## _cr16163 — Executive Summary and Key Findings

## Source details

**Canonical URL:** [_cr16163 — Executive Summary and Key Findings](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2016/_cr16163.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2016/_cr16163.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2016/_cr16163.pdf.json)

---

### FSAP scope and approach
- Conducted comprehensive range of stress tests to assess resilience of the U.K. banking system.
- Top-down (TD) solvency stress test covered seven major U.K. banks and building societies representing over 80 percent of Prudential Regulatory Authority (PRA)-regulated banks’ lending to the real economy.
- Stress testing addressed solvency and liquidity risk and key risk types, including credit risk, market risk, sovereign risk, and funding risk.
- FSAP stress scenario focused on a broad-based dislocation in financial markets, with sizeable jumps in yield curves and spillovers to vulnerable emerging market economies; scenario could be triggered inter alia by disorderly monetary normalization in the U.S.
- Complemented the 2015 concurrent Bank of England (BoE) stress test (same set of banks, different global shock).
- Cross-validated bottom-up (BU) submissions for selected portfolios for the 2015 BoE concurrent stress test.
- Suite of liquidity stress tests conducted for 10 large firms, including three large U.K. foreign-headquartered investment banks.
- Cut-off date for FSAP tests: December 2015 (BoE concurrent stress test cut-off: December 2014).

### Key solvency stress test results
- Aggregate Core Equity Tier I (CET1) ratio:
  - Starts at 12.6 percent in December 2015.
  - Falls to a low point of 8.7 percent under the FSAP stress scenario.
- Leverage ratio:
  - Starts at 5.3 percent in December 2015.
  - Falls to a low point of 4.0 percent under the FSAP stress scenario.
- Aggregate losses:
  - Banks incur GBP 35 billion in losses in the first two years of the stress scenario.
  - All covered banks continue to meet prudential requirements throughout the stress test horizon.
- Risk-weighted assets (RWAs):
  - RWAs peak in 2017 with a 2.3 percentage points (pps) hike.
- Bank group heterogeneity:
  - Aggregate CET1 ratio of major U.K. international banks decreases by 4.3 pps at the peak of the FSAP stress scenario.
  - Aggregate CET1 ratio of major U.K. domestic banks decreases by 2.6 pps at the peak of the FSAP stress scenario.
  - International banks’ larger decline reflects stressed impairment charges in overseas exposures and larger mark-to-market losses in securities portfolios.

### Liquidity stress test results
- Sample: 10 large financial institutions (seven large U.K. firms in solvency test plus three large U.K. subsidiaries of major foreign investment banks).
- LCR outcomes:
  - All firms passed the liquidity coverage ratio (LCR) stress test under the current regulatory 80 percent hurdle rate for the standard Basel III scenario, as well as two tailored scenarios reflecting a U.K. retail deposit run and a U.K. wholesale event.
  - Aggregate LCR ratio under standard Basel III assumptions: about 140 percent.
  - Under “U.K. retail” and “U.K. wholesale” scenarios, average ratio declined to about 95 percent.
- Implied cash flow tests:
  - All firms passed five-day and 30-day implied cash flow tests on an all-currency basis.
  - Average counterbalancing capacity: over 180 percent for five-day test; under 150 percent for 30-day test.
- Single-currency analysis by the PRA:
  - All firms had sufficient liquid buffers in domestic currency (GBP).
  - Only minor shortfall in foreign currency: one firm negative in EUR; two firms slight negative shortfall in USD.

### Model validation and comparison with BoE BU submissions
- FSAP credit risk parameters broadly comparable to those generated by participating banks under BoE’s 2015 concurrent stress test for several parameters, providing independent validation.
- FSAP applied an independent quantitative method to assess BU projections submitted by banks for the 2015 BoE concurrent stress test.
- Caution in comparing BU and TD results due to differences in data, granularity, and modeling approach.
- FSAP projections exhibit somewhat more cyclical behavior than BU projections, but broadly confirm BU calculations subject to caveats.

### Conservative and modeling assumptions
- Key conservative assumptions:
  - Administrative expenses would not decrease in the stress scenario.
  - Economic hedges not allowed to operate in the trading book.
  - Fee income capped.
  - Funding-risk module assumed drying-up of market liquidity, with banks unable to pass-through increases in funding cost to loan rates.
- Some conservative choices deliberate; others driven by data constraints and TD modeling exigencies.

### Financial system context and scale
- Banking sector accounts for over 60 percent of total financial sector assets.
- U.K. financial system (sum of financial assets owned by banks and nonbank financial institutions) about GBP 20 trillion in April 2015, over 10 times U.K. annual GDP.
- Excluding derivatives and cross-border exposures of foreign-owned bank branches:
  - Total assets of the U.K. financial system: GBP 13 trillion.
  - GBP 8 trillion of those assets are held by banks.

*Prepared by the International Monetary Fund (IMF) based on the FSAP technical note and associated analysis.*

---

### FSAP coverage, data sources, and risk scope
- Coverage:
  - Includes Barclays plc, HSBC Holdings plc, Lloyds Banking Group plc, Nationwide Building Society, the Royal Bank of Scotland Group plc, Santander UK plc, and Standard Chartered plc.
  - Covered banks represent about 75 percent of total banking sector assets and over 80 percent of PRA-regulated banks’ lending to the real economy.
  - Inclusion criterion: total retail deposits ≥ GBP 50 billion as of December 2015.
- Data and sources:
  - Granular bank-submitted data: IRB portfolio breakdowns by Basel asset class, sectoral split, currency, geography; securities portfolio at security level; duration of debt portfolio; open positions by market risk factor; RWAs split by STA and IMM; projections of RWAs; maturity gaps for interest rate risk.
  - Public data: Pillar 3 disclosures, Bankscope, SNL, Bloomberg, Datastream, Markit, 2015 EU-wide transparency exercise.
  - Aggregate sources: Haver Analytics, Moody’s KMV, IFS, WEO.
- Risk coverage and methodologies:
  - Hurdle rates: baseline 7 percent CET1; adverse 4.5 percent CET1 and 6 percent Tier 1; leverage hurdle includes 3 percent Tier 1 based on U.K. leverage framework.
  - Leverage ratio in IMF FSAP calculated using quarter-end data.
  - Credit risk: separate analysis for IRB exposures calibrated for five Basel asset classes and 15 geographies (matrix of 75 credit risk models).
  - Market risk: scope included HFT, AFS, FVO positions; exclusions: hedge accounting portfolios, economic hedges, cash flow hedges.
  - Funding risk: analyzed system-wide monetary policy shocks, idiosyncratic solvency events, and contagion; FSAP limits pass-through of higher funding costs to lending rates.

---

### Scenarios and macroeconomic assumptions

### Baseline scenario
- Draws from October 2015 WEO projections; expanded to residential and commercial real estate prices for U.K., U.S., euro area core, euro area periphery; U.K. Bank rate; U.K. credit growth; equity prices for U.K. and U.S.
- Country-specific variables complemented by IMF Global Assumptions (Gas) database including six-month LIBOR by major currency and commodity price forecasts.
- Variables projected by adjusting BoE’s 2015 concurrent stress test baseline to WEO core variables up to 2020.

### Adverse scenario (features and assumptions)
- Explores sharp downturn in emerging markets; sizeable increase in rates and steepening of yield curve in the U.K. and globally; funding risk from rising LIBOR spreads, dry-up of money market issuance, FX swap market disruptions; large correction in U.K. property markets.
- Specific assumptions:
  - Balance sheet recession in the United Kingdom and financial crises in fragile emerging economies.
  - Accelerated monetary normalization in the United States with a 200 basis points policy rate hike during 2016–17.
  - Drying up of liquidity in money markets and steepening of the yield curve.
  - Stock market correction triggered by equity risk premium shocks and FX market spillovers.
  - Sudden stops in vulnerable emerging economies causing equity corrections and sharp currency depreciations.
  - Amplification in the United Kingdom from domestic demand shock, property market confidence loss, funding stress, and decompression of term premium.
- U.K. GDP path under adverse scenario:
  - U.K. GDP contracts by 1.6 percent in 2016.
  - U.K. GDP reaches a peak deviation from baseline levels in 2017 at -7.5 percent.
- Scenario focus differences with BoE:
  - BoE 2015 scenario: synchronized global downturn hitting Asia and euro area.
  - FSAP scenario: disorderly monetary normalization in the U.S. with broad-based market dislocation and spillovers.
  - Severity: IMF scenario = 2.13 standard deviation move in two-year cumulative real GDP growth over first two years; BoE = 2.04 standard deviation move.

---

### Credit risk model, PDs, LGDs, and EAD

### Credit risk model and PD projection approach
- Expected losses computed as EAD * LGD * PD (notation preserved).
- Two-step PD projection:
  - Step 1: Build time series of PD proxies by asset class and geography using market-based PDs and write-off data for U.K. exposures.
  - Step 2: Econometric forecast of bank-specific PDs using banks’ Pillar 3 disclosures as dependent variable and projected PD proxies as main driver.
- PD proxies and data sources:
  - Moody’s Analytics average one-year expected default frequency (EDF) mapped to corporate, financials, and consumer groups.
  - BoE write-off rates on U.K. exposures by main asset class.
  - Sovereign PD proxy from sovereign yields.
  - For overseas exposures, Moody’s one-year EDF for Canada, China, France, Germany, Hong Kong SAR, India, Ireland, Italy, Korea, Netherlands, Singapore, South Africa, Spain, and U.S.
- Econometric/statistical approach:
  - VAR estimation techniques and nonlinear PCA included global conditions and spillovers.
  - Scale: over 900 credit risk specifications; over 700 bilateral VARs; 150 PCA analyses per factor; 50 multivariate VARs per PD proxy/asset/geography.
- Key PD projections:
  - Projected peak for corporate EDF: 5.4 percent in 2017:Q2.
  - Projected peak for write-off rates for U.K. mortgages: 0.08 percent in 2017:Q1 (twofold baseline rate levels).

### LGD methodology for U.K. mortgages
- Fair-value option Merton-based approach: LGD nonlinear in house price at default VT and original LTVt.
- LGD projections depend on:
  - Distribution of original LTV by vintage;
  - Outstanding value net of amortization;
  - Assumed house price path;
  - Forced sales discount.
- Empirical origination LTV distribution:
  - At end-2015, 0.1 percent of regulated mortgages have LTV ratios over 95 percent (relative to over 6 percent in 2007).
  - High-rated mortgages with LTV <= 75 percent account for over 60 percent of originated mortgages at end-2015 (about 50 percent in 2007).
- House price path:
  - Largest yoy decline at end-2016; peak-to-trough decline reaches 20 percent by 2018:Q3.
- Forced sales discount:
  - Conservative forced sales discount set at 30 percent of fair-value U.K. residential property prices.
- Under adverse scenario:
  - Average LGD shock set at 3.5 percent.
  - LGD by vintage non-monotonic: higher for pre-crisis vintages and recent vintages.
  - Estimated average initial maturity for pre-crisis loans between five and seven years, yielding no material outstanding volume prior to 2008:Q3.

### LGD calibration and checks for non-mortgage portfolios
- Robustness checks:
  - Maximum one-year increase in IRB LGDs for U.K. mortgages post-crisis about 3 pps (HSBC Pillar 3).
  - Write-off rate peaks for secured lending to individuals: around 0.14 percent in 1997:Q3 and 2009:Q1.
  - Using estimated PD for U.K. mortgages of 1 percent, a 0.14 percent write-off rate implies LGD of 14 percent, approximating FSAP average stressed LGD.
- Non-mortgage loans:
  - Adverse LGDs calibrated for first two years; recovery over outer three years.
  - Bank-by-bank LGD projections derived from initial bank-specific LGD ratios and aggregate LGD shifts.

### EAD projection, FX structural risk, and credit-line triggers
- EAD drivers include balance-sheet assumptions, structural FX risk, and triggered credit lines and guarantees.
- Currency pairs considered:
  - EUR and USD for GBP-reporting banks.
  - EUR and GBP for USD-reporting banks.
- Credit growth determined by EAD-weighted nominal GDP growth across relevant jurisdictions.
- Calibration for triggered credit lines and guarantees: maximum increase in off-balance-sheet exposures to EAD ratio for corporates reported in Pillar 3 over 2008–15 ≈ 40 percent.

---

### Market risk, traded-risk shock calibration, and P&L drivers

### Market risk scope and calibration
- Scope: sovereign debt securities in 22 jurisdictions; market stress in 4 non-sovereign corporate bond indices (EU IG, EU HY, U.S. IG, U.S. HY).
- Shocks hit throughout five-year horizon; rebalancing disallowed.
- Baseline sovereign YTM changes from WEO 10-year projections adjusted with Bloomberg generic rates.
- Adverse sovereign yield hikes by 2017 (largest reported):
  - United States: 310 bps
  - Ireland: 205 bps
  - Hong Kong SAR: 192 bps
  - U.K.: 168 bps
- Corporate spread shocks by 2016 under adverse path:
  - U.K. HY rises to values including 20.78 (selected figures reported).
  - U.S. IG adverse includes 7.03 (selected figures reported).
- Securities treatment:
  - Full revaluation of HFT, AFS, and FVO; hedge accounting portfolios and economic hedges excluded.
  - AFS constitute 67 percent of aggregate debt securities; HFT 26 percent; HTM 7 percent.
  - AFS portfolio = 9.7 percent of total assets; trading book = 3.8 percent of assets; HTM = 1.0 percent of assets.
  - U.K. sovereign securities = 2.6 percent of assets; other sovereign securities = 8.5 percent; corporate securities = 3.4 percent.

### P&L and profitability context
- Post-crisis deleveraging; profits below pre-crisis levels.
- Average post-tax return on equity (ROE) fell from 17 percent in 2006 to 6 percent in March 2015.
- Asset-weighted average price-to-book (P/B) at end-2015: 0.8 (peak 1.9 in July 2007).
- Five-year CDS at end-2015: 69 basis points.
- Drivers of NII and NIMs under stress:
  - Limited pass-through of funding costs to borrowers.
  - Shrinking income base as performing loans migrate into defaulted categories.
  - Potential spike in funding costs; unstable funding structures exposed to credit-sensitive investors.
- NIM drivers and calibration:
  - Weighted average NIM of U.K. major banks ~2.8 percent in December 2015.
  - Deposit-rate elasticity to LIBOR 6m: 0.5 (regression R2: 0.63, 1999 Q1–2015 Q4).
  - NIM elasticity: LIBOR 6m 0.2; slope of yield curve -0.1 (regression R2: 0.4, 1999:Q1–2015:Q4).
- Dividend payout rules considered:
  - Econometric approach: payout ratio peaks ~80 percent in 2017 in adverse before converging to ~50 percent by 2020.
  - 2016 EU-wide approach: payout = max(30 percent, median observed payout in profitable years over past five years).
  - Fixed payout rule of 30 percent considered for consistency.

---

### Funding costs, contagion, and liquidity structure

### Funding structure shifts
- Since 2008:
  - Major U.K. banks’ funding from customer deposits increased by about GBP 250 billion.
  - Wholesale funding declined by over GBP 1.3 billion.
- Asset-weighted average loan-to-deposit (LTD) ratio: 92 percent in December 2015.

### Econometric findings on funding costs and contagion
- Funding-cost proxy: effective interest rate on interest-bearing liabilities; alternative proxy: five-year senior single name CDS spreads.
- Key estimated impacts:
  - A 1 percentage point rise in unemployment pushes up bank funding costs by around 26 bps (13 bps bi-annual).
  - A 1 percentage point decrease in policy or LIBOR rate reduces interest payments by about 44 bps (22 bps biannual).
  - A 1 percentage point rise in NPL ratio leads to a rise in CDS of around 12 bps.
  - For average bank, a rise in 100 CDS spreads in rest of U.K. system associated with rise in bank spreads of 63 bps.
- Contagion:
  - Contagion material driver of CDS spreads; robust coefficient positive and significant at 1 percent.
  - Contagion not significant determinant of effective interest rates once money market rates included.

