## GLOBAL FINANCIAL STABILITY REPORT—Banks’ Dollar Funding: A Source of Financial Vulnerability

## Source details

**Canonical URL:** [GLOBAL FINANCIAL STABILITY REPORT—Banks’ Dollar Funding: A Source of Financial Vulnerability](https://www.imf.org/-/media/files/publications/gfsr/2019/october/english/onlineannex5.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/gfsr/2019/october/english/onlineannex5.pdf.md)
- [Structured JSON version](/-/media/files/publications/gfsr/2019/october/english/onlineannex5.pdf.json)

---

### Rising US dollar intermediation and drivers
- Empirical finding:
  - Non-US banks’ US dollar intermediation has risen.
- Demand drivers:
  - Corporate borrowers prefer dollar borrowing because the bulk of international trade, and therefore trade finance, is denominated in dollars.
- Supply drivers:
  - Non-US international banks collectively dominate bank intermediation of US dollar credit outside the United States because, for many, international dollar lending is more profitable than their low-margin domestic businesses.
- Funding stress link:
  - International banks’ US dollar funding stress has been related to broader financial system stress.

### Dollar funding “plumbing” and channels
- Stable dollar-denominated savings:
  - Reside mainly in the United States, and mainly in households.
  - These savings are in principle adequate to meet international demand for dollars but are effectively confined to the United States by regulation (Regulation Q prevents banks from deploying insured deposits outside the United States; almost every jurisdiction similarly ‘ring-fences’ insured deposits).
- Two broad approaches for non-US banks to obtain dollars:
  1. Tap international market instruments (interbank markets, money markets, debt capital markets, derivatives).
  2. Fund operations locally in the economy where the loan is booked (often difficult due to insufficient foreign currency savings and limited local funding relationships).
- Characteristics of international-market funding:
  - Either expensive (e.g., long-tenor bonds) or short-tenor and requires frequent refinancing.
  - Refinancing (rollover) risk is high; these instruments are chronically susceptible to fluctuations in international investors’ preferences and risk appetite.
- Cross-currency swaps and marginal cost proxy:
  - As wholesale on-balance-sheet foreign currency funding is exhausted, banks increasingly use cross-currency swaps.
  - Cross-currency swaps are mainly short-tenor and often priced at a premium to interbank cash of equivalent tenor.
  - The premium is captured by the “cross-currency basis” (CCB), used as a proxy for marginal US dollar funding costs throughout the chapter.

### Institution structure: branches versus subsidiaries and funding mixes
- Legal forms and funding implications:
  - Branches (legally subsumed within an overseas parent):
    - More readily transact with overseas affiliates; better for wholesale and corporate banking and foreign currency intermediation.
    - Generally prevented from tapping insured retail deposits; heavily dependent on uninsured (wholesale or corporate) deposits; short-term wholesale funding sourced direct from market instruments; funding raised by the overseas banking group and transferred intra-group into the branch.
    - The latter two sources together represent a high proportion of foreign branches’ total funding in most economies.
  - Subsidiaries (local legal entities able to tap insured deposits but restricted from transferring funding across borders):
    - Fund themselves more like local banks.
- Implication:
  - The bulk of foreign currency intermediation by international banks rests on stress-prone funding mechanisms when booked via branches.

### Intra-group channels, “subsidiarization,” and liquidity effects
- Observed trends:
  - Bank managers report increasing concern that branch operations’ access to intra-group liquidity management may become more restricted (“subsidiarization”).
  - Drivers include regulator actions (liquidity regulations, insolvency regimes, other mechanisms) intended to protect viability of operations in each jurisdiction.
- Evidence and interpretations:
  - Sparse evidence weakly supports concerns: branch operations are generally shifting toward higher net borrowing (lower net ‘due from’ or higher net ‘due to’) from related parties overseas.
  - Possible interpretations:
    - Reduction in branches’ access to market liquidity, forcing more intra-group borrowing.
    - Host supervisors reducing branches’ exposure to parent-group cross-border default risk.
- Gross intra-group connectivity:
  - Data do not unambiguously show shrinking gross cross-border intra-group funding; some economies (Japan, Taiwan Province of China) show increased branch connectedness with overseas related parties.
  - Any ‘subsidiarization’ is nascent in most economies and may be difficult to detect.
- Consequence for dollar liquidity demand:
  - Restriction on intra-group channels—greater net borrowing by branches or shrinking cross-border funding—implies branches must hold more precautionary US dollar liquidity relative to funding commitments.
  - This increases banks’ effective demand for US dollar funding and further pressures the dollar supply-demand balance.
  - Non-US banks are effectively accumulating greater persistent US dollar liquidity on the balance sheets of individual foreign bank offices.

### Key empirical and graphical elements referenced
- Online Annex Figure 5.1.1: Illustrates international non-US banks’ dollar funding channels (black arrows = open market sources; blue arrows = intra-group flows; thick solid arrows = major channels; dotted arrows = smaller/less significant).
- Online Annex Figure 5.1.2: Liabilities structure—Average by Booking Economy (Locational Basis), by Category of Institution (Percent of total liabilities, end-2018). Panels: Foreign Branches; Foreign Subsidiaries; Domestic Banks.
- Online Annex Figure 5.1.3: Change in the Characteristics of Intragroup Relationships–Branches in Selected Host Economies (Locational Basis), 2017 vs. 2010.

### Econometric methodology (ONLINE ANNEX 5.2)
- Cross-currency basis (CCB):
  - Measure of violation of covered interest parity (CIP): price difference between funding directly in US dollars and funding synthetically in US dollars using currency forward contracts.
  - The 3-month Libor basis tenor is used throughout the chapter.
  - Spot, forward and LIBOR rates obtained from Bloomberg.
  - Sample of currencies in panel 1 of Figure 5.3: Australian dollar, Brazilian real, Canadian dollar, Swiss franc, Danish krone, euro, British pound, Hong Kong dollar, Japanese yen, South Korean won, Indian rupee, Mexican peso, Malaysian ringgit, Norwegian krone, New Zealand dollar, Philippine peso, Russian ruble, Swedish krona, Thai baht, Turkish lira, South African rand.
- Drivers and amplification (CCFR):
  - Purpose: understand drivers of the CCB and whether cross-currency funding ratio (CCFR) amplifies effects.
  - Sample period: 01/01/2000 to 03/01/2018.
  - Currencies for this exercise: Australian dollar, Canadian dollar, Swiss franc, euro, British pound, Hong Kong dollar, Japanese yen, Indian rupee, Malaysian ringgit, Swedish krona.
  - Baseline regression: CCB_i,t = alpha_i + beta' X_i,t + epsilon_i,t (alpha_i = currency fixed effects).
  - Drivers included: VIX; US dollar index; Home economy default probability (EDF over 1-year horizon); FX implied volatility; Bid-ask spread; Term-spread differential; Libor-OIS spread; CCFR.
  - Regressors standardized prior to estimation.
  - “High” and “low” CCFR defined as fourth and first quintiles of CCFR historical distribution.
  - Absence of panel unit roots tested with Im–Pesaran–Shin test.
  - Robust standard errors; results robust with Newey-West standard errors.
- Effects of regulation and reserve holding:
  - Variant regression includes month fixed effects (alpha_m) to control quarter- and year-end effects.
  - Sample period: 01/01/2012 to 12/31/2016.
  - Focus currencies: Australian dollar, Canadian dollar, Swiss franc, euro, British pound, Japanese yen, Swedish krona.
  - Regulatory reform dummies:
    - Stress-VaR 01/01/2013 – 12/31/2013
    - Supplementary leverage ratio 01/01/2014 – 12/31/2014
    - Liquidity coverage ratio 01/01/2015 – 12/31/2015
    - Money market mutual fund reform (MMMF) 01/01/2016 – 12/31/2016
  - Values shown in panel 2 of Figure 5.4 refer to the effect of baseline coefficient and cross-term with each regulation dummy.

