demand from foreign investors into US dollars.
  - Foreign and US positions drawn from US Treasury long-term bond holdings (TIC dataset). CCFR and NIP are used at the currency level by aggregating countries within the same currency area using GDP-weighted averages.

### Empirical methodology for macrofinancial uncertainty shocks
- Uncertainty measures and shock definitions:
  - Economic Policy Uncertainty (EPU) — US index (Baker et al. 2016). Shock D_UNCRT_t: AR(1) residual for EPU exceeds two standard deviations above its mean.
  - Financial uncertainty — VIX index. Shock D_FINUNCRT_t: AR(1) residual of VIX exceeds two standard deviations above its mean.
  - Monetary policy uncertainty — MOVE index. Shock D_MOVE_t: AR(1) residual exceeds two standard deviations above its mean. Robustness: alternative monetary policy uncertainty proxy D_MPU_t derived from US EPU index.
- Econometric specification (weekly frequency, January 1, 2015–May 31, 2025):
  - Δlog(InFlow_{c,i,s,t+h}) = α_{c,i,s}^h + λ_{c,i,t}^h + β_c^h · GShock_t + γ_c^h · Control_{c,i,s,t} + δ_c^h · Global_t^h + ν_{c,i,s,t}^h  (equation 2.4.1)
  - Control variables include domestic country term spread (10-year minus 3-month government bond yield) and exchange rate for currency c; FX swap inflow models also control for 3-month CIP deviation.
  - Global_t controls for common factors: commodity price index, Chicago Financial Condition Index, US term spread (US 10-year minus 3-month), end-of-quarter dummies.
  - Fixed effects: α_{c,i,s} (country-sector) and λ_{c,i,t} (country-month-year). Standard errors clustered at the country level.
  - Models estimated for spot and swap transactions and across different sectors; analysis focuses mainly on USD with CHF and JPY also included for safe-haven comparisons.

### Empirical results and key quantitative findings
- Hedging pressure drivers:
  - Beyond term spread differentials, hedging pressure is closely linked to the USD net investment position (NIP).
  - Quantitative estimate: a one–percentage point change in NIP corresponds to a 1.2 percentage point change in hedging pressure.
- Behavioral patterns under macrofinancial uncertainty:
  - When macrofinancial uncertainty rises, nonresident NBFIs tend to increase demand for safe-haven assets, not only US dollar assets but also euro and Swiss franc; net spot purchases of EUR and CHF rise markedly, indicating active diversification across safe-haven assets.
  - Online Annex Figure 2.3.5 reports unconditional correlations between net spot flows of NBFIs to safe-haven currencies and uncertainty measures (VIX, EPU, MOVE) over 2015M1–2025M5.
- Sectoral flow dynamics:
  - Spot market 4-quarter moving-average USD net flows are shown separately for US banks, US nonbank institutions, non-US banks, and non-US nonbank institutions (Online Annex Figure 2.3.3); measures of uncertainty are standardized.
  - Long-term swap (tenor ≥ 35 days) USD net flows by sector are shown as 4-quarter moving averages and indicate non-banks drive net demand for USD hedging (Online Annex Figure 2.3.4).
- Robustness:
  - Baseline model estimates using monetary policy uncertainty proxies (D_MPU_t) mirror main-text patterns, reinforcing robustness (Online Annex Figure 2.4.1).

*Source: ch2annex - 1. CIP Deviation Relative to US Dollar (PDF chapter/section), International Monetary Fund; data and calculations described in the provided content.*

### 1. Change in USD Spot Flows following an Increase in Monetary

### 1. Change in USD Spot Flows following an Increase in Monetary Policy Uncertainty due to Monetary Policy Shocks Based on Gürkaynak, Sack, and Swanson (2005) (Percentage points)

### Methodology
- Shocks constructed by regressing the U.S. EPU index on alternative measures of monetary policy surprises following Gürkaynak, Sack, and Swanson (2005) and Nakamura and Steinsson (2018).
- Panel-model estimates of weekly changes in USD inflows in the spot market for non-US financial and nonfinancial institutions across 15 jurisdictions.
- Model controls include: Chicago Financial Conditions Index, a commodity price index, the U.S. term spread, domestic term spreads, the spot exchange rate.
- Specification includes country-sector fixed effects and country-time fixed effects.
- Uncertainty shocks defined as dummy variables equal to 1 when the AR(1) residual of the respective indicator exceeds two standard deviations.
- Data and sources: Baker, Bloom, and Davis (2016); CLS Group; LSEG Datastream; IMF staff calculations.
- Analysis period for some regressions: January 1, 2006–May 31, 2025.