### FSAP add-on to funding costs under adverse scenario
- Funding shock calibrated to H1 2012 stress:
  - Margin on less stable liabilities assumed to rise by maximum increase observed in H1 2012 = 52 basis points annually.
  - Funding shock applied over first two years of horizon.
- Funding costs decline in outer years as capital buffers rebuilt:
  - Improvement magnitude ≈ 38 bps in 2019 and 33 bps in 2020.

---

### Solvency test decomposition, heterogeneity, and sensitivities

### Aggregate solvency outcomes and drivers
- Baseline projections:
  - Aggregate CET1 stabilizes around 12.6 percent by 2020.
  - Aggregate Tier 1 leverage ratio ~4.8 percent over horizon.
- Adverse scenario outcomes:
  - Aggregate CET1 falls from 12.6 percent to 8.7 percent (low point in 2017).
  - Aggregate Tier 1 leverage ratio falls from 5.3 percent to 4.0 percent in 2017.
  - Aggregate losses GBP 35 billion in first two years.
  - RWAs peak in 2017 with 2.3 pps hike.
- Decomposition of CET1 shortfall at low point relative to end-2015:
  - 2.0 pps from loss provisions.
  - 1.9 pps from shifts to RWAs.
  - Aggregate impact of funding risk ≈ 1.4 pps (driven by add-on funding cost).

### Heterogeneity by bank type
- Major U.K. international banks vs domestic banks:
  - International banks: CET1 decrease by 4.3 pps at peak relative to December 2015.
  - Domestic banks: CET1 decrease by 2.6 pps.
- Drivers: larger overseas impairments, larger mark-to-market losses, higher CCR for internationally active banks.

### Sensitivity tests
- U.K. residential property price shock:
  - Adverse scenario peak-to-trough reduction 20 percent by 2018:Q3 reduces recovery rates by 3.5 pps.
  - Increase of 20 pps in LGD depresses aggregate CET1 ratio by 120 basis points.
  - Impact broadly linear on size of LGD shock.
- Sterling swap curve sensitivity:
  - Increase of 250 bps in risk-free rates would depress aggregate CET1 by about 120 basis points driven by decline in retained capital.
  - Test does not account for propagation from swap curve to other asset classes.

---

### Liquidity stress testing framework and outcomes

### LCR implementation and PRA transitional path
- LCR in effect from October 1, 2015.
- PRA revoked previous liquidity standards on that date; firms continue ILG reporting during interim period until at least October 2016.
- LCR transitional path:
  - 80 percent until end-2016,
  - 90 percent in 2017,
  - 100 percent in January 2018 as required by CRR.
- PRA’s transitional 80 percent starting point above CRR 70 percent due to PRA’s narrower eligible liquid assets.

### Liquidity stress test scenarios and assumptions
- Three LCR scenarios:
  - Standard Basel III LCR scenario.
  - U.K. retail stress (Northern Rock-like): run-off up to 15 percent for retail deposits; 60 percent for corporate deposits; 50 percent for operational deposits; contingent liabilities increases up to 50 percent.
  - U.K. wholesale stress (global financial crisis repeat): corporate deposit run-off up to 100 percent; operational deposits 75 percent; secured funding outflows backed by level-2B up to 100 percent; contingent funding increases up to 75 percent.
- Implied cash-flow tests:
  - Five-day and 30-day tests with severe assumptions (haircuts up to 60 percent for mobilizable securities; no new unsecured issuance; callback rates 100 percent on expected cash inflows related to credit extension without liquid collateral).
- Maturity mismatch test:
  - Calibrated on six maturity buckets (<1 week to >1 year); classification rules for no-maturity outflows and deposits specified.

### Liquidity stress test key results
- All firms passed LCR stress test under 80 percent hurdle and tailored U.K. retail and wholesale scenarios.
- Aggregate coverage ratio (percent) reported: 138 (standard), 97 (U.K. retail), 94 (U.K. wholesale) in reported table.
- Five-day implied cash-flow test:
  - Average counterbalancing capacity > 180 percent.
- 30-day implied cash-flow test:
  - Average counterbalancing capacity < 150 percent.
- Alternative scenario (no retail deposit run):
  - Counterbalancing capacity > 325 percent (five-day) and 250 percent (30-day).
- Maturity mismatch:
  - Five of ten firms show shortfall in one- to four-week bucket (retail-focused commercial banks and building societies).
- Single-currency analysis:
  - No U.K. firm shortfall in GBP.
  - One firm negative in EUR.
  - Two firms slight negative shortfall in USD.

---

### Model validation, comparison with BU, and limitations

### FSAP versus BoE differences (high level)
- Both tests: high granularity for international exposures; comparable impact on U.K. GDP (IMF 2.13 s.d.; BoE 2.04 s.d. over two-year cumulative real GDP growth); dynamic balance sheet assumptions with restrictions on deleveraging; traded risk scenario linked to macro scenario; same Basel III hurdle rate for RWAs and similar leverage hurdle.
- Key differences:
  - FSAP uses simple integrated TD model; BoE uses hybrid approach.
  - Scenarios differ: BoE long-lived U-shape shocks; FSAP speedier V-shape recovery.
  - Risk coverage differences: BoE includes misconduct costs and pension risk; FSAP excludes explicit misconduct stress but projects material impact via expense methodology.
  - Funding risk modeling differs; FSAP excludes management actions in stressed capital ratios while BoE reports before and after management actions.

### FSAP assessment of BU projections
- FSAP applied TD models to assess BU PD/LGD projections for selected IRB exposures (U.K. retail secured, corporate; non-U.K. corporate exposures in China, Hong Kong SAR, Ireland, U.S.).
- Comparison results:
  - FSAP broadly confirms banks’ BU calculations.
  - FSAP projections more cyclical and front-loaded; more conservative for cross-border exposures, relatively less conservative for U.K. exposures.
  - Differences arise from market-based risk-neutral measures versus banks’ physical PDs, granularity, and IRB regulatory calibrations.

### Caveats, data, and interpretation limits
- Results based on banks’ submitted data at cut-off date of December 2015; FSAP submitted data not validated by IMF.
- Data complemented by market-based and publicly available data for calibration.
- Cleaning, validation, matching, and reconciliation of risk data from multiple sources complex and constraining reliability.
- Structural shifts in banks’ business models since the crisis limit reliability of past data for forward-looking projections.
- Stress scenarios replicate historical events or extreme tail events; nonlinear effects can be captured but unknown patterns may emerge in extreme shocks.
- Specific caveat: uncertainties in projected bilateral exchange rates between GBP and USD over horizon since two major U.K. banks report regulatory balance sheets in USD.

*Italic: Source — IMF staff report content unit _cr16163 (extract provided).*

### EXECUTIVE SUMMARY ___________________________________________________________________________ 6

### _cr16163 - EXECUTIVE SUMMARY

### FSAP scope and approach
- FSAP conducted a comprehensive range of stress tests to assess resilience of the U.K. banking system.
- Top-down (TD) solvency stress test covered seven major U.K. banks and building societies representing over 80 percent of Prudential Regulatory Authority (PRA)-regulated banks’ lending to the real economy.
- Stress testing addressed solvency and liquidity risk and key risk types, including credit risk, market risk, sovereign risk, and funding risk.
- The FSAP stress scenario focused on a broad-based dislocation in financial markets, with sizeable jumps in yield curves and spillovers to vulnerable emerging market economies; scenario could be triggered inter alia by disorderly monetary normalization in the U.S.
- The FSAP solvency stress test complemented the 2015 concurrent stress test of the Bank of England (BoE), which covered the same set of banks but focused on a different global shock.
- The FSAP team also used its own models and analysis to cross-validate bottom-up (BU) submissions on selected portfolios for the 2015 BoE concurrent stress test.
- Suite of liquidity stress tests was conducted for 10 large firms, including three large U.K. foreign-headquartered investment banks.
- Cut-off date for the FSAP tests was December 2015; for comparison the 2015 BoE concurrent stress test had a cut-off date of December 2014.

### Key solvency stress test results
- Aggregate Core Equity Tier I (CET1) ratio:
  - Starts at 12.6 percent in December 2015.
  - Falls to a low point of 8.7 percent under the FSAP stress scenario.
- Leverage ratio:
  - Starts at 5.3 percent in December 2015.
  - Falls to a low point of 4.0 percent under the FSAP stress scenario.
- Aggregate losses:
  - Banks incur GBP 35 billion in losses in the first two years of the stress scenario.
  - All covered banks continue to meet prudential requirements throughout the stress test horizon.
- Risk-weighted assets (RWAs):
  - RWAs peak in 2017 with a 2.3 percentage points (pps) hike.
- Bank group heterogeneity:
  - Aggregate CET1 ratio of major U.K. international banks decreases by 4.3 pps at the peak of the FSAP stress scenario.
  - Aggregate CET1 ratio of major U.K. domestic banks decreases by 2.6 pps at the peak of the FSAP stress scenario.
  - International banks’ larger decline reflects stressed impairment charges in overseas exposures and larger mark-to-market losses in securities portfolios.

### Liquidity stress test results
- Tests covered 10 large financial institutions: the seven large U.K. firms in the solvency test plus three large U.K. subsidiaries of major foreign investment banks.
- All firms passed the liquidity coverage ratio (LCR) stress test under the current regulatory 80 percent hurdle rate for the standard Basel III scenario, as well as two tailored scenarios reflecting a U.K. retail deposit run and a U.K. wholesale event.
- All firms passed a five-day and a 30-day implied cash flow on an all-currency basis.
- Single-currency analysis by the PRA:
  - All firms had sufficient liquid buffers in domestic currency.
  - Only a minor shortfall in foreign currency.

### Model validation and comparison with BoE BU submissions
- A number of credit risk parameters estimated by the FSAP team are broadly comparable to those generated by participating banks under the BoE’s 2015 concurrent stress test, providing some degree of independent validation.
- FSAP applied an independent quantitative method to assess BU projections submitted by banks for the 2015 BoE concurrent stress test.
- Caution advised in comparing BU and TD results due to differences in data, granularity, and modeling approach.
- FSAP projections exhibit a somewhat more cyclical behavior than BU projections, but broadly confirm banks’ BU calculations subject to the caveats above.

### Conservative and modeling assumptions
- Key conservative assumptions used in FSAP stress tests:
  - Administrative expenses would not decrease in the stress scenario.
  - Economic hedges were not allowed to operate in the trading book.
  - Fee income was capped.
  - Funding-risk module assumed drying-up of market liquidity, with banks unable to pass-through increases in funding cost to loan rates.
- Some conservative choices were deliberate; others were driven by data constraints and the exigencies of a TD modeling approach.

### Data, interpretation, and caveats
- FSAP stress test results are based on data submitted by U.K. banks for FSAP purposes at the cut-off date of December 2015.
- The submitted data have not been subject to validation by the IMF.
- Data were complemented by market-based and publicly available data to support calibration.
- Cleaning, validation, matching, and reconciliation of risk data from multiple sources is complex and constrains reliability.
- Structural shifts in banks’ business models since the crisis limit the reliability of past data for forward-looking projections.
- Stress test scenarios typically replicate historical events or express extreme “tail events”; nonlinear effects can be captured but unknown patterns may emerge in extreme shocks.

### Financial system context and scale
- The banking sector is a dominant part of the U.K. financial system, accounting for over 60 percent of total financial sector assets.
- The U.K. financial system (sum of financial assets owned by banks and nonbank financial institutions) was about GBP 20 trillion in April 2015, over 10 times U.K. annual GDP.
- Excluding derivatives and cross-border exposures of foreign-owned bank branches:
  - Total assets of the U.K. financial system are GBP 13 trillion.
  - GBP 8 trillion of those assets are held by banks.

*Prepared by the International Monetary Fund (IMF) based on the FSAP technical note and associated analysis.*

### 5.      The U.K. is a global financial hub. As of December 2015, there were 359 monetary financial

### _cr16163 - 5.      The U.K. is a global financial hub. As of December 2015, there were 359 monetary financial

### Financial hub overview and size
- As of December 2015, there were 359 monetary financial institutions in the U.K., of which 239 firms were foreign-headquartered.
- U.K. banks’ consolidated assets amounted to about GBP 5.7 trillion.
- Foreign banks’ consolidated assets accounted for GBP 2.2 trillion, of which over GBP 1.7 trillion were foreign investment banks.
- Exposures to the U.K. economy:
  - U.K.-headquartered banks accounted for GBP 2.9 trillion of U.K. aggregate lending.
  - Branches of foreign-headquartered banks contributed GBP 1.1 trillion.
  - Subsidiaries of foreign-headquartered banks contributed about GBP 500 billion.

### U.K.-headquartered bank groups and asset distribution
- U.K.-headquartered banks categorized into three groups:
  - Major U.K. international banks (HSBC Holdings plc, Barclays plc, the Royal Bank of Scotland Group plc, and Standard Chartered plc) accounted for GBP 4.1 trillion of aggregate assets as of December 2015.
  - Major U.K. domestic banks (Lloyds Banking Group plc, Santander U.K. plc, and Nationwide Building Society) had an aggregate balance sheet of GBP 1.2 trillion.
  - All other U.K.-headquartered banks (including retail and investment banks, and building societies) are in the third group, with aggregate assets of under GBP 300 billion.
- Two banks in the covered group had partial government ownership: Lloyds Banking Group plc and the Royal Bank of Scotland Group plc.

### FSAP stress testing approach (overview)
- The FSAP stress test was conducted on a global consolidated group basis ("on a consolidated basis").
- Banking Reform Act of 2013 ring-fencing requirement (for banking groups with core deposits in excess of GBP 25 billion) will not be implemented until January 2019; thus the FSAP used consolidated basis for the stress test.
- FSAP stress testing components:
  - FSAP team performed its own TD solvency stress test based on data as of December 2015, matched with publicly available data for 2016–20, complementing the BoE/PRA concurrent stress test (based on end-2014 data for 2015–19).
  - Detailed stress test of the U.K. mortgage book using estimations of outstanding balances for different LTV vintages and a structural fair-value option Merton-based approach.
  - Single-factor sensitivity tests on a wide range of stressed U.K. residential property prices and a stressed swap curve.
  - Independent quantitative assessment of selected BU credit risk projections submitted by banks for the 2015 BoE concurrent stress test, focusing on a subset of domestic and cross-border retail and wholesale portfolios.
  - A battery of liquidity stress tests were performed by the BoE using scenarios and tools provided by the FSAP team; liquidity tests used end-December 2015 data for consistency.

### FSAP team solvency stress test — coverage and data
- Coverage:
  - Includes all major U.K. banks, representing over 80 percent of PRA-regulated banks’ lending to the real economy.
  - All major U.K. banks with total retail deposits equal to or greater than GBP 50 billion as of December 2015 are included.
  - The same set of banks as the 2015 BoE concurrent stress test: Barclays plc, HSBC Holdings plc, Lloyds Banking Group plc, Nationwide Building Society, the Royal Bank of Scotland Group plc, Santander UK plc, and Standard Chartered plc.
  - The aggregate size of banks covered represents about 75 percent of total banking sector assets.
- Data and sources:
  - Stress test based on granular bank-submitted data via FSAP templates: IRB portfolio breakdowns by Basel asset class, sectoral split, currency, geography; securities portfolio at security level; duration of debt portfolio; open positions by market risk factor; RWAs split by STA and IMM; projections of RWAs; maturity gaps for interest rate risk.
  - Public data matched to banks’ data: Pillar 3 disclosures, Bankscope, SNL, Bloomberg, Datastream, Markit, 2015 EU-wide transparency exercise.
  - Aggregate sources: Haver Analytics, Moody’s KMV, International Financial Statistics (IFS), World Economic Outlook (WEO).