*Source: Chapter 5 and Online Annex 5.2, October 2019 Global Financial Stability Report (GFSR).*

### 2. USD FUNDING COSTS AND HOME COUNTRY FINANCIAL STRESS — Research questions and baseline
- Research aims:
  - How banking sector stability and domestic financial conditions in home economies of global non-US banks respond to changes in US dollar funding costs.
  - Dependence on non-US banks’ share of US dollar business and USD funding fragility.
  - Policy-related mitigators.
- Baseline regression specification:
  - Dependent variable: change in country-level financial stress ∆1 Y_i,t, captured by either average 1-year ahead bank probability of default PPD_i,t or financial conditions index FCI_i,t.
  - Key regressor: change in US dollar funding cost CCBS_C_i,t (the negative of the CCB_i,t).
  - Controls (CnControls_i,t): interest rate differentials, inflation, log NEER, log VIX, volatility of log VIX, real GDP growth; bank controls: capital to assets ratio, cash to assets ratio, return to assets ratio, deposit to assets ratio, net loan to assets ratio, cost to income ratio.
  - All controls are moving averages from quarter t-4 to t-1.
  - Country fixed effects α_i; four period dummies P1 to P4 with interactions:
    - P1 (2007≤Year_C ≤2009)
    - P2 (2011≤Year_C ≤2012)
    - P3 (Year_C ≥2013)
    - P4 (Year_C = ResC)
  - Standard errors clustered at the economy level.
- Representative exercise:
  - Association between a contemporaneous increase in the change of the US dollar funding cost by 50 basis points and ∆PD or ∆FCI.

### Baseline cross-border variant (recipient economies)
- Traces spillovers to recipient economies that receive cross-border dollar credit from global non-US banks.
- Example: association between a 50 basis point increase in US dollar funding costs across a recipient economy's top 10 lending economies and the probability of default of 38 recipient economies; comparison of top 10 main borrowers versus the rest.

### Amplification effects: share of USD assets and funding fragility
- Augmented specification adds share of US dollar assets to total assets S_i,t−1~−4 and its interaction with ∆1 CCBS_C_i,t.
  - Share S is average for economy i between quarter t-4 and quarter t-1.
- Funding fragility specification replaces share with fragility F_i,t−1~−4 (average t-4 to t-1) measured by one of: CCFR, US dollar liquidity ratio (LR), US dollar stable funding ratio (SFR).
- Asymmetry:
  - Differentiate responses when funding gap (CCFR/CCFG) is positive or negative using indicator [CCFR>0] and interactions.
- Empirical comparisons:
  - Panel 2 of Figure 5.7: effect when CCFR or ratio of cross-currency funding gap (CCFG) to US dollar assets in quarter t-4 to t-1 is low versus high.
  - Panel 3 of Figure 5.7: effect when LR or SFR in quarter t-4 to t-1 is low versus high relative to historical distribution.

### Policy-related variables as mitigators
- Bank health (BH) specification:
  - Bank health variable CCBH_i,t−1~−4 averaged over t-4 to t-1 and its interaction with ∆1 CCBS_C_i,t.
  - Metrics: capital-asset ratio (capital), cash-assets ratio (liquidity), return on assets (ROA).
  - Panel 1 of Figure 5.9: association between a 50 basis point increase in ∆1 CCBS_C_i,t and ∆PD comparing low vs high bank health.
- Swap line (Fed) specification:
  - Dummy SSWAP_i,t indicates swap liquidity agreement between central bank in economy i and the Federal Reserve in quarter t.
  - Model includes separate coefficients on ∆1 CCBS_C_i,t when SSWAP_i,t = 0 and when SSWAP_i,t = 1.
  - Panel 2 of Figure 5.10: compares association for a 50 basis point increase when a swap line arrangement is in effect versus when it is not.
- International reserves (R) specification:
  - Introduce R_i,t−1~−4 (international reserve holdings averaged over t-4 to t-1) and interactions with ∆1 CCBS_C_i,t and fragility F_i,t−1~−4.
  - Panel 3 of Figure 5.10: transmission effect of US dollar funding fragility—with LR or SFR evaluated at their median—on ∆PD when home central bank’s international reserve holding is low or high by historical standards.
- All regressions use standard errors clustered at the economy level.

### Robustness and identification strategies (summary)
- Unrestricted panel VAR with Impulse Response Functions to corroborate basis response to CCFR shocks.
- Instrumental-variable regressions using US monetary policy shocks as instrument for basis; instrumented basis positively and significantly related to home-economy probability of default.
- Euro area-only regressions: effect persists when restricting sample by excluding Germany and France.
- Time fixed effects and additional macro controls confirm core message: conditional on US dollar funding fragility, tightening in US dollar funding conditions transmits to stress in home economies of non-US global banks.
- Control for non-banking financial stress (sovereign CDS) leaves results broadly robust.
- Core identification: difference-in-difference strategy to measure amplification/mitigation effects.

### Data and sources (selected, exact labels)
- BID-Ask Spread — Bloomberg. IMF staff estimates.
- Cross-Currency Basis — Bloomberg, IMF staff estimates.
- Default Probability (EDF 1-year) — Bloomberg. IMF staff estimates.
- Financial Conditions Index — methodology in Annex 3.2 of October 2017 GFSR — IMF staff estimates.
- LIBOR-OIS spread — Bloomberg, IMF staff estimates.
- Term spread differential — Haver Analytics.
- US dollar index — Trade Weighted U.S. Dollar Index: Broad — Haver Analytics.
- VIX — Chicago Board Options Exchange Volatility Index — Haver Analytics.
- US dollar bilateral lending (cross-border) — BIS.
- Swap line arrangement — Board of Governors of the Federal Reserve System.
- International Reserves — International Financial Statistics.
- Probability of Default of Banking Sector (EDF) — The National University of Singapore Risk Management Institute (RMI).
- Banking sector ratios (capital, liquidity, ROA, deposit ratio, cost ratio, net loan ratio) — Fitchconnect, Bankscope, IMF staff estimates.
- EM dummy — Brauning and Ivashina (2019).

### Online Annex 5.3 — Non-US Banks’ US Dollar Balance Sheet and Funding Fragility Measures
- Concepts:
  - International Position (IP): USD operations outside the United States, capturing USD cross-border positions of non-US banks and US dollar local positions of non-US banks outside the United States.
  - Branch operations in the United States.
  - IP + Branches in the United States (IP+B): aggregate of the two above.
  - Foreign Position (FP): balance sheets constructed by adding subsidiary operations in the United States.
- Sample coverage (Panel A sample with all three funding fragility measures includes): Australia, Austria, Canada, France, Germany, India, Italy, Brazil, China, Cyprus, Hong Kong SAR, Luxembourg, Malaysia, Japan, Korea, Netherlands, Spain, Sweden, Switzerland, United Kingdom, Mexico, Norway, Singapore, South Africa, Turkey, Russia.
- Data sources and construction:
  - BIS locational statistics on nationality basis (confidential database) used to capture USD operations outside the United States.
  - US branch operations constructed via bottom-up aggregation based on FFIEC 002 regulatory filings (nationality basis).
  - US subsidiary operations constructed using FFIEC 031/041 call report filings for non-US-owned subsidiaries via S&P Global Market Intelligence.
- Empirical observation:
  - Inclusion of US subsidiaries of non-US banks does not affect analysis materially (comparison of liquidity and stable funding ratios across IP+B and FP concepts).