### Key findings on USD and other safe-haven currency flows
- Heightened global uncertainty increases demand for safe-haven currencies (CHF, JPY, EUR) by nonresidents.
- VIX shock estimated to increase the volume of transactions in the spot market for JPY by about 40 percent.
- Comparison of market sizes and absolute effects:
  - Average weekly USD spot transactions immediately preceding the COVID-19 outbreak: roughly USD 2.2 trillion.
  - A 21 percent rise in USD spot transactions during the crisis corresponded to an additional USD 0.46 trillion in trading volume.
  - Average weekly JPY spot transactions: about USD 0.6 trillion.
  - A 42 percent increase in JPY spot transactions amounted to USD 0.25 trillion.
- Despite larger percentage changes in JPY turnover, the increase in USD inflows was more significant in absolute terms.

### Sectoral differences
- Nonbank financial institutions (NBFIs) show the strongest increases in demand for safe-haven currencies following global uncertainty shocks.
  - Following VIX shocks, volume transacted by NBFIs in the JPY spot market increases by 55 percent, twice as much as the volume transacted by banks.
  - A MOVE shock increases the volume transacted by NBFIs in the CHF spot market by 47 percent; the effect for banks is below 20 percent.
- NBFIs are also more sensitive to large domestic financial uncertainty shocks, with their response in the swap market occurring with a one-period lag relative to the spot market.

### Macro-financial disconnect (US EPU vs VIX)
- Constructed a measure of the disconnect between U.S. policy uncertainty (EPU) and global financial uncertainty (VIX) as an alternative shock variable.
- Empirical patterns:
  - When US EPU is high while VIX is contained (EPU–VIX disconnect is wider), nonresident investors may reduce USD purchases.
  - The effect is nonlinear, increasing in the magnitude of the disconnect.
  - Institutions also react to the disconnect in the swap market by decreasing hedging activity to some extent.

### Domestic uncertainty shocks
- Domestic shocks constructed based on implied volatility of 10 percent out-of-the-money put options specific to the domestic currency area (analogous to VIX shock construction).
- Large domestic financial uncertainty shocks also trigger increased purchases of U.S. dollars in both spot and swap markets.
- NBFIs again appear more sensitive to such shocks.

### Persistence and cross-country heterogeneity
- The impact of VIX shocks peaks at around four weeks and persists for up to 16 weeks; similar patterns observed for US EPU shocks.
- Emerging markets (EMs) face stronger and more persistent impacts from uncertainty shocks across FX market functioning indicators:
  - Treasury CIP deviation widens by 16 basis points on average.
  - Bid-ask spreads increase by 0.04 basis points on average.
  - Excess exchange rate return volatility rises by 0.2 percentage points on average.
- Treasury CIP deviation in EMs interpreted as a signal of dollar funding stress (tighter dollar liquidity) when OIS rates unavailable; regressions control for expected default frequency of banking sector and are robust to sovereign CDS-based controls.

### Robustness and extensions
- Robustness checks include alternative fixed-effect specifications (including country-sector–time fixed effects), controls for end-of-month effects, presence of FX swap lines, inclusion of FX implied volatility, the LIBOR–OIS spread, USD broad index, and alternative clustering of standard errors at the country-sector level.
- Results remain robust in a pre-tariff-period sample (2015:M1–2024:M12) that excludes unusually high EPU around April 2 tariff announcements.
- Equation (2.5.1) estimated across weekly horizons h = 0,...,16 weeks for 11 currency pairs over 2006-01-01 to 2025-05-31.
- Extensions examine whether structural vulnerabilities amplify shock effects via interaction terms (dealer concentration, currency mismatches, NBFI participation share) and whether policy measures (swap-line dummy, international reserves normalized by GDP) influence FX market conditions.
- Standard errors computed using Newey–West HAC estimator.