### Risk coverage and methodologies
- Hurdle rates and regulatory framework:
  - Assessment criteria (“hurdle rates”) include Basel III capital standards implemented via the Capital Requirements Regulation (CRR) using PRA national discretions, and the PRA leverage framework.
  - PRA implemented an end-point definition of CET1 that does not take into account transitional provisions for CET1.
  - All additional Tier 1 instruments issued externally by U.K. banks have a trigger of 7 percent of CET1 rather than the CRR minimum of 5.125 percent.
  - Under the baseline scenario, the hurdle rate applied was set at 7 percent CET1 ratio.
  - Under the adverse scenario, the hurdle rate was set at 4.5 percent CET1 ratio and 6 percent Tier 1 ratio.
  - The hurdle rate does not include G-SIBs’ higher loss-absorbency requirements that began to be phased in January 1, 2016, with full implementation by January 1, 2019.
  - The hurdle rate also includes a 3 percent Tier 1 ratio based on the U.K. leverage framework.
  - For the IMF FSAP stress test, the leverage ratio is calculated using quarter-end data as opposed to average over the quarter.
- Credit risk:
  - For IRB exposures, a separate credit risk analysis was calibrated for five Basel asset classes and the 15 most material geographies, yielding a matrix of 75 credit risk models for IRB exposures.
  - Basel asset classes covered: retail unsecured exposures (including SME credit); retail secured exposures (mortgages); corporate lending (including commercial real estate); lending to institutions; sovereign and central bank lending.
  - Material geography composition for IRB exposures: United Kingdom (60 percent); other advanced economies (30 percent) including Canada, France, Germany, Hong Kong SAR, Ireland, Italy, Korea, Netherlands, Singapore, and the United States; emerging economies (10 percent) including China, India, and South Africa.
  - Credit stress test included securitization and covered bond exposures.
- Market risk:
  - Scope included all positions exposed to changes in market prices: Held for Trading (HFT), Available for Sale (AFS) and fair-value option (FVO), including sovereign and non-sovereign exposures.
  - Exclusions: positions in hedge accounting portfolios, economic hedges, and cash flow hedges.
  - Treatment of sovereign exposures in the banking book follows the Basel III framework; FSAP used the IRB approach relying on FSAP risk assessments of underlying issuers.
- Funding risk:
  - Funding risk analysis examined shocks from: (i) a system-wide event related to monetary policy shocks and adverse movements in LIBOR rates; (ii) an idiosyncratic event linked to concerns over the solvency position of each bank; and (iii) a contagion effect triggered by concerns over the solvency position of vulnerable banks within the U.K. banking system.
  - FSAP funding risk module models contagion from peer banks’ funding pressures and limits pass-through of higher funding costs on bank lending rates; the BoE methodology allows more pass-through where banks judge feasible.
- Differences with BoE concurrent stress test (high level):
  - Both tests: high granularity for international exposures; comparable impact on U.K. GDP (equivalent to a 2.1 standard deviation shock on two-year cumulative real GDP growth during first two years); dynamic balance sheet assumptions with restrictions on deleveraging; traded risk scenario linked to macro scenario; same Basel III hurdle rate for risk-weight capital metric and similar leverage hurdle rate.
  - Key differences: FSAP uses a simple integrated TD model versus BoE’s hybrid approach; scenarios differ (BoE: long-lived U-shape shocks; FSAP: speedier V-shape recovery); risk coverage differences (BoE includes misconduct costs and pension risk; FSAP excludes misconduct stress but projects a material impact from misconduct via expense methodology); traded risk scope and funding risk modeling differ; FSAP excludes management actions when projecting stressed capital ratios while BoE reports before and after management actions.

### FSAP scenarios and macro assumptions
- Two macroeconomic scenarios examined: baseline and adverse.
- Baseline scenario:
  - Draws from the October 2015 WEO projections for key variables, expanded to additional variables relevant to credit risk (residential and commercial real estate prices for the U.K., the U.S., euro area core, and euro area periphery; U.K. Bank rate; U.K. credit growth; equity prices for the U.K. and the U.S.).
  - Variables projected by adjusting BoE’s 2015 concurrent stress test baseline to WEO core variables and spanning horizon through 2020.
  - Country-specific variables complemented by IMF Global Assumptions (Gas) database, including six-month LIBOR by major currency and commodity price forecasts.
- Adverse scenario (features and assumptions):
  - Explores: sharp downturn in emerging markets; sizeable increase in rates and steepening of yield curve in the U.K. and globally, triggering abrupt price corrections; funding risk from rising LIBOR spreads, dry-up of money market issuance, and FX swap market disruptions; large correction in U.K. property markets.
  - More specifically assumes:
    - Balance sheet recession in the United Kingdom and financial crises in fragile emerging economies.
    - Accelerated monetary normalization in the United States with a 200 basis points policy rate hike during 2016–17.
    - Drying up of liquidity in money markets and steepening of the yield curve driven by monetary policy uncertainty, internationally correlated credit risk premium shocks, and duration premium shocks.
    - A stock market correction triggered by equity risk premium shocks that spreads to FX markets driven by currency risk premium shocks.
    - Sudden stops in vulnerable emerging economies causing corrections in equity markets and sharp currency depreciations.
    - Amplification in the United Kingdom from an autonomous domestic demand shock, confidence loss in property markets, stress in funding markets, and decompression of the term premium in debt markets.
  - U.K. GDP path under adverse scenario:
    - U.K. GDP contracts by 1.6 percent in 2016.
    - U.K. GDP reaches a peak deviation from baseline levels in 2017 at -7.5 percent.
- Scenario focus differences with BoE:
  - BoE 2015 scenario focused on synchronized global downturn and correction affecting mainly Asia and the euro area.
  - FSAP scenario features disorderly monetary normalization in the United States with broad-based market dislocation and spillovers to vulnerable emerging market economies.
  - Both scenarios have major impact on the U.K. economy and share a similar degree of severity as measured by effect on U.K. real GDP (the IMF scenario constitutes a 2.13 standard deviation move in two-year cumulative real GDP growth rate over the first two years; the BoE constitutes a 2.04 standard deviation move).

*Source: IMF FSAP material as presented in the supplied content.*

### 16.      The persistence of shocks and the channels of distress are also different. The FSAP

### _cr16163 - 16.      The persistence of shocks and the channels of distress are also different. The FSAP

### FSAP versus BoE adverse scenarios
- The FSAP scenario incorporates a quicker recovery toward steady state and features a severe dislocation in money markets and a steepening in the yield curve.
- Scenarios are designed to capture different risks and assess the U.K. banking system’s resilience against a range of possible stresses.

### Credit risk and regulatory capital overview
- Average risk weights for major U.K. banks:
  - Declined from 70 percent in the mid-1990s to a trough of 29 percent in 2008:Q3.
  - Reached 36.3 percent in December 2015.
- At end-2015, RWAs of the largest seven U.K. banking groups reached GBP 1.9 trillion, of which:
  - 78 percent reflects credit risk excluding counterparty credit risk (CRR).
  - Market risk represented just over 5 percent.
  - CCR amounted to over 7 percent.
  - Operational risk was under 10 percent.
- Shortcomings in RWAs are typically compensated under the Pillar 2 capital framework:
  - Additional minimum requirements (Pillar 2A) average about 2.4 percent of RWAs in terms of Tier 1 across major U.K. banks as of end-2015.

### Booking approaches and exposure composition
- Aggregate credit risk EaD reported in 2014 reached GBP 4.4 trillion for major U.K. banks:
  - GBP 3.3 trillion booked under the IRB approach.
  - GBP 1.1 trillion under the STA approach.
  - Of the GBP 1.1 trillion STA, GBP 400 billion was exposures to sovereign and central banks, with an asset-weighted average risk weight of under 5 percent.
- For IRB exposures, credit risk depends on:
  - Exposure at default (EaD), probability of default (PD), and loss-given default (LGD).
- All new or materially changed IRB capital models require PRA approval.
- For STA exposures, risk depends on banks’ loan classification and provisioning requirements set out by the U.K. authorities.

### Credit risk model used in the FSAP
- Expected losses (impact of stress on regulatory capital through increased provisions) were computed as the level of expected losses: EAD * LGD * PD (notation j,i,t for asset class, bank, time).
- Separate credit risk models were estimated by Basel asset class and geography; all material geographies for U.K. banks are covered.
- For FSAP purposes, unexpected losses for specialized lending (subject to slotting under PRA) were treated under the corporate IRB approach.

### Approach to project PDs (two-step process)
- Step 1: Build time series of PD proxies by asset class and geography using:
  - Market-based PDs across all geographies.
  - Write-off data for U.K. exposures (quarterly frequency).
- Step 2: Econometric forecast of bank-specific PDs using banks’ Pillar 3 disclosures as the dependent variable and projected PD proxies as the main driver.
- Data limitations noted:
  - Pillar 3 disclosures available at annual frequency, short history starting in 2008, and do not cover all portfolios and geographies.

### PD proxies and data sources
- Two PD proxies applied for U.K. exposures:
  - Moody’s Analytics average one-year expected default frequency (EDF) mapped to corporate group, financials group, and consumer nondurables and services group (mapped to corporate exposures, exposures to institutions, and retail unsecured exposures, respectively).
  - BoE write-off rates on U.K. exposures by main asset class (closest to estimated portfolio loss ratio but tends to lag defaults by up to about 24 months).
- For sovereign and central bank exposures, PD proxy extracted from sovereign yields.
- For overseas exposures, a market-based PD proxy was used (Moody’s Analytics one-year EDF) for Canada, China, France, Germany, Hong Kong SAR, India, Ireland, Italy, Korea, the Netherlands, Singapore, South Africa, Spain, and the U.S.
- Note: For household secured loans, market-based proxies for default are not available.

### Econometric and statistical approach
- An integrated approach used bilateral and multivariate VAR estimation techniques and nonlinear principal component analysis (PCA) to project conditional PDs.
- Global conditions and spillovers from relevant geographies were included as factors.
- Scale of estimation:
  - A battery of over 900 credit risk specifications was run.
  - Over 700 bilateral VARs for all pairs of credit risk by PD proxy, asset class, and geography against each macro variable forecast.
  - 150 PCA analyses for each factor, PD proxy, asset class, and geography.
  - 50 multivariate VARs for each PD proxy, asset class, and geography.

### Key model findings and projections
- Drivers of credit risk for U.K. exposures:
  - Financial conditions are the main driver. Rising spreads across money and equity markets (PC1_fin) and tight corporate credit markets (PC1_fin_c) contribute to rising corporate EDFs.
  - Projected peak for corporate EDF: 5.4 percent in 2017:Q2.
  - Write-off rates for U.K. mortgages increase with deteriorating financial conditions in secured lending markets (PC1_fin_m) and with deteriorating macroeconomic conditions; movements in housing prices are not statistically significant.
  - Projected peak for write-off rates for U.K. mortgages: 0.08 percent in 2017:Q1, twofold baseline rate levels.
- Determinants of credit risk in overseas exposures vary by geography:
  - France: domestic macroeconomic conditions and real estate prices are drivers of corporate credit risk (noted that some are not statistically significant at the 5 percent level).
  - China: global financial conditions—particularly FX shocks, steepening of the yield curve, and developments in oil markets—are the most significant drivers of corporate credit risk.
- Model fit statistics reported (examples):
  - R-squared values reported across specifications include 0.874, 0.925, 0.976, 0.692, 0.929 for various equations; adjusted R-squared values include 0.859, 0.916, 0.973, 0.653, 0.920.

### Modeling and projection outputs highlighted
- Projections illustrated across multiple variables (GDP Growth, Inflation, Unemployment, Bank Rate, Exchange Rate, Domestic Credit, Equity Price, 3-month Sterling LIBOR, UK Government Bond Slope) comparing IMF and BoE adverse scenarios over quarterly horizons (selected ranges shown in figures are 2016Q1 through 2020Q4 and 2008Q1 through 2019Q2 for historical comparisons).
- Aggregate PD proxy projections for selected U.K. exposures used principal components:
  - PC1_Macro: deteriorating macroeconomic environment characterized by low growth, high inflation, and high unemployment.
  - PC1_Fin: financial turmoil featuring rising LIBOR rates, GBP depreciation, and a U.K. stock market correction.

*International Monetary Fund: United Kingdom FSAP—credit risk model description and key projections as presented in the supplied content.*

### 6.4 percent...

### 6.4 percent...

### PD estimation, adjustment, and use in stress test
- An econometric approach drawing on banks’ Pillar 3 disclosures was used to avoid a structural shift at the cut-off date between banks’ estimated PDs and FSAP team’s PD projections.
- HSBC Holdings plc Pillar 3 disclosures were used as an extract (HSBC accounts for one-third of U.K. major banks’ assets).
- Peak bank-estimated PDs (from Pillar 3) at the height of the financial crisis:
  - European unsecured retail exposures: 4 percent
  - U.S. mortgage portfolio: 15 percent
  - European mortgage exposures: 3 percent
  - Global corporate exposures: 3 percent
  - Exposures to institutions (mainly financial): 0.5 percent
  - Sovereign exposures average PD peak: 0.2 percent in 2008
- Two key reasons PD proxies and bank-specific PDs differ:
  - PD proxies are for the aggregate market portfolio and may not reflect banks’ portfolio risk profiles.
  - Forecast PD proxies are market-based (Merton approach) or approximate ex post realization of credit losses, while bank-estimated PDs are ex ante IRB-model measures.
- For consistency with the PRA regime, bank-specific PDs by asset class and geography were regressed on historical EDFs and write-off rates; estimated coefficients were used to forecast bank-specific PDs over the stress-test horizon.
- Adjustment for PiT cyclicality: To compute expected losses, an adjustment to bank-specific PD projections is performed to account for the cyclicality of PiT parameters under the adverse scenario.
- Formula used to link bank blended PDs and aggregate PD shifts:
  - j t i PD , = j t PDΔ + PDPD j j t i −1 ,, (presented in source as: jjj ttiti PDPDPDΔ + = −1,,)

### LGD calculation methodology for U.K. mortgages
- FSAP applied a fair-value option approach: LGD is highly nonlinear in house price at default VT and original LTVt.
- LGD projections depend on four key parameters:
  - The distribution of original LTV ratios by vintage;
  - The outstanding value of each loan vintage net of amortization;
  - The house price fall assumed under the scenario;
  - The forced sales discount on property market price under foreclosure.
- Empirical changes in origination LTV distribution:
  - At end-2015, 0.1 percent of regulated mortgages have LTV ratios over 95 percent (relative to over 6 percent in 2007).
  - High-rated mortgages with LTV <= 75 percent account for over 60 percent of originated mortgages at end-2015 (about 50 percent in 2007).
- Rollover and maturity:
  - Principal rollover rate has slowed since the crisis; estimated initial maturity lengthened (prevalence of five-year fixed-rate mortgages).
- House price path and scenario:
  - Path draws on the 2015 BoE scenario adjusted for IMF scenario horizon.
  - Largest yoy decline in house prices at end-2016; peak-to-trough decline reaches 20 percent by 2018:Q3.
- Forced sales discount:
  - A conservative forced sales discount set at 30 percent of fair-value U.K. residential property prices (derived from foreclosure sale comparisons and supported by international evidence).
- Under the adverse scenario, average LGD shock set at 3.5 percent.
  - LGD by vintage pattern is non-monotonic: higher LGD for pre-crisis vintages (driven by high LTV tails) and recent vintages (driven by higher fair-market valuations at origination).
  - Multiplying vintage LGD by outstanding volume yields the average LGD shock of 3.5 percent.
  - Estimated average initial maturity for pre-crisis loans ranges between five and seven years, yielding no material outstanding volume prior to 2008:Q3.