### Construction of the US dollar liquidity and stable funding ratios (LR and SFR)
- Methodology:
  - LR and SFR follow methodology developed in Chapter 1 of the April 2018 GFSR (Online Annex 1.2 of that chapter), with minor modifications and additional LR estimates using alternative outflow assumptions.
- Deposit component adjustments:
  - Refinement: only a fraction of US dollar-denominated bonds and short-term debt securities issued by banks is subtracted from liabilities to nonbanks; fraction corresponds to estimate of portion held by nonbank sector.
  - When residency-basis counterparty breakdown exists as a shorter series, nationality-basis liabilities to nonbanks as a ratio of liabilities to all sectors used to splice series.
- Sensitivity:
  - Online Annex Figure 5.3.3: median liquidity ratios estimated using less stringent alternative assumptions varies by about fifteen percentage points from the baseline (international position plus US branches concept).
  - Scenario assumptions for run-off factors for transaction and non-transactional deposits at branch-level:
    - Scenario 1: 3 and 5
    - Scenario 2: 5 and 10
    - Scenario 3: 5 and 25
- Inclusion of foreign exchange swaps in SFR:
  - In addition to FX swap payables (included in numerator when US dollar claims > liabilities), estimate of FX swap receivables included in denominator (net position when USD liabilities are greater than claims).
- Formulae (as presented in source):
  - Liquidity ratio (LR) estimated as:
    LR = [Bihh QAYC iC y LiL Aidd AssseCCs NeEC pYsh CAACfCCCSSs] * 100
    (definition used in construction provided verbatim in source)
  - Stable funding ratio (SFR) estimated as:
    SFR = [PeS CCs iC s + CCn H CeCA se pACiCie s + FX sSYSs SYyY pc e] / [LCYns + FX sSYS s s Cep eivYpC e] * 100
    (formula components and notes preserved as in source)
- Notes:
  - Cash outflows refer to interbank liabilities, other short-term liabilities, and deposits; cash inflows refer to interbank assets; both incorporate applicable run-off factors.
  - SFR incorporates applicable run-off factors.

### Balance sheet construction (assets and liabilities mapping, exact labels)
- Assets:
  - A1. HQLA: Claims on US official sector, ultimate risk basis (BIS consolidated banking statistics, Table B4 for USA) minus HQLA position from US subsidiaries; supplemented with Call report filings via S&P Global Market Intelligence.
  - A2. Interbank assets: Claims on banks (BIS locational banking statistics, Nationality basis) minus intragroup claims plus US branches' interbank assets (FFIEC 002).
  - A3. Intragroup assets: Intragroup claims (BIS locational banking statistics, Nationality basis) plus US branches' intragroup claims (FFIEC 002).
  - A4. Loans: Claims on nonbanks (BIS locational banking statistics, Nationality basis) plus US branches' loans (FFIEC 002) minus A1. HQLA and minus claims on central bank.
  - A5. Other claims: Claims on banks and nonbanks (BIS locational banking statistics, Nationality basis) minus A1–A4 and plus US branches' other claims (FFIEC 002).
- Liabilities:
  - L1. Interbank liabilities: Liabilities to banks (BIS locational banking statistics, Nationality basis) minus intragroup liabilities plus US branches' interbank liabilities (FFIEC 002) minus long-term and short-term debt securities (held by banks) per BIS debt securities statistics and locational statistics (Nationality and Residency basis).
  - L2. Intragroup liabilities: Intragroup liabilities (BIS nationality basis) plus US branches' intragroup liabilities (FFIEC 002) minus intragroup portions of long- and short-term debt securities (BIS).
  - L3. Bonds: Long-term debt securities issued by banks (BIS debt securities statistics, Nationality basis) plus US branches' long-term securities (FFIEC 002).
  - L4. Other short-term liabilities: Short-term debt securities issued by banks (BIS) plus US branches' short-term securities (FFIEC 002).
  - L5. Deposits: Liabilities to nonbanks (BIS nationality basis) minus long-term and short-term debt securities (held by nonbanks) per BIS debt securities statistics (Nationality and Residency basis).
  - L6. FX swaps: Claims on banks and claims on nonbanks (BIS nationality basis) minus liabilities on banks and liabilities on nonbanks (BIS nationality basis) plus US branches' FX swaps (FFIEC 002).

### Country-level results and heatmaps
- Online Annex Tables 5.3.4 and 5.3.5 present evolution of estimated LR and SFR within each economy, shown as deciles 1–10 across time (examples span from 2005Q1 through 2018Q1).

### FSAP findings on US dollar funding fragility (Online Box 5.1) — conclusions and recommended actions
- FSAP conclusions: some major banks vulnerable due to:
  - (i) mismatches between repo and reverse repo outflows/inflows;
  - (ii) concentrated US dollar wholesale funding (limited counterparties and small share of retail funding);
  - (iii) limited transferability of excess liquidity and collateral between jurisdictions during a crisis.
- Recommended additional actions:
  - (i) collect granular cash flow liquidity data by currency, entity and jurisdiction;
  - (ii) monitor US dollar funding patterns and liquidity metrics for various maturity buckets;
  - (iii) simulate the impact of specific liquidity scenarios for additional time horizons;
  - (iv) ensure that banks engaged in US dollar liquidity transformation have adequate US dollar liquidity buffers to cover wholesale cash outflows for at least a one-week horizon;
  - (v) assess risks related to US dollar liquidity trapping in foreign jurisdictions.

### Marginal and Total US Dollar Funding Cost (MMF evidence)
- Data and sample:
  - End-of-month portfolios of US money market mutual funds from CRANE data, Jan 2012 to Dec 2018.
  - Sample limited to 73 non-US banks who finance roughly one trillion of their US dollar activities in US dollar money markets.
  - Funding instruments: Certificates of Deposits, Commercial Paper, Repurchase Agreements.
  - Fund coverage: 461 funds managed by 81 fund families.
- Baseline econometric specification (equation labels preserved from source):
  - Dependent variable Y = yield to maturity of contract as reported to the SEC.
  - Controls include CCON_i t, CCON_{i t−1}, RemainderMaturity, log(v), X_{i,t}, IsRepo dummy, fixed effects (fund × month; bank; Repo × time; time).
  - Standard errors clustered on currency and time.
- Main empirical finding:
  - A ten-bp widening of the CCB is associated with an increase in wholesale funding costs by one basis point in the following month in the baseline specification.
  - Footnote: γ2 is negative and statistically significant and γ1 is not significant. Since the basis is negative for most currencies a decrease in the CCB is to be interpreted as a widening of the basis. Baseline test limits sample to currencies with a negative basis throughout the sample period (effectively excluding AUD, CNY, SGD, KRW).
- Robustness:
  - Results robust to including bank × fund fixed effects, banks 1-year expected default frequency, and country controls such as baseline interest rate and CCFR.

### Robustness Analysis (additional exercises)
- Exercises include:
  1. Unrestricted panel VAR with IRFs confirming basis responds to CCFR shocks.
  2. Instrumental-variable regressions using US monetary policy shocks; instrumented basis positively and significantly related to home-economy PD.
  3. Euro area-only regressions: effect persists excluding Germany and France.
  4. Time fixed effects and additional macro controls: key message holds.
  5. Controls for non-banking financial stress (sovereign CDS): results broadly robust.
- Core econometric identification: difference-in-difference strategy to measure amplification/mitigation effects.

*Source: Chapter 5 and Online Annex 5.2, October 2019 Global Financial Stability Report (GFSR).*

### Chapter 5 of the October 2019 Global Financial Stability Report (GFSR) documents rising US

### GLOBAL FINANCIAL STABILITY REPORT—Banks’ Dollar Funding: A Source of Financial Vulnerability

### Rising US dollar intermediation and drivers
- Non-US banks’ US dollar intermediation has risen (see Figure 5.1 in the chapter).
- Demand drivers:
  - Corporate borrowers prefer dollar borrowing because the bulk of international trade, and therefore trade finance, is denominated in dollars.
- Supply drivers:
  - Non-US international banks collectively dominate bank intermediation of US dollar credit outside the United States because, for many, international dollar lending is more profitable than their low-margin domestic businesses.
- Funding stress link:
  - International banks’ US dollar funding stress has been related to broader financial system stress.