### Additional stylized facts on market-making capacity
- US Treasury supply has surged in recent years driven by large government borrowing, while dealer balance sheets have remained flat, limiting dealers’ capacity to absorb inventory and intermediate.
- International capital ratio (intermediaries’ equity divided by total assets) has fluctuated but not kept pace with growth in marketable debt, suggesting potential constraints on market-making capacity.

*Source: IMF staff calculations; Baker, Bloom, and Davis (2016); CLS Group; LSEG Datastream.*

### 1. Effect of an increase in VIX on CIP

### 1. Effect of an increase in VIX on CIP

### Effect of uncertainty shocks on FX market conditions
- The panels show the effect of one standard deviation of each uncertainty indicator and corresponding shocks across time on:
  - weekly 3-months OIS CIP deviation (Basis points),
  - excess exchange rate return volatility (Percentage points),
  - bid-ask spreads (Percent).
- High uncertainty shocks are defined as dummy variables equal to 1 when the AR(1) residual of the respective indicator exceeds two standard deviations.
- VIX = CBOE Volatility Index.
- Data sources: Baker, Bloom, and Davis (2016); CLS Group; LSEG Datastream; and IMF staff calculations.

### Effects documented (AE and EM samples)
- For Treasury measures (5-years CIP deviation based on government bond yields to extend coverage among EMs):
  - EM sample comprises 16 emerging market economies, with coverage varying by the availability of the target variable.
  - Estimates use weekly horizons and show effects of one standard deviation increases in uncertainty indicators and corresponding large uncertainty shocks (AR(1) residual exceeds two standard deviations).
  - Whiskers show the 90 percent confidence intervals.
- Notes on specification:
  - The panel shows effects over time on weekly 5-years CIP deviation, excess rate return volatility, and bid-ask spreads.
  - AE = advanced economies; EM = emerging market economies.

### International reserves as a stabilizer
- International reserves mitigate FX market stress by enabling central banks to provide dollar liquidity and by bolstering sovereign creditworthiness, limiting depreciation pressures.
- Economies with reserve buffers about one standard deviation above the average show markedly smaller CIP deviations and lower excess exchange rate return volatility after macro-financial uncertainty shocks.
- Mitigating effect panels:
  - Effect of one-standard-deviation increase in VIX and US EPU and their uncertainty shocks on FX market conditions, with interaction effects of international reserves (normalized by GDP).
  - The FX reserves interaction term captures the incremental impact of an uncertainty shock associated with a one–standard-deviation increase in FX reserves.
  - Whiskers show the 90 percent confidence intervals.
- Data sources: Bloomberg; Refinitiv; and IMF staff calculations.
- EPU = economic policy uncertainty; VIX = CBOE Volatility Index.

### Robustness checks
- Fixed-effects variations: currency, time, and currency–time fixed effects tested.
- Additional controls: end-of-month and end-of-quarter effects.
- Extended specification includes 3-months FX option-implied volatility and the LIBOR–OIS spread.
- Identification/addressing endogeneity:
  - Granular instrumental variables (GIV) methodology (Gabaix and Koijen 2023) deployed to extract idiosyncratic shocks from two granular instruments based on cross-sectional differences in capital ratios of (i) primary dealer and non-dealer banks and (ii) bank asset size-weighted and equal-weighted aggregates.
  - Results remain broadly consistent across all robustness tests.
- Pre-tariff robustness: analysis in Figure 2.9 repeated using data only up to 2024M12; results broadly consistent with baseline.
- Pre-tariff period exclusion robustness:
  - Panels show effect of one standard deviation increase in each uncertainty indicator and high uncertainty shocks on three-month OIS CIP deviation, excess exchange rate return volatility, and bid-ask spreads over a one-week horizon and excluding 2025 observations.
  - MOVE = Merrill Lynch Option Volatility Estimate; OIS = Overnight Index Swap.

### Empirical magnitudes highlighted in text
- Economies with reserve buffers about one standard deviation above the average show markedly smaller CIP deviations and lower excess exchange rate return volatility after macro-financial uncertainty shocks.

### Sources and notes
- Sources: Baker, Bloom, and Davis (2016); CLS Group; LSEG Datastream; Bloomberg; Refinitiv; and IMF staff calculations.
- Notes: Whiskers show the 90 percent confidence intervals where plotted.

### Final robustness takeaway
- Results remain broadly consistent across specifications, additional controls, and identification strategies including GIV and residualization via principal component analysis.