### LGD calibration and cross-checks for mortgages and non-mortgage portfolios
- Robustness checks:
  - Maximum one-year increase in IRB LGDs for U.K. mortgages in the post-crisis period is about 3 pps (HSBC Pillar 3 disclosures).
  - Write-off rate peaks for secured lending to individuals: around 0.14 percent in 1997:Q3 and 2009:Q1.
  - Using an estimated PD for U.K. mortgages of 1 percent, a 0.14 percent write-off rate implies LGD of 14 percent, approximating FSAP average stressed LGD over horizon.
- For non-mortgage loans:
  - Adverse LGDs calibrated for the first two years of the scenario; recovery path over the outer three years.
  - Bank-by-bank LGD projections derived using initial bank-specific LGD ratios and aggregate LGD shifts per portfolio.
  - Formula linking post-credit risk mitigation LGD to aggregate LGD shift:
    - j t i LGD , = j t LGDΔ + LGD j j t i −1 ,, (presented in source as: ttiti LGDLGDLGDΔ + = −1,,)
- Sources used for LGD shocks:
  - Banks’ disclosed IRB-based LGD estimates during stressed periods (e.g., average peak LGD in HSBC wholesale IRB corporate portfolio reached 39.2 percent in 2011, an increase of about 3.6 pps over 2015 levels).
  - LGD floors applied to central government and central bank exposures in 2013 and to institutions in 2014; floors maintained under adverse scenario.
  - Projected increases in corporate LGDs cross-checked against nominal GDP drops under adverse conditions.
  - Retail unsecured LGD increases matched with rising unemployment by IRB jurisdiction (including a rise of over 2 pps in the U.K. unemployment rate in 2017).

### EAD projection, FX structural risk, and credit-line triggers
- EAD projection drivers:
  - Balance sheet assumptions and structural FX risk in foreign currency loans;
  - Triggered credit lines and guarantees.
- Components and notation (from source):
  - i denotes the bank, j denotes the asset class, t is time;
  - t i g , is the growth rate of the IRB portfolio;
  - j FX i f , is the fraction of foreign currency loans;
  - t FX Δ is the shock to foreign currency under the macro scenario;
  - ( j t i PD 1, 1 − − ) represents the non-defaulted portfolio;
  - j t i L , Δ is the shock to triggered credit lines and guarantees;
  - j t i UCL 1, − is the amount of undrawn guarantees.
- Currency considerations:
  - FSAP considers two major currency pairs by bank: EUR and USD for GBP-reporting banks; EUR and GBP for USD-reporting banks.
- Balance-sheet and credit supply assumptions:
  - FSAP scenario assumes constrained balance sheets by geography; bank credit supply behavior is constrained through activation of macroprudential policy tools to avoid credit supply rationing.
  - Credit growth is determined by credit demand shocks from negative consumption and investment shocks under the adverse scenario.
  - Growth rate of credit varies across banks; computed as EAD-weighted nominal GDP growth across relevant jurisdictions.
- FX and off-balance-sheet calibration:
  - Data on FX structural effects from foreign currency loans provided by firms; where gaps exist, loans assumed denominated in domestic currency of the overseas exposure.
  - Shock to triggered credit lines and guarantees calibrated using maximum increase in off-balance-sheet exposures to EAD ratio for corporates reported in Pillar 3 over 2008–15, about 40 percent.

*Source: IMF staff estimates; HSBC Holdings plc Pillar 3 disclosures; Mortgage Lenders and Administrators Return (MLAR) statistics; BoE scenario and IMF staff calculations.*

### 42.      To compute capital requirements, regulatory risk parameters were considered and the

### _cr16163 - 42.      To compute capital requirements, regulatory risk parameters were considered and the

### Capital requirements methodology
- Basel III formula for IRB exposures was applied.
- RWAs derivation depends on estimates of PD, LGD, EAD, correlation assumptions, and effective maturity for each exposure.
- Credit RWAs were computed after applying a scaling factor of 1.06.
- A multiplier of 1.25 was applied to the correlation parameter of:
  - all exposures to large regulated financial institutions, and
  - all unregulated financial institutions.

### Credit Risk Model for STA Exposures — main assumptions and calibration
- Expected losses for STA banking book exposures rise due to:
  - migrations from performing to nonperforming loans (NPLs), and
  - migration effects across rating grades within NPLs.
- NPL ratio caveat: NPL ratios are not disaggregated between IRB and STA exposures.
- For large international banks, NPLs can be larger among STA exposures where the STA approach is used to book exposures warehoused in overseas subsidiaries subject to higher credit risk.
- NPL coverage calibration: Basel findings on NPL coverage ratios during the global financial crisis averaged about 65 percent.

### Two-prong approach to compute impact of migration on capital requirements
- Prong 1 — NPL risk-weight increase:
  - Risk weight of NPL was set at an average 130 percent.
  - The increase in risk weights attributed to STA exposures in default = (130 percent − average risk weight of each bank’s STA exposures (excluding sovereign and central bank exposures)) × nominal amount of NPLs under the scenario.
- Prong 2 — downgrade of non-defaulting portfolio:
  - The non-defaulting portfolio was assumed to downgrade two notches under the adverse scenario.
  - Asset-weighted risk weight of major U.K. banks’ STA exposures (net of sovereign and central bank claims) stands at about 75 percent.
  - A two-notch downgrade in the BBB+ to BB- category represents an increase of 16.67 percent in the average risk weight of the non-defaulted portfolio.
- Table 3 (selected figures; in millions of British Pounds and percent) — United Kingdom: EAD and Risk Weights of STA Exposures:
  - IRB EAD: 1,037,673; 449,608; 478,572; 384,837; 580,080; 199,100; 193,889
  - STA EAD: 378,332; 207,328; 193,869; 66,160; 119,956; 64,600; 23,911
    - of which gov&cb: 121,284; 104,499; 63,596; 11,088; 83,617; ...; 14,423
    - of which other: 164,690; 102,829; 130,273; 35,285; 36,339; 64,600; 9,488
  - Total EAD: 907,230; 656,936; 672,441; 288,953; 700,036; 263,700; 217,800
  - STA (in percent of total EAD): 41.7%; 31.6%; 28.8%; 22.9%; 17.1%; 24.5%; 11.0%
  - RWA STA total: 228,665; 74,540; 101,496; 38,874; 25,444; 37,500; 3,621
    - of which other: 138,418; 71,712; 101,412; 24,151; 25,433; 37,500; 3,621
  - RW density gov&cb: 10.4%; 2.7%; 0.1%; 10.6%; 0.0%; ...; 0.0%
  - RW density excl gov&cb: 84.0%; 69.7%; 77.8%; 68.4%; 70.0%; 40.0%; 38.2%
  - ∆ RWAs due to ∆ NPL: 46.0%; 60.3%; 52.2%; 61.6%; 60.0%; 90.0%; 91.8%
  - ∆ RWAs due to rating effects on PL: 16.7%; 16.7%; 16.7%; 16.7%; 16.7%; 16.7%; 16.7%

### Drivers of STA capital requirements (decomposed)
- Capital requirements driven by:
  - changes in provisioning rates,
  - growth of EAD,
  - structural FX risk,
  - triggered credit lines and guarantees,
  - migration effects.
- RWAs in STA portfolio composed of four main components (described in equation form in source):
  - motion from provisions, portfolio growth, and FX effects;
  - increase in risk weights from triggered off-balance sheet credit lines and guarantees;
  - increase in risk density from migration of loans from performing to nonperforming;
  - change in risk density from the transition matrix estimated for performing exposures.
- Notation preserved in source: ti, PrΔ; ti g; EURi f, (USDi f,); EUR FXΔ (USD FXΔ); STA ti RWD1,-; OBS ti RWD1,-; OBSSTA ti L,Δ; NPL ti RWD1,-; PL ti RWD1,-; ti NPL,Δ; sti RWD1,-; b ti RWD1,-.

### Market risk approach
- Market liquidity context:
  - Market liquidity is not low in most markets when compared to historical averages, using measures such as bid–ask spreads, imputed round-trip costs, and Corwin and Schultz’s high-low spreads.
  - Current levels supported by benign cyclical conditions including accommodative monetary policy; resilience challenged by structural changes (asset manager concentration, banks’ retrenchment from trading, proliferation of smaller or higher-risk issues).
- Scope of market risk analysis:
  - Shocks to sovereign debt securities in 22 jurisdictions.
  - Market stress in 4 non-sovereign corporate bond indices: EU investment grade, EU high-yield, U.S. investment grade, and U.S. high-yield.
  - Banks’ corporate fixed income securities are mapped into these corporate indices.
- Stress shock timing and treatment:
  - Shocks hit banks throughout the five-year stress test horizon; rebalancing of the portfolio is disallowed.
  - Market shock applied as an instantaneous shock to all positions covered by the market risk analysis each year of the horizon, with losses fully recognized each year.
- Notional values growth under constrained balance sheet: notional values grow according to bank i’s growth of interest-bearing assets (ti g) and FX fractions (EURi f, USDi f) with FX shocks (EUR FXΔ, USD FXΔ); no portfolio rebalancing or liquidation allowed.
- Calibration of market risk factors:
  - Baseline sovereign yield to maturity (YTM) changes extracted from WEO projections for 10-year bond yields and adjusted using Bloomberg generic bond rates.
  - Adverse scenario shocks calibrated using the IMF in-house structural model; adverse shifts stem from monetary policy shocks and internationally correlated term premium shocks.
  - Model assumes no flight-to-quality effects; all sovereign yields edge up relative to December 2015 rates.
- Largest year-on-year hikes in yields under the adverse scenario by 2017 (as reported):
  - United States: 310 bps
  - Ireland: 205 bps
  - Hong Kong SAR: 192 bps
  - U.K.: 168 bps
- Corporate spread shocks:
  - Under the adverse scenario, U.K. corporate spreads reach levels seen at the height of the financial crisis with spreads rising between 353 basis points (for U.S. investment grade) and 1,463 basis points (for U.K. high yield) by 2016.
- Treatment of securities and hedges:
  - Full revaluation of securities for HFT, AFS, and FVO.
  - Granular approach applied to all firms, including those whose market risk capital requirement is lower than 5 percent of the total capital requirement.
  - Banks not allowed to rebalance; mitigating effect of hedge accounting portfolios excluded.
  - Offsetting impact from derivatives classified as economic hedges excluded.
  - Prudential filters excluded; positions with prudential filters were nevertheless included.
- Distribution of debt securities portfolio (as of December 2015):
  - AFS: 67 percent of aggregate debt securities portfolio for U.K. major banks.
  - HFT securities: 26 percent of the portfolio.
  - HTM securities: 7 percent.
  - Relative to bank total assets:
    - AFS portfolio = 9.7 percent of total assets.
    - Trading book = 3.8 percent of assets.
    - HTM securities = 1.0 percent of assets.
  - By asset class:
    - U.K. sovereign securities = 2.6 percent of assets.
    - Other sovereign securities = 8.5 percent of assets.
    - Corporate securities = 3.4 percent of assets.

### Market and other risk-factor shocks (selected)
- Table 4 corporate spread projections (selected rows shown in source):
  - UK IG: Dec-15 3.26; 2016 3.48; 2017 3.63; 2018 3.78; 2019 3.92; 2020 4.04; adverse path includes 7.80; 6.05; 4.59; 4.33; 4.32 (years shown in full table).
  - UK HY: Dec-15 6.15; 2016 6.75; 2017 7.21; 2018 7.67; 2019 8.11; 2020 8.56; adverse includes 20.78; 12.86; 8.48; 8.53; 8.83.
  - US IG: Dec-15 3.50; 2016 4.22; 2017 4.94; 2018 5.56; 2019 6.12; 2020 6.37; adverse includes 7.03; 7.13; 6.56; 6.49; 6.44.
  - US HY: Dec-15 8.33; 2016 9.39; 2017 10.21; 2018 10.94; 2019 11.60; 2020 11.95; adverse includes 17.97; 14.33; 11.94; 11.97; 12.02.
- Shocks to other risk factors (selected excerpts from Table 5; in percent):
  - FX shocks (negative denotes appreciation of USD relative to FX): examples
    - Hong Kong HKD: 2016 baseline -0.09; 2017 baseline 0.56; 2016 adverse -0.43; 2017 adverse 0.19
    - Canada CAD: 2016 baseline 0.49; 2017 baseline 0.74; 2016 adverse -7.04; 2017 adverse -7.33
    - UK GBP: 2016 baseline 1.17; 2017 baseline 1.65; 2016 adverse -7.64; 2017 adverse -6.57
  - Shock to U.K. risk-free interest rate:
    - baseline: 2016 0.33; 2017 0.25; 2018 0.16; 2019 0.10; 2020 0.14
    - adverse: 2016 0.33; 2017 -1.16; 2018 -0.22; 2019 0.99; 2020 0.95
  - Shock to Gold:
    - baseline: 2016 15.55; 2017 11.52; 2018 6.76; 2019 2.75; 2020 0.33
    - adverse: 2016 -10.43; 2017 -20.54; 2018 -12.56; 2019 -1.93; 2020 1.33
  - Shock to Oil: adverse plunge by 2017 (source states "commodity prices for metals and fuel plunging by 2017").

### Market risk modeling and RWA calculation
- Impact of traded risk stress test on P&L differentiates between:
  - general interest rate impact, and
  - credit spread impact.
- Haircut calculation: maximum modified duration for each residual maturity × change in yield to maturity at stress-test levels.
- Shocks to market yields derive from:
  - general interest rate impact assumed under the scenario, and
  - credit spread impact for each individual security.
- Calculation of RWAs for market risk:
  - Drew on firms’ projections for the 2015 BoE stress test, adjusted for changes in the IMF market scenario relative to the BoE traded risk scenario.
  - RWAs projections for market risk include VaR, Stressed Value at Risk (sVaR), incremental risk charge (IRC), comprehensive risk measure (CRM), structural FX risk, CVA, PVA, and CCR.

### P&L approach and profitability context
- Structural factors:
  - Major U.K. banks have continued to deleverage; profits of major U.K. banks have been notably lower than pre-crisis.
  - Part of pre-crisis profitability driven by excessive risk-taking and was unsustainable.
  - Banks have been exiting lower-return businesses and non-core foreign operations, reducing exposures to other financial institutions and interconnectedness relative to 2008.
  - Accumulated charges relating to past misconduct have weighed on profitability.
- Cyclical factors:
  - Sluggish economic recovery and sustained period of low interest rates pressured profitability.
  - Low policy rates’ impact on net interest margins (NIMs) has not been material, partly due to hedging.
  - Market-based indicators show below-par price-to-book (P/B) ratios.
    - Asset-weighted average P/B for major U.K. banks at end-2015 stood at 0.8 (from a peak of 1.9 in July 2007).
    - Five-year CDS stood at 69 basis points.