### Dollar funding “plumbing” and channels
- Stable dollar-denominated savings:
  - Reside mainly in the United States, and mainly in households.
  - These savings are in principle adequate to meet international demand for dollars but are effectively confined to the United States by regulation (Regulation Q prevents banks from deploying insured deposits outside the United States; almost every jurisdiction similarly ‘ring-fences’ insured deposits).
- Two broad approaches for non-US banks to obtain dollars:
  1. Tap international market instruments (interbank markets, money markets, debt capital markets, derivatives).
  2. Fund operations locally in the economy where the loan is booked (often difficult due to insufficient foreign currency savings and limited local funding relationships).
- Characteristics of international-market funding:
  - Either expensive (e.g., long-tenor bonds) or short-tenor and requires frequent refinancing.
  - Refinancing (rollover) risk is high; these instruments are chronically susceptible to fluctuations in international investors’ preferences and risk appetite.
- As wholesale on-balance-sheet foreign currency funding is exhausted, banks increasingly use cross-currency swaps:
  - Cross-currency swaps are mainly short-tenor and often priced at a premium to interbank cash of equivalent tenor.
  - The premium is captured by the “cross-currency basis” (CCB), used as a proxy for marginal US dollar funding costs throughout the chapter.

### Institution structure: branches versus subsidiaries and funding mixes
- Two legal forms for international banks’ operations:
  - Branches (legally subsumed within an overseas parent):
    - More readily transact with overseas affiliates; better for wholesale and corporate banking and foreign currency intermediation.
    - Generally prevented from tapping insured retail deposits; heavily dependent on:
      - Uninsured (wholesale or corporate) deposits.
      - Short-term wholesale funding sourced direct from market instruments.
      - Funding raised by the overseas banking group and transferred intra-group into the branch.
    - The latter two sources together represent a high proportion of foreign branches’ total funding in most economies.
  - Subsidiaries (local legal entities able to tap insured deposits but restricted from transferring funding across borders):
    - Fund themselves more like local banks.
- Implication:
  - The bulk of foreign currency intermediation by international banks rests on stress-prone funding mechanisms when booked via branches.

### Intra-group channels, “subsidiarization,” and liquidity effects
- Observed trends and concerns:
  - Bank managers report increasing concern that branch operations’ access to intra-group liquidity management may become more restricted (“subsidiarization”).
  - Drivers include regulator actions (liquidity regulations, insolvency regimes, other mechanisms) intended to protect viability of operations in each jurisdiction.
  - Evidence (sparse) weakly supports concerns: branch operations are generally shifting toward higher net borrowing (lower net ‘due from’ or higher net ‘due to’) from related parties overseas.
- Possible interpretations of branch net borrowing shift:
  - Reduction in branches’ access to market liquidity, forcing more intra-group borrowing.
  - Host supervisors reducing branches’ exposure to parent-group cross-border default risk.
- Gross intra-group connectivity:
  - Data do not unambiguously show shrinking gross cross-border intra-group funding; some economies (Japan, Taiwan Province of China) show increased branch connectedness with overseas related parties.
  - Any ‘subsidiarization’ is nascent in most economies and may be difficult to detect.
- Consequence for dollar liquidity demand:
  - Either restriction on intra-group channels—greater net borrowing by branches or shrinking cross-border funding—implies branches must hold more precautionary US dollar liquidity relative to funding commitments.
  - This increases banks’ effective demand for US dollar funding and further pressures the dollar supply-demand balance.
  - Non-US banks are effectively accumulating greater persistent US dollar liquidity on the balance sheets of individual foreign bank offices.

### Key empirical and graphical elements referenced
- Online Annex Figure 5.1.1: Illustrates international non-US banks’ dollar funding channels (black arrows = open market sources; blue arrows = intra-group flows; thick solid arrows = major channels; dotted arrows = smaller/less significant).
- Online Annex Figure 5.1.2: Liabilities structure—Average by Booking Economy (Locational Basis), by Category of Institution (Percent of total liabilities, end-2018). Panels:
  1. Foreign Branches
  2. Foreign Subsidiaries
  3. Domestic Banks
- Online Annex Figure 5.1.3: Change in the Characteristics of Intragroup Relationships–Branches in Selected Host Economies (Locational Basis), 2017 vs. 2010.

### Econometric methodology (ONLINE ANNEX 5.2)
- Definition of cross-currency basis (CCB):
  - CCB is a measure of violation of covered interest parity (CIP): the price difference between funding directly in US dollars and funding synthetically in US dollars using currency forward contracts.
  - The 3-month Libor basis tenor is used throughout the chapter as it represents the main benchmark in market practice and for a large part of the literature.
  - Spot, forward and LIBOR rates used to construct the CCB for different currencies are obtained from Bloomberg.
  - Sample of currencies included in panel 1 of Figure 5.3: Australian dollar, Brazilian real, Canadian dollar, Swiss franc, Danish krone, euro, British pound, Hong Kong dollar, Japanese yen, South Korean won, Indian rupee, Mexican peso, Malaysian ringgit, Norwegian krone, New Zealand dollar, Philippine peso, Russian ruble, Swedish krona, Thai baht, Turkish lira, and the South African rand.
- Drivers and amplification effect of the cross-currency funding ratio (CCFR):
  - Purpose: understand which major factors drive the CCB and whether the CCFR serves as an amplifier.
  - Sample period: 01/01/2000 to 03/01/2018.
  - Currencies used in this exercise (data availability): Australian dollar, Canadian dollar, Swiss franc, euro, British pound, Hong Kong dollar, Japanese yen, Indian rupee, Malaysian ringgit, Swedish krona.
  - Baseline level regression estimated:
    - CCB_i,t = alpha_i + beta' X_i,t + epsilon_i,t
    - Where alpha_i are currency fixed effects; X_i,t is the set of drivers; epsilon_i,t is the error term.
  - Set of drivers includes:
    - VIX
    - US dollar index
    - Home economy default probability (measured with expected-default frequency (EDF) over 1-year time horizon)
    - FX implied volatility
    - Bid-ask spread
    - Term-spread differential (defined as the difference between the term spread in the home economy and the term spread in US; term spread = 10-year government bond yield – 3-month government bond yield)
    - Libor-OIS spread
    - Cross-currency funding ratio (CCFR)
  - Estimation details:
    - Regressors are standardized prior to estimation.
    - “High” and “low” levels of CCFR are defined as the fourth and first quintiles, respectively, of the historical distribution of the CCFR.
    - Values shown in panel 2 of Figure 5.3 refer to the aggregate effect of the baseline coefficient (beta') of each x_j and estimated cross-term coefficients (lambda') conditioning on “high” and “low” CCFR.
    - Absence of panel unit roots tested with Im–Pesaran–Shin test.
    - Robust standard errors are used; results are robust with Newey-West standard errors.
- Effects of regulation and reserve holding:
  - Variant of baseline regression estimated to test effects of introduction of regulatory changes on coefficients on CCFR.
  - Regression includes month fixed effects (alpha_m) to control for quarter- and year-end effects.
  - Sample period for this exercise: 01/01/2012 to 12/31/2016.
  - Focus currencies: Australian dollar, Canadian dollar, Swiss franc, euro, British pound, Japanese yen, Swedish krona.
  - Regulatory reforms coded as dummies:
    - Stress-VaR 01/01/2013 – 12/31/2013
    - Supplementary leverage ratio 01/01/2014 – 12/31/2014
    - Liquidity coverage ratio 01/01/2015 – 12/31/2015
    - Money market mutual fund reform (MMMF) 01/01/2016 – 12/31/2016
  - Values shown in panel 2 of Figure 5.4 refer to the effect of the baseline coefficient (beta') and each cross-term coefficient (lambda') with the dummy for the time of each regulation change.

*Source: Chapter 5 and Online Annex 5.2, October 2019 Global Financial Stability Report (GFSR).*

### 2. USD FUNDING COSTS AND HOME COUNTRY FINANCIAL STRESS

### 2. USD FUNDING COSTS AND HOME COUNTRY FINANCIAL STRESS

### Research questions / aims
- How do banking sector stability and domestic financial conditions in home economies of global non-US banks respond to changes in US dollar funding costs?
- To what extent does this relationship depend on non-US banks’ share of US dollar business?
- To what extent does it depend on USD funding fragility?
- What policy-related variables serve to mitigate the association of variations in the US dollar funding cost and banking sector stress/domestic financial conditions in home economies of global non-US banks?