### Attribution
*International Monetary Fund | October 2025 — Annex: Effect of an increase in VIX on CIP*

### 4. Net yen cumulative spot flows by sector and nationality

### 4. Net yen cumulative spot flows by sector and nationality

### Panels, data coverage, and methodological notes
- Data source: CLS Group; IMF staff calculations.
- Panels 1 and 2 display cumulative flows beginning on November 1, 2024.
- Panels 3 and 4 show cumulative flows starting from October 2019.
- Arrows indicating tariff-related events correspond to April 3, 2025, for Panels 1 and 2, and March 9, 2020, for the COVID-19 market turmoil.
- Country names are represented by ISO 3-digits code except EA (DE, FR, IT)=Germany, France, and Italy; EA Other=euro area except Germany, France, and Italy; ROW=Rest of the World.
- Statistics are descriptive and do not account for other potential drivers of FX market dynamics beyond the tariff and COVID-19 shocks—such as interest rate differentials or other macroeconomic factors.

### Key findings on yen flows and composition of buyers
- For the Japanese yen:
  - Net purchases by foreign institutions have been steadily increasing, a trend that has continued following the tariff announcement.
  - This rise occurred without any notable shifts in the composition of buyers (referenced: Online Annex Figure 2.7.2, panel 2).
- Cross-episode comparisons:
  - Compared to the market turmoil during the COVID-19 episode in March 2020, the cumulative volume of flows from non-euro countries into the euro appears notably smaller during the recent tariff-related episode.
  - The composition of foreign buyers during the tariff-related episode has been more diversified.
  - Net flows into the yen since April 2 have significantly exceeded those observed during the COVID-19 period, with substantial inflows originating from the United Kingdom and major euro area economies.

### Displayed chart scales (as shown in source figures)
- Example axis tick values presented in the figure panels:
  - Panel axes include values such as -35,000; -25,000; -15,000; -5,000; 5,000; 15,000; 25,000; 35,000.
  - Other panel axes include values such as -400; -300; -200; -100; 0; 100; 200; 300; 400.
  - Additional axis ticks shown: -150; 50; -35,000; -15,000; 5,000; 25,000.
- These numeric scales are part of the descriptive charts of cumulative spot flows by sector and nationality.

### Related empirical evidence on settlement risk and market resilience (Online Annex 2.8 summary)
- Two empirical approaches are used to assess the effect of settlement risk (CLS/PvP) on excess FX returns and volatility:
  - (i) Difference-in-differences (DID) case study for the Hungarian Forint (HUF) around Hungary’s accession to CLS on November 16, 2015.
  - (ii) Panel regression across an expanded currency sample over January 1, 2000 to May 31, 2025.
- Case study: Hungarian Forint (DID)
  - Control currencies: Czech koruna (CZK) and Polish zloty (PLN).
  - Pre-treatment correlations of exchange rate returns with HUF/USD over the one-year pre-event period: 0.8 for CZK/USD and 0.9 for PLN/USD.
  - Parallel trends test: interaction coefficient 훽3 = 8.3 × 10⁻⁸, p-value = 0.99 (not statistically significant), supporting parallel trends.
  - Estimated effects of joining CLS (average daily DID impact):
    - Daily excess return declined by 11 basis points one month after the entry date.
    - Daily excess return declined by 10.7 basis points three months after the entry date.
    - Excess HUF/USD return volatility decreased by about 3.4 basis points in both the one-month and the three-month periods following CLS participation.
    - The volatility decline represents a 4 percent reduction relative to the average volatility observed during the one-year period prior to Hungary’s accession to CLS.
  - Time windows: primary estimation over October 16–December 15, 2015; robustness window August 16, 2015 to February 15, 2016.
- Panel analysis (PvP participation across currencies)
  - Sample period: January 1, 2000 to May 31, 2025.
  - Currency sample: 20 currencies settled through PvP arrangements (CLS, CHATS, CCIL, B3) and 6 currencies not settled through PvP arrangements, totaling 26 currencies.
  - Treatment indicator: PvP dummy equals 1 from the date a currency joined a PvP arrangement onward (examples and join dates are enumerated in source).
  - Estimated average effects of PvP participation:
    - Significant reduction of 34.1 basis points on excess returns associated with PvP system participation (both over 2000-2025 and 2002–2015 subsample).
    - Excess FX return volatility declines by about 3.1 basis points.
  - Regression controls include stock price returns and 3-month interest rate differential with the US; regressions include country-week-year, country, and time fixed effects; standard errors clustered at the country level.
- Statistical significance: reported effects are statistically significant at the 10 percent significance level or below in the presented analyses.