*Italic: Source — IMF staff report content unit _cr16163 (extract provided).*

### 2007. Average post-tax return on equity (ROE) fell from 17 percent in 2006 to 6 percent in March 2015 (BoE’s

### _cr16163 - 2007. Average post-tax return on equity (ROE) fell from 17 percent in 2006 to 6 percent in March 2015 (BoE’s

### Banking profitability, divestments, and market indicators
- Average post-tax return on equity (ROE) fell from 17 percent in 2006 to 6 percent in March 2015.
- Notable divestments and reductions in geographic footprint:
  - Barclays plc: reduced exposure to periphery Europe; announced on March 1, 2016, plans to divest part of its shareholding in Barclays Africa over the next two to three years.
  - Royal Bank of Scotland Group plc: will reduce its geographic footprint from 38 to 13 countries.
  - Lloyds Banking Group plc: reduced its international presence from 30 countries in 2011 to seven countries.
- Sample of U.K. banks in market-based indicators includes HSBC Holdings plc, Barclays plc, the Royal Bank of Scotland Group plc, Lloyds Banking Group plc, and Standard Chartered plc.

### Net interest income (NII) and net interest margins (NIMs): drivers and projection approach
- Banks relying on net interest income are particularly exposed to low NIMs; adverse macroeconomic conditions and money market stress were used to project NII.
- Key channels reducing profitability under stress:
  - Limited pass-through of funding costs to borrowers when borrower creditworthiness deteriorates.
  - Shrinking income base as performing loans migrate into exposures at default.
  - Decline in asset prices and rise in loan delinquencies increasing charge-offs and loan-loss provisions.
  - Potential spike in bank funding costs, with banks having unstable funding structures particularly exposed to credit-sensitive investors.
- FSAP approach to net interest income:
  - Funding costs proxied by changes to deposit rates determined by policy rates, LIBOR rates, and firm-specific conditions (asset quality, capital adequacy, contagion). An additional add-on funding cost calibrated to funding stress observed during the Eurozone crisis in H1 2012 was included.
  - NIMs projected based on the policy rate, money market rates, and the term structure assumed under the macroeconomic scenario; yield curve steepening (flattening) typically raises (lowers) NIMs, though crisis-period term premium spikes may weaken pass-through.
  - Interest income on trading assets and liquid buffers projected by regulatory book.
  - Short-term interest rate risk in the banking book captured using a gap risk approach; bank hedges excluded for stress test calculations.

### Deposit rates: projection formula and parameters
- Deposit rates projected at the bank level; changes to deposit rates are driven by:
  - Changes to LIBOR 6m rates, with elasticity of 0.5.
  - Bank aggregate capital adequacy reflecting solvency–funding linkages.
- Regression period for deposit-rate formulation: 1999 Q1 through 2015 Q4. Reported R2: 0.63. Policy rate not statistically significant in regression.

### NIM projection details and parameters
- NIMs regressed on money market rates and the slope of the yield curve.
- Bank-level projection anchored on reported NIM in December 2015; weighted average NIM of U.K. major banks was around 2.8 percent in December 2015.
- Changes to NIMs mainly explained by:
  - Changes to LIBOR 6m rates with elasticity 0.2.
  - Slope of the yield curve with elasticity -0.1.
- Regression period for NIM projection: 1999:Q1 through 2015:Q4. Reported R2: 0.4. Note on unexpected negative slope sign explained by limited pass-through during the global financial crisis.

### Lending rates and nominal shocks (summary)
- Lending rates driven by shocks to funding costs and shocks to NIMs.
- Under the adverse scenario:
  - First year: lending rates outpace baseline due to rise in funding costs and increase in NIMs.
  - Second year: NIMs compress as borrowers’ creditworthiness downgrades (reflected in yield-curve slope changes).
- Interest accrued on defaulted loans is not recognized under the baseline and adverse scenario (contrast with 2016 EU-wide stress methodology).

- Table 6: Nominal shocks to rates (deviations from end-2015 value) — selected entries shown in source:
  - ∆D t-2015 (Baseline): 14 16 13 12 8
  - ∆D t-2015 (Adverse): 5 11 7 4 2 5 1 4 4
  - ∆NIM t-2015 (Baseline): 3 4 3 2 0
  - ∆NIM t-2015 (Adverse): 1 5 -1 1 3 1 7 1 3
  - ∆L t-2015 (Baseline): 17 20 16 14 8
  - ∆L t-2015 (Adverse): 6 7 16 5 5 6 8 5 8

(Note: Table 6 in the source presents nominal shocks across 2016–2020; above are the numeric entries as reported.)

### Interest income from securities and banking book interest-rate risk
- Implicit interest rate computed for HFT, AFS, and HTM regulatory books by asset class: own sovereign bonds, foreign sovereign bonds, corporate bonds, and other securities.
- Implicit interest rate projected forward assuming no changes in portfolio composition during the stress test period (conservative assumption).
- Interest rate in the banking book driven by policy rate shocks and banks’ repricing gaps across six maturity buckets: <1m, 1m-2m, 2m-3m, 3m-6m, 6m-12m, and >12m. Net income computed using gap approach with bucket mid-points and shocks to risk-free rates.

### Bank tax rate and other P&L items
- Effective bank tax rate fluctuates with GDP growth and money market rates; projected tax rate declines by 4 pps in 2017 under the adverse scenario (relative to baseline) before picking up by 2019.
- Non-interest income:
  - Ratio to total assets kept constant at 2015 value under baseline.
  - Under adverse scenario, projected ratio is the minimum between 2015 ratio and the average of the two years with the smallest value over the past five years (min_i r).
  - Weighted average non-interest income ratio for U.K. banks in December 2015: 0.82 percent of total assets (down from 0.91 percent in June 2015).
- Non-interest expenses:
  - Kept constant relative to total assets under baseline. Under adverse scenario, cannot be below 2015 value; the 2015 value contains material misconduct costs and significant restructuring costs for some banks.

### Dividend payout rules considered
- Three approaches considered for payout ratio projections:
  - Econometric approach: regression on macrofinancial variables; suggests payout ratio peaks at about 80 percent in 2017 in the adverse scenario before converging to baseline levels at about 50 percent by 2020. Feasibility limited by CRD restrictions and declared bank policies.
  - 2016 EU-wide stress test approach: where no declared policy, payout ratio = max(30 percent, median observed payout in profitable years over past five years). Limited applicability in U.K. due to Royal Bank of Scotland plc history and Nationwide mutual structure; approach would generate substantial variability (17 percent standard deviation in payout ratios).
  - Fixed dividend payout rule of 30 percent: reflects lower bound of 2016 EU-wide stress test methodology and ensures level playing field.

### Funding structure shifts and liquidity vulnerability
- Post-crisis funding shifts:
  - Major U.K. banks’ funding from customer deposits increased by about GBP 250 billion since 2008.
  - Wholesale funding declined by over GBP 1.3 billion over the same period.
  - Asset-weighted average loan-to-deposit (LTD) ratio stood at 92 percent in December 2015.
- A generalized sell-off in fixed-income securities driven by solvency valuation uncertainty could pressure funding schemes and escalate bank funding costs.

### Econometric approach to funding costs and contagion
- Risk factors examined in three categories: (1) bank-specific (asset quality, leverage, regulatory capital, funding structure, business model, earning capacity); (2) country-specific (macro variables, real estate prices, equity prices, corporate spreads, credit growth); (3) global variables (world GDP growth, commodity prices, VIX, U.S. equity prices, emerging market exchange rates).
- Contagion captured via two-prong approach:
  - For each bank, peer group defined as the U.K. banking system excluding that bank; average peer funding costs regressed against explanatory variables.
  - Orthogonal residuals from that regression used as proxy for contagion from funding pressures in other U.K. banks.
- Funding-cost proxy: effective interest rate paid on interest-bearing liabilities (panel model with fixed effects). Robustness check: five-year senior single name CDS spreads used as alternative proxy. Data source: Bloomberg over 2000:Q1 through 2015:Q2. Interest payments computed biannually.

### Key econometric findings on funding costs and spreads
- Unemployment: a 1 percentage point rise in unemployment pushes up bank funding costs by around 26 bps (13 bps bi-annual).
- Money market conditions: a decrease in the policy or LIBOR rate by 1 percentage point reduces interest payments by about 44 bps (22 bps biannual).
- Asset quality and CDS spreads:
  - Asset quality is the only bank-specific indicator that drives CDS spreads across specifications.
  - A rise of one percentage point in the NPL ratio leads to a rise of CDS by around 12 basis points.
  - Investors (CDS market) appear more credit sensitive than depositors.
- Contagion:
  - Contagion from other U.K. banks is a key driver of CDS spreads; coefficient for contagion is positive and statistically significant at the one percent level.
  - For the average bank, a rise in 100 CDS spreads in the rest of the U.K. system is associated with a rise in bank spreads of 63 bps.
  - Contagion is not a significant determinant of effective interest rates once money market rates are included in the econometric specification.

*Italic: Source: IMF staff estimates and FSAP team analysis (source content provided).*

### 72.      A bank-by-bank econometric analysis reveals a diversity of funding cost drivers across

### _cr16163 - 72.      A bank-by-bank econometric analysis reveals a diversity of funding cost drivers across

### Bank-by-bank econometric analysis: key findings
- The analysis was replicated using funding costs on a range of bank-by-bank individual regressions to distill cross-sectional variation in determinants of funding costs.
- The impact of bank specific determinants on funding costs differs across types of banks (Appendix II, Table 2).
  - This reflects diversity in banks’ liability structure, capital retention, and recapitalization efforts in publicly intervened banks.
- The effect of the U.K. housing market is broadly significant.
  - Rising prices in the U.K. residential real estate market tend to push down funding costs, except for U.K. banks with significant operations outside the U.K.
- The effect of contagion from peer banks’ funding pressure is robust for specific banks.
  - Even after controlling for conditions in money markets, contagion from peer banks’ funding costs affects interest payments of selected U.K. banks, with an elasticity of about 0.9.
  - This analysis may help identify banks potentially exposed to funding pressures in peer banks.

*Data note*: The data are sourced from Bloomberg at the biannual frequency. A cubic spline method is used to convert biannual data at the quarterly frequency in line with macroeconomic data.

### Projection of an add-on to funding costs (FSAP approach)
- The FSAP team considered market stress prevailing in H1 2012, banks’ funding structure, and capital adequacy.
- Under the adverse scenario:
  - A funding shock was applied to less stable liabilities. The funding model varies across banks.
  - The median value of the most stable source of funding, that is, small, insured deposits, is estimated at 46 percent of interest-bearing liabilities as of end-2015 (Figure 18).
  - The magnitude of the funding shock replicates the funding stress observed at the height of the European sovereign crisis.
    - Under the adverse scenario, the margin paid on less stable liabilities was assumed to rise by the maximum increase observed in H1 2012, that is, 52 basis points annually.
    - The funding shock is applied over the first two years of the horizon.
- Funding costs edge down in the outer years of the stress test horizon as the banking sector rebuilds capital buffers.
  - The gradual improvement in bank solvency and asset quality during 2018–19 pushes down funding costs in 2019 and 2020.
  - The magnitude of this effect is around 38 bps in 2019 and 33 bps in 2020, respectively.
- Reference for methodology: Babihuga, R., and M. Spaltro, 2014, “Bank Funding Costs for International Banks,” IMF Working Paper 14/71.

### Solvency stress test results (aggregate and comparative)
- Baseline projections:
  - Aggregate CET1 ratios stabilize at around 12.6 percent by 2020 as capital buffers grow in line with regulatory balance sheet exposures.
  - The aggregate Tier 1 leverage ratio hovers around 4.8 percent over the stress test horizon.
- Under the stress scenario:
  - The aggregate CET1 ratio falls from 12.6 percent in December 2015 to a low point of 8.7 percent in 2017.
  - The aggregate Tier 1 leverage ratio falls from 5.3 percent in December 2015 to a low point of 4.0 percent in 2017.
  - In the aggregate, banks incur losses amounting to GBP 35 billion in the first two years of the scenario.
  - RWAs peak in 2017 with a 2.3 pps hike.
- Comparison with 2015 BoE concurrent stress test:
  - FSAP stress scenario: aggregate CET1 ratio of U.K. major banks falls by 3.9 pps at the low point; Tier 1 leverage ratio falls from 5.3 percent to 4.0 percent in 2017.
  - BoE stress scenario (pre-management actions basis): aggregate CET1 ratio falls by 4.0 pps (from 11.2 percent at end-2014 to 7.2 percent in 2016); PRA Tier 1 leverage ratio falls to a low of 3.4 percent.
  - Despite differences in starting dates, nature of shocks, transmission channels, and models, results on aggregate system capital impact are similar.

### Heterogeneity by bank type
- Major U.K. international banks are relatively more vulnerable to global shocks than major U.K. domestic banks.
  - Asset-weighted CET1 ratio under FSAP stress test:
    - Major U.K. international banks: decrease by 4.3 pps at the peak relative to December 2015.
    - Major U.K. domestic banks: decrease by 2.6 pps.
  - BoE stress test by type of bank: international banks’ CET1 ratios decrease by 3.7 pps at peak versus 2.6 pps for domestic banks.
- Drivers: internationally active banks face larger impairment charges, larger mark-to-market losses, sharper corrections in asset prices, and higher counterparty credit risk due to greater market activity and interconnectedness.

### Decomposition of CET1 shortfall and stress contributions
- Over the first two years of the FSAP stress test, the fall in aggregate capital ratios is mainly driven by:
  - Stressed credit losses.
  - RWAs increases.
  - Funding costs.
- At the low point of the stress, the cumulative impact on aggregate CET1 relative to end-2015 capital ratios:
  - 2.0 pps from loss provisions.
  - 1.9 pps from shifts to RWAs.
  - Aggregate impact of funding risk is around 1.4 pps, driven by the add-on funding cost induced by money market stress.
- Methodological note: similarity in the impact of both credit risk measures is partly driven by the FSAP methodological approach to compute unexpected losses, which draws mainly on point-in-time projections.

### Sensitivity tests: housing market and swap rate shocks
- U.K. residential real estate sensitivity:
  - Adverse scenario includes a cumulative peak-to-trough reduction in residential property prices of 20 percent in 2018:Q3.
  - This reduction is estimated to reduce recovery rates by 3.5 pps.
  - An increase of 20 pps in LGD would depress the aggregate CET1 ratio by 120 bps.
  - The most significant contributor to the decline of capital buffers is a widening of required regulatory capital through a spike in RWAs.
  - Results show the impact is broadly linear on the size of the LGD shock.
- Sterling swap curve sensitivity:
  - A separate sensitivity test assessed shocks in the swap curve at the one-year maturity.
  - Transmission channels: repricing of AFS and trading-book portfolio; maturity gaps for interest rate risk calculations.
  - Net effect shows concave and convex sections with an inflection point at about a 200 bps increase.
  - An increase of 250 bps in risk-free rates would depress aggregate CET1 ratio by about 120 basis points driven by a decline in retained capital.
  - Caveat: this sensitivity test does not take into account propagation effects from the swap curve to other asset classes or broader risk types.