### a. Baseline regression specification
- Dependent variable: change in country-level financial stress ∆1 Y_i,t, captured by either:
  - average 1-year ahead bank probability of default PPD_i,t, or
  - financial conditions index FCI_i,t.
- Key regressor: change in the US dollar funding cost CCBS_C_i,t (the negative of the CCB_i,t).
- Controls (vector CnControls_i,t include):
  - Macro controls: interest rate differentials, inflation, logarithm of nominal effective exchange rate (NEER), logarithm of VIX, volatility of logarithm of VIX, and real GDP growth.
  - Bank controls: bank capital to assets ratio, cash to assets ratio, return to assets ratio, deposit to assets ratio, net loan to assets ratio, cost to income ratio.
- All controls are moving averages from quarter t-4 to t-1.
- Country fixed effects: α_i.
- Time variation: four period dummies P1 to P4 and interactions with ∆1 CCBS_C_i,t to capture changing associations over time; P1 (2007≤Year_C ≤2009), P2 (2011≤Year_C ≤2012), P3 (Year_C ≥2013), P4 (Year_C = ResC).
- Estimation: standard errors clustered at the economy level.
- Representative exercise: association between a contemporaneous increase in the change of the US dollar funding cost by 50 basis points and ∆PD or ∆FCI (Figure references in source).

### Baseline cross-border variant (link to recipient economies)
- A similar exercise traces spillovers to recipient economies that receive cross-border dollar credit from global non-US banks.
- Example: association between a 50 basis point increase in US dollar funding costs across a recipient economy's top 10 lending economies and the probability of default of 38 recipient economies; comparison of top 10 main borrowers versus the rest.

### b. Amplification effects: share of USD assets and funding fragility
- Augmented specification adds the share of US dollar assets to total assets S_i,t−1~−4 and its interaction with ∆1 CCBS_C_i,t.
  - Share S is the average for economy i between quarter t-4 and quarter t-1.
- Representative scenario: association between a 50 basis point increase in US dollar funding costs and the change in the probability of default when the share of US dollar business in quarter t-4 to quarter t-1 is low versus high (full-sample distribution) (Figure 5.7 Panel 1).
- Funding fragility specification replaces share with fragility F_i,t−1~−4 (average between quarter t-4 and quarter t-1) where fragility is measured by one of:
  - CCFR (cross-currency funding ratio),
  - US dollar liquidity ratio (LR),
  - US dollar stable funding ratio (SFR).
- Asymmetry: differentiate responses when the funding gap (CCFR/CCFG) is positive or negative; specification includes indicator [CCFR>0] and interactions to capture asymmetric effects.
- Empirical comparisons shown in:
  - Panel 2 of Figure 5.7: effect when CCFR or ratio of cross-currency funding gap (CCFG) to US dollar assets in quarter t-4 to t-1 is low versus high (distribution of positive-value CCFGs).
  - Panel 3 of Figure 5.7: effect when LR or SFR in quarter t-4 to t-1 is low versus high relative to historical distribution for each economy.

### c. Policy-related variables as mitigators (home-country bank health, swap lines, reserves)
- Bank health (BH) specification:
  - Introduce bank health variable CCBH_i,t−1~−4 averaged over t-4 to t-1 and its interaction with ∆1 CCBS_C_i,t.
  - Bank health metrics considered: capital-asset ratio (capital), cash-assets ratio (liquidity), return on assets (ROA).
  - Panel 1 of Figure 5.9: association between a 50 basis point increase in ∆1 CCBS_C_i,t and ∆PD comparing low vs high bank health (quarter t-4 to t-1, full-sample distribution).
- Swap line (Fed) specification:
  - Dummy SSWAP_i,t indicates if there is a swap liquidity agreement between central bank in economy i and the Federal Reserve in quarter t.
  - Model includes separate coefficients on ∆1 CCBS_C_i,t when SSWAP_i,t = 0 and when SSWAP_i,t = 1.
  - SSWAP_i,t varies across economies and time.
  - Panel 2 of Figure 5.10: compares association for a 50 basis point increase in US dollar funding costs when a swap line arrangement is in effect versus when it is not.
- International reserves (R) specification:
  - Introduce R_i,t−1~−4 (international reserve holdings averaged over t-4 to t-1) and interactions with ∆1 CCBS_C_i,t and fragility F_i,t−1~−4.
  - Panel 3 of Figure 5.10: transmission effect of US dollar funding fragility—with LR or SFR evaluated at their median—on ∆PD when home central bank’s international reserve holding is low or high by historical standards.
- All regressions: standard errors clustered at the economy level.

---

### 3. USD FUNDING COSTS AND CROSS-BORDER LENDING

### Research questions / aims
- How large is the spillover impact on cross-border US dollar lending of shocks to US dollar funding conditions?
- Do EM lenders cut cross-border lending by more when the US dollar funding cost increases?
- Are EM recipients more vulnerable to their lending partners’ US dollar funding shocks?
- Do we observe non-linear relations in the transmission channel?
- Can recipient countries fully substitute US dollar credit across different sources (other lending countries, local position US dollar credit, or other currency cross-border lending) after their cross-border US dollar lenders experience funding cost shocks?
- Do US dollar funding fragilities amplify the cross-border spillover transmission channel?
- Do bank conditions in home economies act as mitigators or amplifiers?
- Are home countries’ central bank swap line arrangements and international reserve holdings effective in mitigating the negative impact of USD funding shocks?
- Are recipient countries with higher historical corporate, sovereign, or bank risks more vulnerable?

### a. Baseline regression for cross-border lending spillovers
- Dependent variable: log( L_i,j,t+1,t+4_USUSD ) × 100 = average cross-border US dollar bilateral lending during t+1 to t+4 from home economy i to recipient economy j (expressed as log percentage).
- Key regressor: ∆4 CCBS_C_i,t = one-year cumulative change in US dollar funding costs (the negative of the CCB).
- Controls CnControls_i,t include:
  - Home macro controls: domestic credit growth rate, real GDP growth rate %∆RealGDP, logarithm of nominal exchange rate log(FX).
  - Home banking controls: log total banking sector assets, average equity-asset ratio, deposit-assets ratio, cash-assets ratio, return on assets (ROA).
- Interpretation: β < 0 implies shocks to USD funding costs disrupt cross-border USD lending (panel 1 of Figure 5.6).

### Heterogeneity tests: EM lenders and EM recipients
- Specification with EM lender dummy (EM_i) or EM recipient dummy (EM_j):
  - If β2 < 0 with EM_i, EM lenders cut cross-border US dollar lending by more than other lenders when their USD funding cost increases.
  - If β2 < 0 with EM_j, EM recipients suffer more adverse impacts than other USD recipients.

### Nonlinear specifications
- Specification (15) includes a quadratic term (∆4 CCBS_C)^2 to test for non-linear effects: β2 ≠ 0 implies non-linear effects.
- Specification (16) interacts ∆4 CCBS_C with an indicator of initial funding cost being high/low (initial CCB_i,t−1 < S−Sp_C threshold) to test whether higher initial funding costs amplify responses; β2 < 0 would indicate larger effects when initial funding costs are higher (lower initial CCB).