### Operational disruptions in market venues and liquidity impacts (Online Annex 2.9 summary)
- Two primary outages analyzed:
  - FX Matching outage on June 30, 2015.
  - EBS outage on July 25, 2023.
- Currencies analyzed against the US dollar: EUR, JPY, GBP, CHF, CAD, AUD, NZD, SEK, NOK.
- Market condition metrics and sampling:
  - Bid-ask spread in the spot market sampled at 30-minute intervals (standardized within 181-day windows).
  - Realized illiquidity measure (Ranaldo and Santucci de Magistris 2022) computed from 30-minute returns and CLS spot market volume.
  - Price dispersion measured by the coefficient of variation of volume-weighted spot transaction prices.
  - Sample windows: day of each outage plus the 90 days preceding and following it.
- Key methodological specifications: several standardized regression specifications estimate the outage effect on market conditions, controlling for date, currency-time-year, and currency-year fixed effects as appropriate (equations (2.9.1)–(2.9.9), (2.9.10) in source).
- Implications for non-CLS currencies
  - BIS 2022 Triennial Central Bank Survey: interbank transactions as percentage of total spot volume:
    - 61% for the CLS currencies.
    - 66% for the non-CLS currencies (sample of non-CLS currencies with available interbank share: BRL, CNY, INR, PLN, RUB, and TRY).
  - Estimated differential effects of interdealer platform disruptions:
    - For currencies with an interbank share equal to that for the non-CLS currencies, the increase in bid-ask spreads in the spot (forward) market due to an interdealer platform disruption is four (three) times larger than for the CLS currencies.
    - The increase in price impact of trading volume is estimated to be four times larger for the non-CLS than for the CLS currencies.

*International Monetary Fund | October 2025 — ch2annex (figure and accompanying annex text)*