### FSAP team analysis of bottom-up (BU) projections: overview and findings
- Context:
  - FSAP applied a quantitative tool to assess BU projections generated by banks in the context of the 2015 BoE concurrent stress test.
  - The BoE uses a hybrid approach: TD models and BU banks’ models inform stress test results.
  - FSAP used its own TD models to assess selected BU projections using publicly available data for cross-validation of aggregate BU projected credit risk parameters for unexpected losses under the 2015 BoE scenario.
- Focus of FSAP analysis:
  - Stressed PDs and LGDs for selected IRB exposures relevant to major U.K. banks across:
    - U.K. exposures: retail secured lending and corporate exposures.
    - Non-U.K. exposures: corporate exposures in China, Hong Kong SAR, Ireland, and the U.S.
- Reasons for BU vs FSAP differences (cautions):
  1. FSAP projections are based on risk-neutral default measures extracted from market prices or ex post write-off rates for selected exposures rather than on physical default probabilities of banks’ portfolios.
  2. FSAP results provide aggregate estimates of credit risk, while BU projections are granular at borrower or loan level and account for diversification, grade migration, LTV, credit risk mitigation, and other loan-level factors.
  3. Modeling differences:
     - FSAP projections use IMF in-house TD models and PiT estimates for both PD and LGD.
     - BU projections use banks’ PRA-approved IRB models with regulatory calibrations required for capital calculation, including a long-run average calibration for PD and a downturn calibration for LGD.
- Comparison results:
  - FSAP analysis broadly confirms banks’ BU calculations, although FSAP projections exhibit a more cyclical behavior.
    - Peak of stressed PDs between the two sets of projections is similar: the average maximum PD across portfolios reaches the same value, with variation across portfolios.
    - BU estimates of LGD for corporate exposures reflect a weighted average of FSAP team LGD projections for investment grade (IG) and high yield (HY) corporate bonds.
    - FSAP projections are more conservative for cross-border exposures, relatively less conservative for U.K. exposures.
    - FSAP projections are relatively front-loaded during the first period of stress and rebound in the outer years of the horizon, explained partly by the PiT approach and market-based data reflecting nonlinear effects from cyclical shocks on credit risk parameters.

*Italic source: Extracted from IMF staff analysis in the specified content unit.*

### 83.      Figure 24 summarizes the framework and main steps involved in the FSAP team credit

### _cr16163 - 83.      Figure 24 summarizes the framework and main steps involved in the FSAP team credit

### Framework and main steps of FSAP credit risk analysis
- Two-step process to forecast PDs for each portfolio:
  - Regression equations used to project aggregate PD proxies.
  - Time series regression comparing PD proxies and bank reported PDs; the average forecast across banks used to compute aggregate PD projections.
- LGD forecasting:
  - Structural approach applied to extract losses from corporate spreads using the 2015 BoE scenario.
  - Structural approach yields expected losses by asset class as a function of corporate PDs and LGDs.
  - Aggregate BU PD estimates used to back out implied LGD projections underlying projected expected losses.

### Data used in estimation
- Three data types combined:
  - Macroeconomic and financial market variables drawing on the 2015 BoE scenario; quarterly frequency over 2000:Q1 through 2019:Q4. Variables used are a subset of those included in the scenario.
  - Aggregate credit risk data from BoE write-off data on U.K. exposures by asset class; data run over 1995:Q1 through 2014:Q4.
  - For banks’ non-U.K. exposures, Moody’s EDF rates used as a PD proxy. Estimation period starts in 2004:Q1 for U.K. exposures and in 2002:Q1 for non-U.K. exposures, and ends in 2014:Q4.
  - Bank-level credit risk data drawn from banks’ Pillar 3 annual disclosures over 2008–14.

### 2015 BoE adverse scenario (overview and key projected impacts)
- Scenario explores deterioration of global economy over 2015–19; output falls in the U.K., the euro area, and emerging markets.
- U.K.-specific:
  - Growth turns negative in 2015:Q3 as export demand falls and higher risk premia push up banks’ funding costs and credit growth contracts.
  - U.K. inflation turns negative for the first seven quarters of the scenario.
  - Bank rate is set at the zero bound.
- Regional troughs cited:
  - Aggregate euro-area real GDP growth troughs at -2.1 percent in 2016:Q1.
  - Real GDP growth in China falls to a rate of 1.7 percent in 2015:Q4.
  - Hong Kong SAR real GDP growth reaches about -6 percent at its trough.
- Financial variables and traded risk scenario included detailed projections for:
  - U.K. yield curve, German yield curve, average euro area excluding periphery government bond spreads, average European periphery (ex-Greece) government bond spreads, U.S. yield curve, 3 million LIBOR rates across major currencies, Sterling LIBOR swap rates, EURIBOR swap rates, and U.S. dollar LIBOR swap rates.
  - Traded risk shocks to FX, equities, credit spreads, commodities, rates, FX volatility, equity volatility, rates volatility, and structured finance shock across currencies, maturities, and liquidity horizons.

### Econometric methodology for PD projections
- U.K. exposures:
  - VAR model combined with PCA analysis.
  - VAR structure: each variable as linear function of its own past values, past values of other variables, and a serially uncorrelated error term.
  - For each asset class, VAR involved four equations reflecting four factors: credit risk, macroeconomic conditions, financial conditions, and real estate conditions.
  - PCA used for dimensionality reduction to identify variables in the 2015 BoE scenario that account for maximum variability with minimum factors.
  - Different PCA variable sets across portfolios (example: household income included for mortgage macro risk factor; corporate profits for corporate macro factor).
  - Key financial risk factor loadings differ by portfolio (example: residential prices and credit growth in secured lending to individuals for mortgage book; commercial real estate prices and credit growth in corporate lending for corporate portfolio).
  - Econometric results: mortgage loss rates driven mainly by tight financial conditions for households; corporate risk rises with widening corporate spreads and CRE correction.
- Overseas corporate exposures in advanced countries (Ireland and the U.S.):
  - Same econometric strategy applied.
  - Ireland: shocks to residential prices and monetary policy shocks are key drivers of corporate stress.
  - U.S.: PDs in corporate exposures explained mainly by shocks to IG credit spreads.
  - Credit risk across loan categories is highly autoregressive.
- Asian exposures:
  - Quantile regression approach to capture tail of distribution; semi-parametric estimation at the 70th percentile chosen to capture tail of extreme PD realizations.
  - Results: China distress driven mainly by volatility in U.S. markets and FX depreciation; Hong Kong SAR impacted by corporate distress in mainland China, indicating sizable spillovers.
  - Caveat: large confidence intervals associated with quantile regression due to forecast uncertainty in the tail.

### LGD projections using corporate spreads and a structural (Merton) approach
- BoE scenario projects sharp increase in corporate bond spreads:
  - Yields on high-yield U.K. corporate bonds widening by over 1,000 bps in 2015.
  - U.K. investment grade corporate spreads increase to about 100 bps.
- Merton structural model applied to relate credit spreads to corporate default rates and loss rates; used to extract corporate default risk.
- Procedure to back out LGDs:
  - Use five-year time to maturity assumption.
  - Apply structural model to estimate path of LGDs for:
    - sterling IG corporate bonds,
    - sterling HY corporate bonds,
    - U.S. dollar IG corporate bonds,
    - U.S. dollar HY corporate bonds.
  - Plugging path of BU PDs into estimated path of expected losses yields path of estimated LGDs.
  - Using BU cumulative LGD shock to the U.K. (U.S.) corporate portfolio, implicit rating of the U.K. (U.S.) loan portfolio can be backed out.
- Key findings and caveats:
  - LGD shocks significant for low-rated asset classes but relatively moderate for investment grade bonds.
  - Shocks are front-loaded during the first year of the scenario, with a quick recovery in outer years.
  - Caveat 1: liquidity premia may distort the theoretical relationship between credit spreads and expected losses.
  - Caveat 2: differential tax treatment of corporate relative to risk-free bonds may explain variation in credit spreads.
  - Caveat 3: idiosyncratic default risk on individual corporate issuers is not priced-in because the BoE scenario focuses on corporate bond indices rather than single-name issuers.

### Key elements of the structural model (Box 2 summary)
- Risk-neutral no-arbitrage condition links risky bond yield spreads to expected loss from default, expressed in terms of cumulative risk-neutral PD and LGD.
- Relationship used by FSAP team to solve for estimated LGD embedded in BoE projected credit spreads:
  - With notation preserved from source, the no-arbitrage and solved expressions relate T-period risky bond prices and yields to PD and LGD, and the FSAP team used submitted average PD by portfolio to solve for LGD.

### Liquidity stress tests and LCR implementation
- LCR implementation:
  - LCR came into effect on October 1, 2015.
  - PRA revoked liquidity standards in U.K. liquidity regulation, including firms’ individual liquidity guidance (ILG) on that date.
  - Firms continue to submit ILG reporting returns during an interim period until at least October 2016.
  - LCR ratio transitional path:
    - Set at 80 percent until end-2016,
    - Rise to 90 percent in 2017,
    - Reach 100 percent in January 2018, as required by the CRR.
  - The 80 percent starting point is above the 70 percent ratio set out by the CRR based on the narrower range of liquid assets allowed by the PRA’s former regime.
- Liquidity stress tests conducted:
  - Suite of liquidity stress tests carried out by the BoE on scenarios calibrated by the FSAP team, drawing on interim LCR report and PRA liquidity return.
  - PRA ILG return PRA 048 runs in parallel to COREP LCR returns until at least October 2016; both returns used to populate IMF liquidity stress test templates.
  - Sample for PRA run specified to provide at least 80 percent coverage of total U.K. banking assets as measured by PRA liquidity returns and to include a diversified sample of firms.
  - Sample consists of 10 large financial institutions: seven large U.K. firms covered in the solvency stress test plus three large U.K. subsidiaries of major foreign investment banks.
  - Data sources for tests:
    - Interim LCR return as of end-December 2015 used to populate LCR liquidity stress test data; PRA ensured consistent mapping of asset haircuts, outflows, and inflow rates between IMF template and EU LCR template.
    - Implied cash flow tests and maturity mismatch test rely on PRA 048 returns as of January 1, 2016.
    - Single currency analysis based on PRA ILG returns.

### PRA liquidity regime (Box 3 key comparisons with Basel III LCR)
- PRA revoked liquidity standards in Prudential sourcebook (BIPRU 12) in October 2015; BIPRU 12 had been effective in mid-2010 for largest U.K. banks and building societies.
- Basel III LCR transitional path reiterated: 80 percent in October 2015, 90 percent in 2017, 100 percent in 2018.
- Key differences between PRA regime and Basel III LCR:
  - PRA used three liquidity horizons for stress test: two-week, one-month, and three-month; LCR uses a single one-month horizon.
  - PRA scenarios are not fully prescribed and vary according to funding vulnerabilities exposed in supervisory review; LCR scenarios are almost fully prescribed.
  - PRA counterbalancing capacity includes a narrower range of eligible liquid assets than the LCR; assets broadly confined to LCR’s Level 1 assets.
  - PRA’s buffer requirement applies a suite of “add-ons” to capture risks not included in the LCR (including intraday risk); these carried forward as interim Pillar 2 requirements for LCR.
  - ILG regime was somewhat stricter than LCR, applying stresses to retail deposits maturing beyond the LCR’s 30-day time horizon, with all retail and corporate lending rolling over in the stress.
  - PRA adopted an interim Pillar 2 approach to reflect risks not captured in LCR, based on firms’ existing ILG add-ons; PRA’s Pillar 2 regime for liquidity risk is currently under review and includes an add-on for risks not reflected in LCR using PRA’s liquidity regime.

### Liquidity stress test scenarios (predefined and tailored)
- Three predefined LCR liquidity stress scenarios, including two tailored to U.K. banks:
  - LCR scenario with standard parameters set out by Basel III LCR 2013.
  - U.K. retail stress scenario (replicates a “Northern Rock event”):
    - Run-off rates up to 15 percent for retail deposits.
    - Rates of 60 percent for corporate deposits.
    - Rates of 50 percent for operational deposits generated by clearing, custody, and cash management activities.
    - Trade finance contingent liabilities of 10 percent.
    - Increase of outflows from retail and corporate committed but undrawn credit and liquidity facilities of up to 50 percent.
  - U.K. wholesale stress scenario (repeat of global financial crisis liquidity stress):
    - Run-off rates up to 100 percent for corporate deposits.
    - Rates of 75 percent for operational deposits generated by clearing, custody, and cash management activities.
    - Outflows of secured funding backed by level-2B assets of up to 100 percent.
    - Increase of contingent funding liabilities of up to 75 percent.

### Implied cash flow tests and 30-day test assumptions
- Two implied cash flow tests simulate gradual outflow of funding over five consecutive days and over a 30-day time horizon; assumptions are more severe than money market conditions observed during the 2008 crisis.
- 30-day test key assumptions:
  - Haircuts of up to 60 percent for securities and bank loans mobilizable in repo transactions backed by other than general government collateral.
  - No issuance of new unsecured funding and freeze of securitization markets.
  - Callback rates of 100 percent on expected cash inflows related to credit extension without liquid financial assets as collateral, and no offsetting cash inflows from renewed wholesale lending.
  - Cash outflows of up to 75 percent of maturing and non-maturity funding without liquid financial assets as collateral, and no renewal of term retail and wholesale deposits.

*UNITED KINGDOM — INTERNATIONAL MONETARY FUND*

### 98.      The maturity mismatch test is calibrated on six different maturity buckets, ranging

### _cr16163 - 98.      The maturity mismatch test is calibrated on six different maturity buckets, ranging

### Maturity mismatch test—calibration and assumptions
- Six maturity buckets ranging from less than one week to over one year.
- Key assumptions:
  - All outflows with no maturity (retail accounts, derivative outflows, and corporate current accounts) are classified in the “less than one week” bucket.
  - Half of the deposits maturing in less than two weeks are reported in the “deposit under one week” bucket.
  - All securities are classified according to their residual maturity, including highly rated long-term tenor gilts.

### Liquidity stress testing framework
- Single-currency analysis:
  - Based on an internal measure used by the PRA under the previous PRA liquidity regime.
  - Scenario replicates very severe wholesale stress, with no rollover of contractual maturities and immediate withdrawal of all on-call wholesale funding.
  - Counterbalancing capacity is narrowly defined to include only cash and the highest quality of liquid assets.
  - Analysis conducted by significant currency (those comprising more than 20 percent of the balance sheet); for the sample, the three significant currencies are British pounds, the euro, and the U.S. dollar.

### Liquidity stress test results — key findings
- LCR and aggregate results:
  - All firms passed the LCR stress test for all three scenarios under the current 80 percent LCR hurdle rate requirement in the United Kingdom.
  - The aggregate LCR ratio under the standard Basel III assumptions stands at about 140 percent.
  - Under the “U.K. retail” scenario and the “U.K. wholesale” scenario, the average ratio declined to about 95 percent.
  - The LCR ratio’s sensitivity to a retail event and a wholesale event was similar.
- Table 8 (as reported)
  - Banks Passing101010
  - Banks Not Passing000
  - Net Cash Outflows510722745
  - Value of Liquid Assets, After Haircuts701701701
  - Aggregate Coverage Ratio (percent)1389794
- Implied cash-flow tests:
  - All firms passed the five-day and 30-day implied cash flow tests.
  - The average liquidity risk-bearing counterbalancing capacity to stressed cash flows:
    - over 180 percent for the five-day test
    - under 150 percent for the 30-day test
  - Under an alternative scenario of stress with no retail deposit run:
    - average counterbalancing capacity would rise to over 325 percent (five-day test)
    - and 250 percent (30-day test)
  - Results suggest protracted noncumulative stressed cash flows over a longer horizon weaken liquid buffers more than cumulative stress over a shorter period.
  - Greater variation among banks’ resilience to short-lived stress than to sustained stress, reflecting variability in short-end funding models.
- Maturity mismatch analysis:
  - Five of the ten covered firms show a shortfall in the one- to four-week bucket.
  - This is explained by retail-focused commercial banks and building societies that focus on long-term lending funded with short-term deposits.
  - The scenario assumes that all securities-related flows, including HQLA, can only be realized when the securities mature (very severe assumption).
- Single-currency results:
  - All firms have sufficient liquid buffers in domestic currency (GBP).
  - Only a minor shortfall in foreign currency:
    - No U.K. firm has a shortfall in GBP.
    - Only one firm out of 10 has a negative position in EUR.
    - Two firms show a slight negative shortfall in USD.