### b. Substitutability tests
- Test (i): substitution across lending partners
  - Dependent variable: log( AAA_CAdapter_-i,j,t+1,t+4 ) = average total cross-border US dollar borrowing by recipient j from all lending partners except i during t+1 to t+4.
  - Regressors include ∆4 CCBS_C_i,t and the weighted sum ∑ S_-i,j,t ∆4 CCBS_C_-i,t−i where S_-i,j,t is the share of cross-border USD borrowing from partner −i in total USD borrowing.
  - Expectation: β1 > 0 indicates a substitution effect across US dollar lending partners.
  - Degree of substitution θ = −β1 / β (where β is from baseline specification (13)); θ = 1 implies full substitution, 0<θ<1 implies partial substitution.
  - Degree of substitution presented in panel 2 of Figure 5.6.
- Test (ii): substitution to local position USD borrowing or other-currency cross-border borrowing
  - Specification (18): dependent variable ∆(CC + LG)_j,t+4 / (CC_j,t + LG_j,t) regressed on weighted sum ∑ S_i,j,t ∆4 CCBS_C_i,t and controls; if β1 < 0 recipient j cannot fully compensate loss by local position USD borrowing or other-currency cross-border borrowing.
  - Specification (19): dependent variable ∆AllCurrencyCrossBorder_j,t+4 / AllCurrencyCrossBorder_j,t regressed similarly to test substitution to other-currency cross-border credit.
  - Degrees of substitution computed as θ = −β1_LG / β1_CC for local position vs cross-border USD, and θ = −β1_Other / β1_USUSD for USD vs other-currency cross-border credit.
  - Results and degrees of substitution presented in panels 2 and 3 of Figure 5.6 (EM recipients shown separately).

### c. Amplifiers and mitigators of effects on cross-border lending
- Funding fragility as amplifier/mitigator (specification 20):
  - Dependent variable: log(L_i,j,t+4_USUSD) ×100.
  - Regressors include:
    - ∆ CCBS_C_i,t+3,t+4 (short cumulative change),
    - ∆1 CCBS_C_i,t+3 × F_i,t−1 (interaction term),
    - F_i,t−1 (fragility quintile; CCFR, LCR, SFR expressed as quintiles across historical levels).
  - Lagging fragility by 4 quarters to mitigate endogeneity.
  - Expected signs:
    - β2 < 0 for amplifiers (e.g., CCFR),
    - β2 > 0 for mitigators (e.g., LCR, SFR, HQLA ratio).
  - Results summarized in panel 4 of Figure 5.7.
- Bank health as mitigator (specification 21):
  - Regressors include ∆ CCBS_C_i,t+3,t+4, ∆1 CCBS_C_i,t+3 × CCB_i,t−1, and CCB_i,t−1 (bank health measure expressed as quintiles across historical levels; capital ratio, deposit-to-asset ratio, liquidity ratio, ROA).
  - Expectation: β2 > 0 if bank health mitigates USD funding cost shocks.
  - Results shown in panel 2 of Figure 5.9.

### d. Policy-related variables as mitigators (swap lines and reserves for cross-border lending)
- Specification (22) tests whether swap line arrangement or international reserve holdings mitigate disruptive effects of USD funding cost shocks:
  - Regressors include ∆4 CCBS_C_i,t, ∆4 CCBS_C_i,t × PP_i,t−1, and PP_i,t−1 where PP_i,t−1 is either:
    - dummy for existence of swap line arrangement between lending economy i and the Fed at t-1, or
    - quintiles of the ratio of international reserves to GDP across lending country i’s historical levels.
  - If β2 > 0 then swap lines and/or international reserve holdings mitigate disruptive effects of the US dollar funding shocks.
  - Empirical comparisons presented in panel 4 of Figure 5.7 (and referenced panels in Figures 5.9–5.10).

*Source: onlineannex5 - 2. USD FUNDING COSTS AND HOME COUNTRY FINANCIAL STRESS (International Monetary Fund | October 2019).*

### 5.10 for the results.

### onlineannex5 - 5.10 for the results.

### Testing for Differences Across Economy Groups and for Recipient Economy Risk
- Specification used to test whether recipient economies’ historical risk matters for the transmission of USD funding shocks:
  - Equation (23): log(LL_{i,j,t+1,t+4}^{USD})×100 = α_{j,t} + ρ_{i,j} + β1 Δ4 CCB_sCC_{i,t} + β2 Δ4 CCB_sCC_{i,t} R_isR_{j,t-1} + η′ CCAnnCov_{i,t} + ε_{i,j,t+4}
  - 푅푅푖푖푠푠 푅푅_{푗푗,푡푡−1} is the historical average of the risk measure (corporate spreads, sovereign spreads, and banking sector EDF) from 2000Q1 up to t-1, expressed as a dummy equal to 1 if higher than the top 20 percentile across all economies in the same quarter.
  - Expectation: β2 < 0 if recipients’ risk amplifies the negative impact of USD funding shocks.
- Horse-race specification across the three risk measures:
  - Equation (24): log(LL_{i,j,t+1,t+4}^{USD})×100 = α_{j,t} + ρ_{i,j} + β1 Δ4 CCB_sCC_{i,t} + ∑_{C=1}^3 β_C Δ4 CCB_sCC_{i,t} × R_isR_{j,t-1}^C + η′ CCAnnCov_{i,t} + ε_{i,j,t+4}
  - R_isR_{j,t-1}^C denotes one of the three risk measure dummies.
- Note: Results will be presented in the forthcoming working paper.
- Footnote: Results are unchanged in signs if deciles of risk measures are used, with slightly lower significance levels; possible non-linearity of recipient country risk impact.

### Marginal and Total US Dollar Funding Cost
- Objective: assess whether the CCB is a reasonable proxy for marginal US dollar funding cost and whether it is related to total US dollar funding costs.
- Data and sample:
  - End-of-month portfolios of US money market mutual funds from CRANE data, Jan 2012 to Dec 2018.
  - Sample limited to 73 non-US banks who finance roughly one trillion of their US dollar activities in US dollar money markets.
  - Funding instruments: Certificates of Deposits, Commercial Paper, and Repurchase Agreements.
  - Fund coverage: 461 funds managed by 81 fund families.
  - Level of observation: individual contract (e.g., a repo).
  - Dataset based on regulatory filings; includes universe of US money market mutual funds.
- Baseline econometric specification:
  - Equation (25): Y_{i e C d_{i j D t}} = γ1 ⋅ CCON_i t + γ2 ⋅ CCON_{i t−1} + γ3 RemainderMaturity_{i j D t} + γ4 log(v_{i j D t}) + β ⋅ X_{i,t} + δ_i × ν_t + F_{sReSe C_D} × ν_t + μ_i + ε_{i j D t}
  - Dependent variable: Y = yield to maturity of contract as reported to the SEC.
  - Remainder maturity: remaining maturity in days.
  - log(v) = natural logarithm of the total face value of the contract in USD.
  - X_{i,t} = vector of bank, currency and country controls.
  - IsRepo dummy for repo contracts.
  - Fixed effects: fund × month; bank; Repo × time; time fixed effects.
  - Standard errors clustered on currency and time.
- Main empirical finding:
  - A ten-bp widening of the CCB is associated with an increase in wholesale funding costs by one basis point in the following month in the baseline specification.
  - Footnote clarifies: γ2 is negative and statistically significant and γ1 is not significant. Since the basis is negative for most currencies a decrease in the CCB is to be interpreted as a widening of the basis. Baseline test limits sample to currencies with a negative basis throughout the sample period (effectively excluding AUD, CNY, SGD, KRW).
- Robustness of the result:
  - Results robust to including bank × fund fixed effects, banks 1-year expected default frequency (limits sample to listed banks) and country controls such as the country’s baseline interest rate and CCFR of the country’s banking system.
- Note: US MMFs are not the sole provider of USD wholesale funding to non-US banks; Eurodollar interbank market also used (see Online Annex 3.1 for more on funding channels).

### Robustness Analysis
- Motivations: address potential simultaneity/reverse causality bias between CCFR and CCB, and between CCB and home-economy financial stress.
- Robustness exercises:
  1. Unrestricted panel VAR treating variables as endogenous and interdependent; Impulse Response Functions (IRFs) estimated to corroborate that the basis responds to shocks to the CCFR.
  2. Instrumental-variable regressions: US monetary policy shocks used as an instrument for the basis in determining home country probability of default; yields a positive and significant relationship between the instrumented basis and probability of default in the home economy.
  3. Euro area-only regressions: exploit common basis across Euro Area countries but differing probabilities of banking default/domestic conditions; positive and significant effect of increases in US dollar funding costs on financial stress persists even when restricting sample by excluding Germany and France.
  4. Time fixed effects (TFEs) and inclusion of various time-varying macroeconomic controls; full set of robustness tests including TFEs confirm key message: conditional on US dollar funding fragility, tightening in US dollar funding conditions will be transmitted to stress in home economy of non-US global banks.
  5. Control for non-banking financial stress: regressions that control for sovereign CDS spreads; results broadly robust.
- Core econometric identification: difference-in-difference strategy used to measure amplification/mitigation effects of US dollar funding fragility and other factors; this design is not affected by possible simultaneity.