### References

### ch2annex - References

### Referenced studies and topics
- Barajas, Adolfo, Andrea Deghi, Claudio Raddatz, Dulani Seneviratne, Peichu Xie, and Yizhi Xu. 2020. “Global Banks’ Dollar Funding: A Source of Financial Vulnerability.” IMF Working Paper WP/20/113. Washington DC: International Monetary Fund.
- Baker, Scott, Nicholas Bloom, Steven Davis. 2016. “Measuring Economic Policy Uncertainty,” The Quarterly Journal of Economics, 131, 4, pp. 1593–1636.
- Bank for International Settlements. 2022. BIS Quarterly Review. December. https://www.bis.org/publ/qtrpdf/r_qt2212.htm
- Caldara, Dario, and Matteo Iacoviello. 2022. “Measuring Geopolitical Risk.” American Economic Review 112 (4): 1194–225
- Copeland, Adam, Darrell Duffie, Yilin Yang. 2021. “What Quantity of Reserves Is Sufficient.” Liberty Street Economics, Federal Reserve Bank of New York.
- Dao, Mai, and Pierre-Olivier Gourinchas. 2025. “Covered Interest Parity Deviations in Emerging Markets: Measurement and Drivers.” IMF Working Paper 25/057, International Monetary Fund, Washington, DC.
- Diamond, William, Zhengyang Jiang, and Yiming Ma. 2024. “The Reserve Supply Channel of Unconventional Monetary Policy.” Journal of Financial Economics, 159.
- Du Wenxin, and Amy Huber. 2024. “Dollar Asset Holdings and Hedging Around the Globe.” NBER Working Paper 32453.
- Du, Wenxin, Joanne Im, and Jesse Schreger. 2018. “The U.S. Treasury Premium.” Journal of International Economics, 112: 167–181
- Du, Wenxin, and Jesse Schreger. 2016. “Local Currency Sovereign Risk.” The Journal of Finance 71, no. 3: 1027-1070.
- Du, Wenxin, and Jesse Schreger. 2022. “Sovereign Risk, Currency Risk, and Corporate Balance Sheets.” The Review of Financial Studies 35, no. 10: 4587-4629.
- Du Wenxin, Alexander Tepper, and Adrien Verdelhan. 2018. “Deviations from Covered Interest Rate Parity.” The Journal of Finance, 73(3): 915-957.
- Eguren‐Martin, O. Fernando, B. Matias, D. Reinhardt. 2023. “Global Banks and Synthetic Funding: The Benefits of Foreign Relatives”, Journal of Money, Credit and Banking, 56(1): 115-152.
- Gabaix, X., and R. Koijen. “Granular Instrumental Variables.” Journal of Political Economy 132(7): 2274-2303.
- Gürkaynak, Refet S., Brian Sack, and Eric Swanson. 2005. “The Sensitivity of Long-Term Interest Rates to Economic News: Evidence and Implications for Macroeconomic Models." American economic review 95, no. 1: 425-436.
- Hasbrouck, Joel, and Richard M. Levich. 2021. “Network Structure and Pricing in the FX Market.” Journal of Financial Economics 141, no. 2 705-729.
- He, Zhiguo, Bryan Kelly, and Asaf Manela. 2017. “Intermediary Asset Pricing: New Evidence from Many Asset Classes.” Journal of Financial Economics, 126(1): 1–35.
- Kang, Wenjin, K Geert Rouwenhorst, and Ke Tang. 2020. “A Tale of Two Premiums: The Role of Hedgers and Speculators in Commodity Futures Markets.” The Journal of Finance, 75:377–417.
- Kloks, Pēteris, Patrick McGuire, Angelo Ranaldo, and Vladyslav Sushko. 2023. “Bank Positions in FX Swaps: Insights from CLS.” BIS Quarterly Review: 17-31.
- Liao, Gordon Y., and Tony Zhang. 2025. “The Hedging Channel of Exchange Rate Determination.” The Review of Financial Studies 38, no. 1: 1-38.
- Nakamura, Emi, and Jón Steinsson. 2018. “High-Frequency Identification of Monetary Non-Neutrality: The Information Effect.” The Quarterly Journal of Economics 133, no. 3: 1283-1330.
- Ranaldo, Angelo, and Paolo Santucci de Magistris. 2022. “Liquidity in the Global Currency Market.” Journal of Financial Economics 146, no. 3: 859-883.
- Ranaldo, Angelo, and Fabricius Somogyi. 2021. “Asymmetric Information Risk in FX Markets.” Journal of Financial Economics 140, no. 2: 391-411.

### Online Annex Figure 2.9.1 — Effect of Interdealer Platform Disruptions on Market Liquidity
- Main finding described: Bid-ask spreads increase, and market liquidity deteriorates during the outages more strongly for currencies with higher interbank share of transactions.
- Title used in figure: Effect of FX Interdealer Platform Disruption for Currencies Primarily Traded on the Platform with Higher Share of Interbank Transactions (Standard deviations)
- Data sources: Bloomberg; CLS Group; and IMF staff calculations.
- Specification and sample details:
  - The bars represent the effect of the platform outage on the outcome variable, estimated using the specification in (2.9.10), evaluated at 푠
푐
−푠̅ equal to 4.3 percent, the difference in the interbank shares of transaction volumes between CLS and non-CLS currencies, according to the BIS 2022 Triennial Central Bank Survey.
  - Realized illiquidity is defined as in Ranaldo and Santucci de Magistris (2022) and refers to the ratio between the realized absolute variation of intraday returns and the volume of transactions in billions of USD, and measures the price impact of trading volume.
  - The bid-ask spreads are sampled at 30-minute intervals while realized illiquidity is constructed at a daily frequency.
  - The sample period comprises the day of each outage, and 90 days before and after it.
  - All the measures are standardized separately in each of the two 181-day windows and for each currency.
  - The currencies in the sample include the euro, the Japanese yen, the British pound, the Swiss franc, the Canadian dollar, the Australian dollar, and the New Zealand dollar against the USD.
  - The specifications include time and currency-year effects. The specifications for the bid-ask spreads additionally include currency-time of the day-year effects.
  - The error bars represent 90 percent confidence interval, obtained using Driscoll-Kraay standard errors, with the number of lags equal to √ 푇 4, where T denotes the number of time periods in the sample.

*International Monetary Fund | October 2025 — Chapter 2 Annex References*

---


_Source: https://www.imf.org/-/media/files/publications/gfsr/2025/october/english/ch2annex.pdf_