### Overall assessment from FSAP stress tests
- Capital and leverage:
  - The FSAP macroeconomic stress scenario reduces the aggregate CET1 ratio from 12.6 percent to a low point of 8.7 percent.
  - The leverage ratio falls from 5.4 percent in December 2015 to a low point of 4.0 percent.
- Drivers of resilience:
  - Strong initial positions (risk-weighted capital ratios more than doubled from pre-crisis levels).
  - De-risking through reductions in international exposures, including global investment bank activities and non-core foreign operations.
  - Improved LTV distributions in U.K. mortgages, more stable funding structures, and reduced leverage.
- Liquidity stability:
  - Positive liquidity stress test results are consistent with more stable funding structures.
  - All firms pass the LCR stress test under the current regulatory 80 percent hurdle rate and tailored U.K. retail and wholesale scenarios.
  - All firms pass five-day and 30-day implied cash flow tests on an all-currency basis; PRA single-currency analysis confirms sufficient GBP liquidity.

### Caveats and limitations
- Stress test interpretation caveats:
  - Results are based on banks’ submitted data at the cut-off date, complemented by market-based and publicly available data.
  - Matching and reconciliation of risk data from multiple sources is complex; caveats are in place.
  - Major U.K. banks have been changing business models to improve profitability and comply with new prudential frameworks; structural shifts constrain the reliability of past data for forward-looking projections.
  - Specific caveat: uncertainties associated with the projected path of nominal bilateral exchange rates between the GBP and the USD over the stress test horizon, since two major U.K. banks report regulatory balance sheets in USD.

### Appendix I — IMF credit risk model for IRB exposures (methodology)
- Modeling approach:
  - Battery of credit risk models to project PDs using a multivariate VAR approach combined with PCA.
  - VAR: each equation is a function of its own lags and lags of other variables in the system; error vector with zero mean and covariance matrix Ω.
- Conditional PD construction:
  - Series denoted by {absPDj,sti, , } at time t for country i, asset class j, and scenario s=b (baseline) and s=a (adverse).
  - Two methods used:
    - Comprehensive range of bilateral VARs for all pairs of credit risk and each macro/financial variable in the scenario; average forecast across specifications reflects deterioration conditions.
      - For U.K. exposures, over 170 VARs were estimated linking a PD proxy (EDF or write-off rate) for each corporate class (mortgages, corporate, retail unsecured, and financial) and a set of macroeconomic and financial variables generated by the scenario (24 key variables).
      - Separate estimations and forecasts for baseline and adverse scenarios.
      - Same approach applied for international exposures in 14 geographies.
    - Multivariate VAR broken down by asset class and geography, combined with a factor model reflecting deteriorating conditions in four categories: macroeconomic variables, financial variables, real estate prices, and global conditions.
- Data and coverage:
  - Appendix II, Table 1: summary descriptive statistics of key variables used to feed the battery of U.K. credit risk models.
  - Data run from 1995:Q1 to 2015:Q2.
- Factor model and PCA:
  - Four factors:
    - Macro factor: GDP growth, inflation, unemployment, FX rates, policy rates.
    - Financial factor: money markets (LIBOR rates), term structure, credit spreads, equity returns, credit availability/cost.
    - Real estate factor: residential housing and CRE developments.
    - Global factor: interest rate swap markets, U.S. equity markets, commodity prices (fuel and nonfuel).
  - Dimensionality reduction:
    - First factor of each principal component included as endogenous driver.
    - First factor explains about:
      - 65 percent of macroeconomic conditions
      - 50 percent of financial conditions
      - 90 percent of real estate conditions
      - over 70 percent of global conditions for the United Kingdom

### Appendix II — funding cost variables and bank-specific results
- Funding cost model dependent and explanatory variables (selected labels as reported):
  - Dependent variable:
    - Funding cost = Interest expense to interest-bearing liabilities (interest)
    - 5 year single name CDS spread (CDS)
  - Contagion:
    - Orthogonalized residual of regressing the rest of the UK banking system funding costs on all fundamental drivers given specification j (cont_spec j)
  - Bank-specific variables (examples):
    - NPL ratio (npl)
    - Loss provision to total loans (loss_r)
    - Tangible equity to total assets (te_ta)
    - CT1 ratio (ct1)
    - Short-term borrowing to total funding (stb_tf)
    - LTD ratio (ltd)
    - ROA (roa)
  - Country-specific variables (examples):
    - GDP growth (g)
    - Inflation rate (inf)
    - Unemployment rate (unem)
    - Residential property price growth (housing)
    - FTSE price index growth (equity)
    - Policy rate (pol), 3m LIBOR GBP (libor)
  - Global variables (examples):
    - GDP growth - world (g_w)
    - Oil price (oil)
    - S&P 500 growth (equity_us)
    - VIX (VIX)
- Bank-by-bank regression results (summary metadata and key points):
  - Estimation period: 2000Q1-2015Q2
  - Table 2 reports bank-specific regressions across 12 model columns for six banks (Bank 1 through Bank 6).
  - Selected coefficient examples (as reported in table rows):
    - ct1: coefficients include -0.008, -0.004, 0.003, -0.011, -0.096***, -0.051*, -0.009, -0.018**, -0.049, -0.019, 0.090**, 0.111*** (standard errors in parentheses where shown)
    - lev: coefficients include -0.017***, -0.015***, -0.003, -0.005, -0.071***, -0.051***, -0.019***, -0.020***, -0.110**, -0.097**, 0.027*, 0.038** (standard errors in parentheses where shown)
    - libor: coefficients include 0.162***, 0.164***, 0.204***, 0.189***, 0.190***, 0.230***, 0.111***, 0.085***, 0.222***, 0.250***, 0.089, 0.031 (standard errors reported)
    - cont_spec j: coefficients include 0.668***, 0.648***, -0.141, 0.286, -0.083, 0.043, 0.386***, 0.355***, 1.172***, 0.893***, 1.343***, 1.306*** (standard errors reported)
  - Observations and fit:
    - Observations reported per regression: 56, 56, 54, 54, 44, 44, 44, 44, 58, 58, 36, 36
    - R-squared values per regression: 0.960, 0.961, 0.845, 0.867, 0.902, 0.911, 0.982, 0.984, 0.880, 0.898, 0.914, 0.922
  - Robust standard errors; significance notation: *** p<0.01, ** p<0.05, * p<0.1

*Italic: IMF staff report content as provided in the source document.*

### Appendix III. BoE Scenario

### Appendix III. BoE Scenario

### BoE Scenario — U.K. Projections and Foreign Adverse Projections
- Figures referenced: "Figure 1. BoE Scenario—U.K. Projections" and "Figure 2. BoE Scenario—Foreign Adverse Projections" (figures shown in source).

### Tail shock calibration and macro outcomes
- Scenario characterized by a broad-based global recession with major adverse implications for China and the euro area, generating a domestic U.K. recession and disinflationary pressures.
- U.K. real GDP growth:
  - Contracts by 2.3 percent year-on-year from 2014:Q4 to 2015:Q4.
  - Reaches a peak deviation from baseline in 2017 at -7.7 percent.
- Unemployment:
  - Rises by 3.5 pps by 2017:Q3.
- Asset prices:
  - Equity prices decline by 36 percent peak-to-trough.
  - Residential property prices fall by 20 percent.
- Alternative/severe scenario features:
  - Disorderly accelerated monetary normalization in the U.S., abrupt asset price correction, and financial crises in fragile emerging economies.
  - Scenario constitutes a 2.1 standard deviation move in two-year cumulative real GDP growth rate for 2016–17.
- Additional scenario variant outcomes (more severe):
  - Euro area year-on-year real GDP growth troughs at -2.1 percent in 2016:Q1.
  - China year-on-year real GDP growth falls to 1.7 percent in 2015:Q4.
  - Peak-to-trough fall in real GDP: about 7 percent in Brazil and 4 percent in South Africa.
  - VIX peaks at 46 pps in 2015.
  - Oil price troughs at USD 38.
  - Renminbi depreciates 10 percent against the USD by end-2015.
  - Euro depreciates by about 25 percent against the USD and by 15 percent against sterling in 2015.
- Severe U.K. variant:
  - U.K. GDP growth contracts by 1.6 percent in 2016.
  - Reaches a peak deviation from baseline in 2017 at -6.9 percent.
  - Real house price falling by 40 percent in 2016 and 2017.
  - Real equity price falls by 40 percent during 2016.
  - Banks’ lending rates rise by 1.1 pps by 2017.
  - Bank credit falls by 6.7 percent.
- Emerging markets output:
  - Output falls by 11.9 percent in the fragile three (Brazil, South Africa, and Turkey).
  - Output falls by 1.3 percent in other emerging economies.
- Scenario includes an additional idiosyncratic and system-wide funding risk shock linked to money market dislocation and banks’ capital ratios under stress.

---

### Appendix IV. LCR Scenarios and Implied Cash-Flow Assumptions

### Table 1. LCR Scenario — Inflows (eligibility and haircuts)
- A. Eligibility of liquid assets (percent listed per scenario):
  - Level 1 Assets 100% 100% 100%
  - Level 2a Assets 85% 85% 85%
  - Level 2b Assets 75% 75% 75%
  - (two subsequent rows) 50% 50% 50% and 50% 50% 50%
- B. Haircuts on Inflows of liquid assets (over 30 days) (percent listed per scenario):
  - Level 1 assets 0% 0% 0%
  - Level 2a assets 15% 15% 15%
  - Eligible RMBS 25% 25% 25%
  - Other Level 2b assets 50% 50% 50%
  - Margin lending backed by all other collateral 50% 50% 50%
  - All other assets 100% 100% 100%
  - Credit or liquidity facilities 0% 0% 0%
  - Operational deposits held at other financial institutions 0% 0% 0%
  - Other inflows, by counterparty:
    - Retail counterparties 50% 50% 50%
    - Nonfinancial wholesale counterparties, transactions not listed above 50% 50% 50%
    - Financial institutions and central banks, transactions not listed above 100% 100% 100%
  - Net derivative cash inflows 100% 100% 100%
  - Other (contractual) cash inflows 100% 100% 100%
- Qualifying asset categories listed (no numeric changes beyond above): Qualifying covered bonds rated AA- or better; Qualifying Mortgage Backed Securities; Qualifying corporate debt securities rated between A+ and BBB-; Qualifying common equity shares; Coins and bank notes; Qualifying marketable securities from sovereigns, central banks, PSEs; Qualifying central bank reserves; Domestic sovereign or central bank debt for nonzero risk-weighted entities; Qualifying corporate debt securities rated AA- or higher.

### Table 2. LCR Scenario — Outflows (percent assumptions by scenario)
- C. Outflows of liquid assets (over 30 days)
- Retail Deposits:
  - Demand deposits — Stable deposits 5% 10% 5%
  - Less stable retail deposits 10% 15% 10%
  - Term deposits, residual maturity > 30d 0% 0% 0%
- Unsecured Wholesale Funding:
  - Demand and term deposits, residual maturity < 30d, small business:
    - Stable deposits 5% 10% 5%
    - Less stable deposits 10% 15% 10%
  - Operational deposits generated by clearing, custody, and cash management activities 25% 50% 75%
  - Portion covered by deposit insurance 5% 5% 50%
  - Cooperative banks in an institutional network 25% 25% 100%
  - Nonfinancial corporates, sovereigns, central banks, multilat development banks, PSEs:
    - Fully covered by deposit insurance 20% 20% 50%
    - Not fully covered by deposit insurance 40% 60% 100%
  - Other legal entity customers 100% 100% 100%
- Secured Funding:
  - Secured funding with a central bank, or backed by Level 1 assets 0% 0% 0%
  - Secured funding backed by Level 2A assets 15% 15% 15%
  - Secured funding backed by non-Level 1 or non-Level 2A asset, with domestic sovereign, mu 25% 25% 50%
  - Funding backed by RMBS eligible for Level 2B 25% 25% 50%
  - Funding backed by other Level 2B assets 50% 50% 100%
  - Other secured funding transactions 100% 100% 100%
- Additional Requirements (percent assumptions):
  - Valuation changes on non-Level 1 posted collateral securing derivatives 20% 20% 20%
  - Excess collateral held by bank related to derivative transactions that could be called anytime 100% 100% 100%
  - Liquidity needs related to collateral contractually due on derivatives transactions 100% 100% 100%
  - Increased liquidity needs related to derivative transactions allowing collateral substitution 100% 100% 100%
- ABCP, SIVs, conduits, SPVs, or similar:
  - Liabilities from maturing 100% 100% 100%
  - Asset backed securities 100% 100% 100%
- Undrawn but committed credit and liquidity facilities:
  - Retail and small business 5% 10% 5%
  - Nonfinancial corporates, sovereigns, central banks, multilat dev. banks, PSEs:
    - Credit facilities 10% 30% 10%
    - Liquidity facilities 30% 50% 30%
  - Supervised banks 40% 50% 50%
  - Other financial institutions:
    - Credit facilities 40% 50% 50%
    - Liquidity facilities 100% 100% 100%
  - Other legal entity customers, credit and liquidity facilities 100% 100% 100%
- Other contingent funding liabilities:
  - Trade finance 5% 10% 5%
  - Customer short positions covered by customers' collateral 50% 75% 75%
  - Additional contractual outflows 100% 100% 100%
  - Net derivative cash outflows 100% 100% 100%
  - Any other contractual cash outflows (not listed above) 100% 100% 100%

---

### Appendix IV. Cash Flow Assumptions — Implied Cash Flow Assumptions (Table 3)

### 5-day implied cash flow (ICF) test — cumulative inflow and outflow over 5 consecutive days
- Liquid financial assets definitions and haircuts/ranges:
  - (i) cash and cash balances with central banks [haircut: 0 percent]
  - (ii) securities and bank loans eligible at major central banks [0-15]
  - (iii) securities and bank loans which can be mobilized in repo transactions (or another type of lending against financial collateral) [5-30]
  - (iv) marketable securities [10-35]
- Cumulative cash inflows:
  - (i) expected cash inflows related to credit extension without liquid financial assets as collateral [call-back rate: 20 percent per day]
  - (ii) expected inflows of cash and liquid assets related to maturing transactions with liquid securities and bank loans (e.g., reverse repo and securities borrowing transactions) [20]
  - (iii) expected and potential net cash flows related to derivatives (excl. credit derivatives) – net contractual cash flows [20]
  - (iv) potential inflows from committed/uncommitted credit lines to related and third parties [5/3]
- Cumulative cash outflows:
  - (i) maturing and non-maturity funding without liquid financial assets as collateral [discount factor: 5 percent per day] (i.e., all deposits and funding from financial and non-financial corporates as well as private households and SME clients) with the exception of sovereign and other public sector and central bank clients [0]
  - (ii) expected outflows of cash and liquid assets related to transactions with liquid securities and bank loans (e.g., repo and securities lending transactions) [20]
  - (iii) maturing outflows to related parties [20]
  - (iv) committed/uncommitted contingent claims to related and third parties [5]

### 30-day implied cash flow (ICF) test — non-cumulative
- Liquid financial assets definitions and ranges:
  - (i) cash and cash balances with central banks [0]
  - (ii) securities and bank loans eligible at major central banks [0-20]
  - (iii) securities and bank loans which can be mobilized in repo transactions (or another type of lending against financial collateral) [10-60]
  - (iv) marketable securities [20-70]
- Non-cumulative cash inflows:
  - (i) expected cash inflows related to credit extension without liquid financial assets as collateral [call-back rate: 100 percent]
  - (ii) expected inflows of cash and liquid assets related to maturing transactions with liquid securities and bank loans (e.g., reverse repo and securities borrowing transactions) [100]
  - (iii) expected and potential net cash flows related to derivatives (excl. credit derivatives) – net contractual cash flows [100]
  - (iv) potential inflows from committed/uncommitted credit lines to related and third parties [23/12]
- Non-cumulative cash outflows:
  - (i) maturing and non-maturity funding without liquid financial assets as collateral [10-75] (i.e., all deposits and funding from financial and non-financial corporates as well as private households and SME clients) with the exception of sovereign and other public sector and central bank clients [0]
  - (ii) expected outflows of cash and liquid assets related to transactions with liquid securities and bank loans (e.g., repo and securities lending transactions) [100]
  - (iii) maturing outflows to related parties [100]
  - (iv) committed/uncommitted contingent claims to related and third parties [23]
- Test interpretation and basic assumptions:
  - A ratio lower than 100 percent implies a liquidity shortage if the stress scenario would materialize at the reporting date.
  - Only unencumbered liquid assets (generating cash inflows) are included in the test.
  - New unsecured financing and securitization impossible within the time horizon.
  - No offsetting cash inflows from new or renewed (secured/unsecured) wholesale lending (at contractual maturities) but full renewal of secured retail lending (e.g., secured lending with illiquid collateral (residential mortgages)).
  - Central bank eligible collateral can be monetized at appropriate haircuts; repo markets are open at appropriate haircuts; fire-sale of assets possible at appropriate haircuts.
  - No consideration of funding via potentially re-usable securities received as collateral ("rehypothecation").
  - Limited potential unsecured support in convertible currencies from related and third parties (e.g., in the form of committed lines).
  - No renewal of term retail and wholesale deposits.
  - Full convertibility between currencies (within one week).