### Data and Sources
- Main datasets and transformations summarized (variables, descriptions, sources):
  - BID-Ask Spread: unit point difference between ask price and bid price — Bloomberg. IMF staff estimates.
  - Cross-Currency Basis: difference between the direct dollar funding and the synthetic dollar funding — Bloomberg, IMF staff estimates.
  - Default Probability: country average expected default frequency (EDF) over 1-year — Bloomberg. IMF staff estimates.
  - Financial Conditions Index: methodology in Annex 3.2 of October 2017 GFSR — IMF staff estimates.
  - LIBOR-OIS spread: unit point difference between LIBOR and OIS rates — Bloomberg, IMF staff estimates.
  - Term spread differential: Difference between 10-year government bond yield and the 3-month government bond yield — Haver Analytics.
  - US dollar index: Trade Weighted U.S. Dollar Index: Broad — Haver Analytics.
  - VIX: Chicago Board Options Exchange Volatility Index — Haver Analytics.
  - US dollar bilateral lending (cross-border): Amount from locational banking statistics (national basis) — BIS.
  - US dollar bilateral lending (local position): Amount in local positions from locational banking statistics (national basis) — BIS.
  - Other currency bilateral lending (cross-border): Amount denominated in all other currencies from locational banking statistics (national basis) — BIS.
  - Swap line arrangement: dummy when central bank swap line with the Fed is available — Board of Governors of the Federal Reserve System.
  - International Reserves: Reserve Assets (Includes reserve position in the Fund and SDR holdings) — International Financial Statistics.
  - Probability of Default of Banking Sector: banking sector expected default frequency (EDF) over 1-year ahead at country level — The National University of Singapore Risk Management Institute (RMI).
  - Interest rate differentials (Home - US): 1-year government bond yield difference — International Financial Statistics.
  - Inflation differentials (Home - US): inflation difference based on quarterly CPI (period average) — International Financial Statistics.
  - Nominal Effective Exchange Rate: nominal effective exchange rate of United States in logarithm — International Financial Statistics.
  - Real GDP: gross domestic product, constant prices in national currency in logarithm — International Financial Statistics.
  - Capital ratio: capital to assets ratio of banking sector at nationality level — Fitchconnect, Bankscope, IMF staff estimates.
  - Liquidity ratio: cash to assets ratio of banking sector at nationality level — Fitchconnect, Bankscope, IMF staff estimates.
  - ROA: return over assets ratio of banking sector at nationality level — Fitchconnect, Bankscope, IMF staff estimates.
  - Deposit ratio: deposit to assets ratio of banking sector at nationality level — Fitchconnect, Bankscope, IMF staff estimates.
  - Cost ratio: cost to income ratio of banking sector at nationality level — Fitchconnect, Bankscope, IMF staff estimates.
  - Net loan ratio: net loan to assets ratio of banking sector at nationality level — Fitchconnect, Bankscope, IMF staff estimates.
  - EM: dummy if the country belongs to the emerging market country list — Brauning and Ivashina (2019).
- Source: IMF staff.

### Online Annex 5.3 — Non-US Banks’ US Dollar Balance Sheet and Funding Fragility Measures: Methodology
- Purpose: describe methodology and concepts behind US dollar funding fragility measures of non-US banks (USD liquidity ratio (LR), USD stable funding ratio (SFR), USD cross-currency funding ratio (CCFR)).
- Concepts and aggregates:
  - International Position (IP): USD operations outside the United States, capturing USD cross-border positions of non-US banks and US dollar local positions of non-US banks outside the United States.
  - Branch operations in the United States.
  - IP + Branches in the United States (IP+B): aggregate of the two above.
  - Foreign Position (FP): balance sheets constructed by adding subsidiary operations in the United States.
  - USD funding fragility can also be estimated at the foreign position level.
- Sample coverage (selected jurisdictions listed in Online Annex Table 5.3.1):
  - Panel A (sample with all three funding fragility measures): Australia, Austria, Canada, France, Germany, India, Italy, Brazil, China, Cyprus, Hong Kong SAR, Luxembourg, Malaysia, Japan, Korea, Netherlands, Spain, Sweden, Switzerland, United Kingdom, Mexico, Norway, Singapore, South Africa, Turkey, Russia.
  - Panel B (sample with CCFR only) — jurisdictions included as in table.
- Data sources and construction:
  - Due to data limitations on bank-level USD balance sheets outside the United States, analysis is performed at jurisdiction level.
  - BIS locational statistics on nationality basis (confidential database) used to capture USD operations outside the United States; includes currency composition and counterparty sector granularity.
  - US branch operations constructed via bottom-up aggregation based on FFIEC 002 regulatory filings (nationality basis).
  - US subsidiary operations constructed using FFIEC 031/041 call report filings for non-US-owned subsidiaries via S&P Global Market Intelligence.
- Note on limitations:
  - BIS consolidated statistics lack required granularity; BIS confidential locational statistics used instead.
  - United States does not report non-US banks’ operations within the United States to BIS locational statistics; hence bottom-up construction used.
- Empirical observation:
  - Inclusion of US subsidiaries of non-US banks does not affect the analysis materially (comparison of liquidity and stable funding ratios across IP+B and FP concepts).
- Visualization:
  - Online Annex Figure 5.3.1: International Position vs. Foreign Position (graphical representation).
  - Online Annex Figure 5.3.2: US Dollar Liquidity and Stable Funding Ratios — inclusion of US subsidiaries of non-US banks; 1. Liquidity ratio, 2018 (Percent).

*Source: IMF staff.*

### 2. Stable Funding Ratio, 2018

### 2. Stable Funding Ratio, 2018

### Construction of the US dollar liquidity and stable funding ratios
- Liquidity (LR) and stable funding (SFR) ratios follow the methodology developed in Chapter 1 of the April 2018 GFSR (see Online Annex 1.2 of that chapter).
- Analysis constrained by lack of granularity in BIS statistics when constructing US dollar balance sheets; additional LR estimates using alternative outflow assumptions are computed for robustness (see Online Annex Figure 5.3.3).
- Minor modifications made to balance sheet components to refine estimated ratios.

### Adjustments to the deposit component
- April 2018 approach: estimated deposit component by subtracting USD-denominated bonds and short-term debt securities issued by banks from liabilities to nonbanks in BIS locational banking statistics.
- Refinement in this analysis: only a fraction of US dollar-denominated bonds and short-term debt securities issued by banks is subtracted.
- The fraction corresponds to an estimate of the portion held by the nonbank sector.
- Due to lack of data on holders, assumptions:
  - If counterparty breakdown of USD-denominated debt securities is available on a residency basis in BIS locational banking statistics, assume the nationality basis ratio of debt securities held by the nonbank sector to debt securities held by all sectors is the same as in residency basis.
  - When counterparty sector breakdown is not available on residency basis, assume the nationality basis ratio of debt securities held by the nonbank sector to debt securities held by all sectors is same as the nationality basis liabilities to the non-bank sector as a ratio of liabilities to all sectors.
- Footnote: When the residency basis series exists as a shorter series, we use the nationality basis liabilities to nonbanks as a ratio of liabilities to all sectors to splice the former series.

### Sensitivity of LR to deposit assumptions
- Online Annex Figure 5.3.3: The median liquidity ratios estimated using less stringent alternative assumptions varies by about fifteen percentage points from the baseline.
- Note: The figure corresponds to the international position plus US branches concept.
- Scenario assumptions for run-off factors for transaction and non-transactional deposits at branch-level:
  - Scenario 1: 3 and 5
  - Scenario 2: 5 and 10
  - Scenario 3: 5 and 25

### Inclusion of foreign exchange swaps in SFR
- In addition to foreign exchange swap payables (defined in April 2018 as the net position when US dollar claims are greater than liabilities and included in the numerator), an estimate of foreign exchange swap receivables are now included in the denominator.
- The swap receivable element is the net position when USD liabilities are greater than the claims.