---

### Appendix V. Stress Test Matrix (STeM): Solvency and Liquidity Risks — Banking Sector Solvency Test

### 1. Institutional perimeter, data, and stress testing process
- Institutions included (seven major banks and building societies):
  - Barclays plc, HSBC Holdings plc, Lloyds Banking Group plc, Nationwide Building Society, the Royal Bank of Scotland Group plc, Santander UK plc, and Standard Chartered plc.
- Criteria for inclusion: firms’ balance sheet, role in the U.K. payment system, and firms’ plans to grow their balance sheet.
- Market share: Approximately 80 percent of PRA-regulated banks’ lending to the UK real economy.
- Data (authorities’ BU effective dates and FSAP team data):
  - BU effective date: end-December 2014.
  - BU effective date for market risk: February 20, 2015.
  - Data: Stress testing templates and associated documents submitted by participating banks for the 2015 Stress Test.
  - BU scope of consolidation: Global consolidated group basis, except for Santander UK plc (parent supervised by a foreign authority). Perimeter of the banking group (CRD IV). Insurance activities excluded but impact must be assessed.
  - TD effective date: end-December 2015.
  - TD effective date for market risk: end-December 2015.
  - TD data: Firm-by-firm confidential data at the cut-off date, BoE data on aggregate write-off rates for UK exposures, and publicly available data (Pillar 3, Bloomberg, Datastream, Markit, 2015 EU-wide transparency exercise, Haver Analytics, Mortgage Lenders and Administrators Statistics, Moody’s KMV, Bankscope, SNL, IFS, IMF Global Assumptions (GAS), and IMF WEO).
  - TD scope of consolidation mirrors BU.
- Stress testing process (hybrid approach):
  - Scenarios designed by the BoE and approved by the FPC and PRA Board.
  - Each participating firm generates additional scenario variables as required.
  - Firms perform constrained BU stress tests based on internal risk-management infrastructure.
  - BoE runs in-house challenger models to cross-validate firms’ results.
  - The FSAP team conducted its own TD macroprudential stress test based on IMF generated scenarios.
  - For IRB exposures, separate credit risk model calibrated for 5 Basel asset classes and 15 geographies.
  - For STA exposures, project stressed NPL ratios, stressed coverage ratios, and transition matrix for performing exposures.
  - For market risk, stress to 22 sovereign issuers and major corporate indices modeled separately.
  - TD stress test includes detailed stress test of UK mortgage book by LTV vintage using structural Merton-based approach.
  - Aggregation by BoE includes adjustments from peer comparison, challenger models, and BoE judgment.
  - FSAP team used its own credit risk models to project stressed credit risk parameters based on the 2015 BoE scenario and compared FSAP projections against banks’ BU projections.
  - Sensitivity tests included a range of stressed UK residential house prices and shocks to the swap curve.

### 2. Channels of risk propagation and methodology
- Methodology elements:
  - Risks projected using a variety of models, approaches, and judgments from banks and BoE/PRA.
  - Banks’ internal models translate scenarios into credit risk losses across asset classes and geographies.
  - Traded risk elements modeled: impact on market risk positions depending on liquidity; valuation adjustments (FVO, CVA, PVA, and bid/offer reserve); prescribed number of counterparty defaults in impacted geographies.
  - Banks model impact on funding costs and pass-through to customers.
  - BoE in-house suite of supervisory risk models challenges banks’ projections and provides system-wide view.
  - Supervisory input accounts for individual banks’ business models.
  - A comprehensive battery of econometric and structural models developed and calibrated for the 2016 U.K. FSAP.
  - Over 75 credit risk models and 900 econometric specifications for PDs based on VAR, PCA, and quartile-based regressions, and a structural Merton-based approach for LGDs.
  - Lending rates linked to shocks to deposit rates (projected in line with the macro scenario), bank-specific solvency ratios, funding stress in peer banks, and shocks to NIMs (affected by base rate, money market shocks, and slope of yield curve), with pass-through estimated empirically.
  - Add-on funding shock related to funding shock in H1 2012 with disallowed pass-through.
  - Mark-to-market losses from full revaluation of sovereign securities (22 jurisdictions) and corporate fixed income debt securities, excluding hedges, under each scenario.

### 3. Tail shocks, scenario analysis, and sensitivities
- Scenario characteristics (see BoE Scenario section for detailed macro outcomes).
- Sensitivity analysis includes:
  - Stressed projections for misconduct costs.
  - Shocks to U.K. residential house prices affecting stressed LGDs.
  - Shocks to the swap curve.

### 4. Risks and buffers — positions, regulatory impact, and behavioral adjustments
- Positions/risk factors assessed:
  - Credit risk estimated according to Basel III framework under IRB advanced, IRB foundation, and Standardized approach.
  - Positions include cross-border loan exposures, interbank and public sector loans, covered bonds, securitization exposures, and off-balance sheet exposures using baseline and stressed CCFs.
  - Sovereign risk: issuer risk from shocks to yield curves across advanced and emerging economies; mark-to-market valuation of securities in trading book and AFS/FVO linked to macro scenario.
  - Market risk (non-sovereign): shocks to FX markets, corporate spreads, commodities, and money markets (IMF stress test); market stress from asset prices and volatilities calibrated by BoE and firms (BoE stress test); counterparty credit risk losses covering all trading book and banking book derivatives and securities financing transactions, as well as CVA (BoE stress test); shocks to PVA for own funding costs (BoE stress test).
  - Profits: income from investment banking activities; interest income declines from defaulted loans; interest income from non-defaulting loans estimated via satellite models (IMF stress test); interest expenses increase due to rising funding costs with empirically estimated pass-through and add-on funding stress from a market event with no pass-through to lending rates (IMF stress test); banks project interest income/expense accounting for balance sheet and funding evolution, product interest rate and margin movements, FX movements, and structural hedging (BoE stress test); net fee and commission income evolve with macro conditions and balance sheets.
- Regulatory impact:
  - Effects of phase-out of no-longer-eligible additional Tier 1 and Tier 2 capital included.
  - No conversion of additional Tier 1 capital is assumed during the stress horizon (IMF stress test).
  - Additional Tier 1 instruments can convert if trigger level is breached after management actions are taken into account (BoE stress test).
- Behavioral adjustments and dynamic balance sheets:
  - BoE: balance sheet evolves under stress per each bank’s corporate plan, adjusted for the scenario.
  - BoE scenario includes paths for U.K. lending to individuals and PNFCs calibrated to reflect reduced credit demand, with guidance that individual banks’ market share of lending should not decline under stress to ensure aggregate consistency.
  - Size and composition of non-U.K. lending allowed to vary: (i) if non-U.K. lending has positive growth under baseline, slower growth allowed under stress but no contraction; (ii) if non-U.K. lending has negative growth under baseline, no further contraction allowed under stress.
  - Dividend payout in stress aligned with publicly disclosed dividend policies; further cuts possible as strategic management actions. Banks may propose realistic management actions; BoE assesses timing and benefit; actions expected to be part of recovery and resolution plans.
  - Maturing assets replaced by exposures of same type and risk.
  - Dividends linked to banks’ net profits: under positive profits, dividend payout set at 30 percent; otherwise, no dividend payout assumed.
  - Effective tax rate evolves with macro scenario.
  - Losses are recognized in the same year that a shock hits.
  - If banks’ capital ratio falls below regulatory minimum during the stress test horizon, no prompt corrective action is assumed.

*Source: Appendix III–V, _cr16163 - Appendix III. BoE Scenario.*

### 5. Regulatory and market-based standards and parameters

### 5. Regulatory and market-based standards and parameters

### Calibration of risk parameters
- Parameter definition
  - Banks project PDs and LGDs across the stress scenario for both expected loan losses (impairment charges) and capital requirements.
  - PDs and LGDs are projected by asset class and geography for defaulted and non-defaulted exposures.
  - Changes in capital requirements (RWAs) driven in part by banks’ PRA approved regulatory IRB models.
- Parameter calibration
  - PDs and LGDs evolve with the macroeconomic and financial variables of the scenario, per banks’ models.
- Parameter definition (additional)
  - PiT PDs and LGDs for both expected and unexpected losses.
  - PDs are blended PDs (i.e., include both defaulted and non-defaulted counterparties) by asset class and geography.
  - LGDs are calculated post credit risk mitigation by asset class and geography.
- Parameter calibration (additional)
  - For IRB exposures, changes in PDs are proxied by shocks to write-off rates for U.K. exposures, Moody’s EDFs for cross-border exposures, and banks’ computed PDs in historical stressed episodes.
  - Shocks to LGDs are projected using a Merton-based approach for mortgage exposures, shocks to unemployment for retail unsecured exposures, and shocks to GDP for corporate exposures.
  - PDs and LGDs evolve with the macroeconomic and financial variables of the scenario.
  - For STA exposures, inflows into NPL categories are based on panel regression, including risk migration for performing exposures.

### Regulatory standards
- Capital definition according to Basel III/CRD IV/PRA rulebook, including CET1, Tier 1, and total CAR.
- The CET1 ratio is computed using CRD IV end-point definition. This follows PRA’s decision not to make use of transitional provisions for CET1. Specifically, unrealized gains/losses in AFS and equities and reserves arising from revaluation of property are recognized since January 2015.
- Capital components that are no longer eligible for additional Tier 1 and Tier 2 capital components follow CRD IV transitional path.
- CET1 ratio hurdle rate are 4.5 percent of RWAs in stress and 7% in the baseline (BoE stress test). Hurdle rates follow Basel III (including Capital Conservation Buffer under the baseline) (IMF stress test).
- Leverage ratio (3 percent hurdle rate met with Tier 1 capital) using two definitions:
  - (i) the leverage ratio set out by the PRA in SS3/13 (BoE stress test):
    - Tier 1 (end-point definition as set out in CRR).
    - Leverage exposure (CRR delegated act definition).
  - (ii) a Tier 1 (CET1 end-point definition and Additional Tier 1 transitional definition) ratio relative to interest-bearing assets (IMF stress test).

### Reporting format for results (capital)
- Output presentation
  - Minimum stressed CET1, Tier 1, CAR, and leverage ratio by bank. Evolution of stressed CET1 and leverage ratio for aggregate of all seven banks (BoE stress test).
  - Evolution of CET1, Tier 1, CAR, and leverage ratio, for the aggregate banking system and type of bank, i.e., major U.K. international banks and major U.K. domestic banks (IMF stress test).
  - Contribution of key drivers to aggregate net profits and aggregate CET1 capital ratios.
  - Cumulative impairment charges by bank for the U.K. and specific other countries impacted by the scenario (BoE stress test).
  - Number of banks and share of total assets below hurdle rates (IMF stress test).

### Liquidity Stress Testing Matrix (IMF designed stress test undertaken by the BoE on behalf of the FSAP team)

- 1. Institutional perimeter
  - Institutions
    - Number of firms: 10, consisting of seven major banks and building societies, and the 3 largest subsidiaries of foreign investment banks.
    - Selection criteria: The sample firms have been selected to provide 80 percent coverage of total U.K. banking assets as measured by the PRA048 liquidity returns.
  - Market share
    - About 80 percent of banking sector total assets.
  - Data and base date
    - The LCR Liquidity Stress Test is based on the data as of December 31, 2015, received by the PRA as Interim LCR reporting from U.K. firms on an all-currency basis. This PRA return is based on the EU Delegated Act (Commission Delegated Regulation (EU) no 2015/61), which implements LCR in the United Kingdom.
    - The cash flow liquidity stress test relies on the PRA048 return as of January 1, 2016.

- 2. Channels of risk propagation
  - Methodology
    - Basel III measures of liquidity risk—the LCR on three scenarios.
    - Two implied cash flow tests: 5 days (cumulative) and 30 days (noncumulative).
    - A general maturity mismatch analysis by maturity bucket.
    - A single currency analysis based on PRA’s ILG regime.

- 3. Risks and buffers
  - Risks
    - Funding liquidity risk, rollover risk, and market liquidity risk.
  - Buffers
    - HQLA securities assessed at market values net of haircut on a security-by-security basis.

- 4. Tail shocks
  - Size of the shock
    - A range of adverse scenarios:
      - LCR Scenario under standard assumptions calibrated by BCBS.
      - An LCR “U.K. retail stress” scenario. The calibration of this deposit run-off scenario replicates the peak stress during the 2007 Northern Rock run, with run-off rates for retail deposits of up to 15 percent and for corporate deposits of 60 percent, and with liquidity risk from committed but undrawn liquidity facilities of 50 percent.
      - An LCR “U.K. wholesale stress” scenario. This scenario replicates the liquidity stress observed during the global financial crisis. It is characterized by: (i) a freeze of wholesale funding on the interbank market, secured funding market via repo and covered bonds, and the commercial paper market (with run-off rate for operational deposits of 75 percent and for not-fully covered corporate deposits of 100 percent), and (ii) liquidity risk from sizeable margin calls related to secured funding, derivatives and foreign currency funding due to market liquidity shocks, derivative assignments, and unwinds and disruptions in the FX swap market (with rollover of secured funding backed by other than Level 1 and Level 2A assets of up to 0 percent).
    - Implied cash flow assumptions include haircuts of up to 60 percent for securities and bank loans that can be mobilized in repos, no issuance of new unsecured funding and freeze of securitization markets, call-back rates of up to 100 percent, and cash outflows of up to 75 percent.

- 5. Regulatory standards (liquidity)
  - Regulatory standards
    - Counterbalancing capacity above net cash outflows under stress scenario.
    - The PRA’s transitional arrangement for the LCR ratio, which is more front-loaded than that prescribed by the CRR (Art. 460). It is set at 80 percent in October 2015 above the 60 percent threshold under the CRR.

- 6. Reporting format for results (liquidity)
  - Output presentation
    - Changes in average liquidity position and counterbalancing capacity for each scenario.
    - Distribution of banks’ liquidity position for each scenario.
    - Number of banks with counterbalancing capacity below net cash outflows.
    - Banks’ post-shock net liquidity position.

*Source: _cr16163 - 5. Regulatory and market-based standards and parameters*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2016/_cr16163.pdf_