### LR and SFR formulae (as presented)
- Liquidity ratio (LR) estimated as:
  퐿퐿푅푅  = 
  퐵퐵푖푖퐻퐻ℎ 푄푄퐴퐴푌푌퐶퐶푖푖퐶퐶푦푦 퐿퐿푖푖퐿퐿 퐴퐴푖푖푑푑 퐴퐴푠푠푠푠푒푒퐶퐶푠푠 
  푁푁푒푒퐶퐶 푝푝푌푌푠푠 ℎ 퐶퐶퐴퐴퐶퐶푓푓퐶퐶퐶퐶푆푆푠푠
  ∗100
- Definition used in construction:
  푁푁푒푒퐶퐶 푝푝푌푌푠푠 ℎ 퐶퐶퐴퐴퐶퐶푓푓퐶퐶퐶퐶푆푆푠푠
  = 퐶퐶푌푌푠푠ℎ 퐶퐶퐴퐴퐶퐶푓푓퐶퐶퐶퐶푆푆푠푠 − min (푝푝푌푌푠푠 ℎ 푖푖푛푛푓푓퐶퐶퐶퐶푆푆푠푠, 0.75 ∗ 푝푝푌푌푠푠 ℎ 퐶퐶퐴퐴퐶퐶푓푓퐶퐶퐶퐶푆푆푠푠)
- Stable funding ratio (SFR) estimated as:
  푆푆퐹퐹푅푅
  = 
  푃푃푒푒푆푆퐶퐶푠푠푖푖퐶퐶푠푠+퐶퐶퐶퐶푛푛 퐻퐻 퐶퐶푒푒퐶퐶퐴퐴 푠푠푒푒푝푝퐴퐴퐶퐶푖푖퐶퐶𝑖𝑖𝑒푒   푠푠+퐹퐹푋푋 푠푠푆푆푌푌푆푆푠푠 푆푆푌푌푦푦푌푌  푝푝퐶퐶푒푒
  퐿퐿퐶퐶푌푌푛푛푠푠+퐹퐹푋푋 푠푠푆푆푌푌푆푆  푠푠 퐶퐶푒푒푝푝푒푒𝑖𝑖𝑣𝑣푌푌𝑝𝑝퐶퐶푒푒
  ∗100
- Notes:
  - Cash outflows refer to interbank liabilities, other short-term liabilities, and deposits. Cash inflows refer to interbank assets. Cash outflows and inflows also incorporate applicable run-off factors.
  - SFR also incorporates applicable run-off factors.

### Balance sheet construction (summarized from Online Annex Tables 5.3.2 and 5.3.3)
- Data sources: Bank for International Settlements (consolidated and locational banking statistics, nationality and residency basis), Federal Financial Institutions Examination Council (FFIEC 002), Call report filings via S&P Global Market Intelligence.
- Non-US banks' operations outside the United States and in US branches: key balance sheet items and data series:
  - Assets:
    - A1. HQLA: Claims on US official sector, ultimate risk basis (BIS consolidated banking statistics, Table B4 for USA) minus HQLA position from US subsidiaries; supplemented with Call report filings via S&P Global Market Intelligence.
    - A2. Interbank assets: Claims on banks (BIS locational banking statistics, Nationality basis) minus intragroup claims plus US branches' interbank assets (FFIEC 002).
    - A3. Intragroup assets: Intragroup claims (BIS locational banking statistics, Nationality basis) plus US branches' intragroup claims (FFIEC 002).
    - A4. Loans: Claims on nonbanks (BIS locational banking statistics, Nationality basis) plus US branches' loans (FFIEC 002) minus A1. HQLA and minus claims on central bank (BIS locational banking statistics, Nationality basis).
    - A5. Other claims: Claims on banks and nonbanks (BIS locational banking statistics, Nationality basis) minus A1–A4 and plus US branches' other claims (FFIEC 002).
  - Liabilities:
    - L1. Interbank liabilities: Liabilities to banks (BIS locational banking statistics, Nationality basis) minus intragroup liabilities plus US branches' interbank liabilities (FFIEC 002) minus long-term and short-term debt securities (held by banks) per BIS debt securities statistics and locational statistics (Nationality and Residency basis).
    - L2. Intragroup liabilities: Intragroup liabilities (BIS nationality basis) plus US branches' intragroup liabilities (FFIEC 002) minus intragroup portions of long- and short-term debt securities (BIS).
    - L3. Bonds: Long-term debt securities issued by banks (BIS debt securities statistics, Nationality basis) plus US branches' long-term securities (FFIEC 002).
    - L4. Other short-term liabilities: Short-term debt securities issued by banks (BIS) plus US branches' short-term securities (FFIEC 002).
    - L5. Deposits: Liabilities to nonbanks (BIS nationality basis) minus long-term and short-term debt securities (held by nonbanks) per BIS debt securities statistics (Nationality and Residency basis).
    - L6. FX swaps: Claims on banks and claims on nonbanks (BIS nationality basis) minus liabilities on banks and liabilities on nonbanks (BIS nationality basis) plus US branches' FX swaps (FFIEC 002).
- For non-US banks' operations including US subsidiaries, the same structure applies with additional inclusion of US subsidiaries' items (Call report filings via S&P Global Market Intelligence) in assets and liabilities.

### Country-level results and heatmaps
- Online Annex Tables 5.3.4 and 5.3.5 present evolution of estimated LR and SFR within each economy in the sample, shown as deciles within each country.
- Heatmaps use deciles 1–10 to indicate relative LR and SFR levels across time (examples in tables span from 2005Q1 through 2018Q1 with decile values recorded for multiple economies).

### FSAP findings on US dollar funding fragility (Online Box 5.1)
- Several recent FSAPs assessed adequacy of banking system US dollar liquidity; Basel liquidity requirements by significant currency facilitated monitoring.
- Analyses covered structural liquidity ratios (Liquidity Coverage Ratio and Net Stable Funding Ratio), asset encumbrance, and cash-flow based IMF top-down liquidity stress tests.
- LCR in US dollars used to identify potential currency mismatches under stress; no requirement to reach a predefined LCR target in US dollars, and for some major banks the ratio falls below 100 percent.
- Practices when LCR in US dollars exceeds 100 percent: banks commonly extend maturities of outflows above 30 days or use collateral swap facilities to meet end-of-month LCR reporting.
- Top-down cash-flow based USD liquidity stress tests considered scenarios including closure of unsecured and/or secured funding markets and wholesale US dollar deposit outflow rates observed during the global financial crisis and the US dollar funding crisis in Europe.
- FSAP results: some major banks vulnerable due to:
  - (i) mismatches between repo and reverse repo outflows/inflows;
  - (ii) concentrated US dollar wholesale funding (limited number of counterparties and small share of retail funding);
  - (iii) limited transferability of excess liquidity and collateral between jurisdictions during a crisis.
- FSAP conclusions and recommended additional actions:
  - (i) collect granular cash flow liquidity data by currency, entity and jurisdiction;
  - (ii) monitor US dollar funding patterns and liquidity metrics for various maturity buckets;
  - (iii) simulate the impact of specific liquidity scenarios for additional time horizons;
  - (iv) ensure that banks engaged in US dollar liquidity transformation have adequate US dollar liquidity buffers to cover wholesale cash outflows for at least a one-week horizon;
  - (v) assess risks related to US dollar liquidity trapping in foreign jurisdictions.

*Sources: Bank for International Settlements, locational banking statistics (nationality basis); Federal Financial Institutions Examination Council; S&P Global, Market Intelligence; IMF staff calculations.*

---


_Source: https://www.imf.org/-/media/files/publications/gfsr/2019/october/english/onlineannex5.pdf_