A. Stress Testing under the Financial Sector Assessment Program (FSAP)

1. The aim of the FSAP stress test is to assess the resilience of the banking sector as a whole rather than the capital adequacy of individual institutions. The FSAP approach to stress testing is essentially macroprudential: it focuses on the resilience of the broader financial system to adverse macrofinancial conditions rather than on the resilience of individual banks to specific shocks. The FSAP stress test ensures consistency in macroeconomic scenarios and metrics across firms to facilitate the assessment of the banking system as a whole. The stress test analysis is intended to help country authorities to identify key sources of systemic risk in the banking sector and inform macroprudential policies to enhance its resilience to absorb shocks. The FSAP stress test assesses solvency and liquidity risk and covers key risk types, including credit risk, market risk, sovereign risk, and funding risk.

2. The FSAP stress tests of the U.K. banking system should be seen in conjunction with the analysis undertaken by the BoE. The FSAP stress test scenarios cover the key macrofinancial risks identified in the FSAP’s Risk Assessment Matrix (RAM) for the U.K., with additional scenarios or single-factor shocks included as necessary, and have a balance sheet cut-off date of December 2015. Nevertheless, to facilitate comparison between the 2015 BoE concurrent stress test result (that had a cut-off date of December 2014), the institutional perimeter of the two tests is the same.

3. As with all stress tests, the FSAP stress test results should be interpreted with caution. The FSAP stress test results on the U.K. banking system are based on data submitted by U.K. banks for FSAP purposes at the cut-off date of the stress test, December 2015. Notwithstanding the benefits from the submission of a dedicated data template, the data have not been subject to validation by the IMF. These data are complemented by market-based and publicly available data to support the FSAP team’s calibration of quantitative projections. Despite the FSAP team’s best efforts to build a consistent database, the cleaning, validation, matching, and reconciliation of risk data extracted from multiple data sources, and collected with different purposes and at different frequencies, is a complex exercise. This reflects the difficulties inherent in matching data from different sources. Moreover, major U.K. banks have been changing their business models relative to the crisis period to improve their profitability, enhance resilience, and comply with the new prudential regulatory framework. Structural shifts place further constraints on the reliability of past data to inform forward-looking projections. More generally, stress test scenarios typically replicate historical events or express extreme “tail events” based on an historical distribution, even though it is well known that the nature of crises is to have unanticipated shocks and unexpected interrelationships where the past offers limited guidance. While some nonlinear effects can be captured in stress tests, it is always possible that that unknown patterns emerge, especially if extreme shocks materialize.

B. Financial System Structure

4. 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, defined as the 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.² Excluding derivatives and cross-border exposures of foreign-owned bank branches, the total assets of the U.K. financial system are smaller, at GBP 13 trillion, of which GBP 8 trillion are held by banks.

5. The U.K. is a global financial hub. As of December 2015, there were 359 monetary financial institutions in the U.K., out of which 239 firms were foreign-headquartered. In terms of assets, U.K. banks’ consolidated assets amounted to about GBP 5.7 trillion, whereas foreign banks’ accounted for GBP 2.2 trillion, of which over GBP 1.7 trillion were foreign investment banks. In terms of exposures to the U.K. economy, U.K.-headquartered banks accounted for GBP 2.9 trillion of U.K. aggregate lending, with branches of foreign-headquartered banks contributing GBP 1.1 trillion and subsidiaries of foreign-headquartered banks about GBP 500 billion.

6. U.K.-headquartered banks feature a diverse range of business models and operate in a broad range of international markets. U.K.-headquartered banks can be categorized into three groups. The first two groups include the seven firms that took part in the 2015 concurrent BoE/PRA stress testing exercise, distinguished by geographic footprint: the first group has the major U.K. international banks (HSBC Holdings plc, Barclays plc, the Royal Bank of Scotland Group plc, and Standard Chartered plc), which accounted for GBP 4.1 trillion of aggregate assets as of December 2015; and the second group has the major U.K. domestic banks (Lloyds Banking Group plc, Santander U.K. plc, and Nationwide Building Society), with an aggregate balance sheet of GBP 1.2 trillion. All other U.K.-headquartered banks, including retail and investment banks, as well as building societies, are included in the third group, with aggregate assets of under GBP 300 billion.

7. The FSAP stress test was conducted on a global consolidated group basis,³ (hereafter ‘on a consolidated basis’). Although the Banking Reform Act of 2013 requires banking groups with core deposits⁴ in excess of GBP 25 billion to ring-fence their core activities, this will not be implemented until January 2019. Therefore, for the purpose of the FSAP, the stress test was conducted on a consolidated basis.

C. FSAP Stress Testing Approach

8. The resilience of the U.K. banking system was assessed under the FSAP through both solvency and liquidity stress tests (Figure 1):

View Full Size

United Kingdom: Overview of FSAP Stress Testing

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

• Download Figure
• Download figure as PowerPoint slide

• The FSAP team performed its own TD solvency stress test, based on data submitted by the covered banks using the FSAP team’s data templates as of December 2015, matched with publicly available data for the period 2016–20. This stress test complements the BoE/PRA concurrent stress test conducted in 2015 (and published in December 2015), based on end-2014 data for the period 2015–19. For details on the similarities and key differences of the BoE and FSAP solvency stress tests, see Box 1.

• The FSAP team solvency stress test includes a detailed stress test of the U.K. mortgage book drawing on estimations of outstanding balances for different LTV vintages, using a structural fair-value option Merton-based approach. The macro stress test has been complemented by single-factor sensitivity tests on a wide range of stressed U.K. residential property prices, and a stressed swap curve.

• The FSAP team applied an independent quantitative method to assess selected BU credit risk projections submitted by banks for the 2015 BoE concurrent stress test. The analysis focused 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 stress testing tools provided by the FSAP team to assess the resilience of large U.K. banks to a sudden withdrawal of funding. For consistency with the FSAP solvency stress test, liquidity stress tests were based on end-December 2015 data.

A. Overview

9. The FSAP stress test includes all major U.K. banks, representing over 80 percent of PRA-regulated banks’ lending to the real economy.⁵ All major U.K. banks with total retail deposits equal to or greater than GBP 50 billion, whether on an individual or consolidated basis as of December 2015, are included. The coverage of the FSAP stress test is the same set of banks included in the 2015 BoE concurrent stress test. Specifically, seven banks and building societies are included: 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. This group includes two banks with partial government ownership, namely Lloyds Banking Group plc and the Royal Bank of Scotland Group plc.⁶

10. The stress test is based on granular data submitted by banks on the basis of a data template provided by the FSAP team. The template included granular credit risk information for exposures booked in the internal ratings-based portfolio (IRB) by Basel asset class, sectoral split, currency breakdown, and geography. In addition, the template included the breakdown of the securities portfolio at the security level, the duration of the debt portfolio, detailed information on open positions by market risk factor, the split of RWAs for market risk by category under the standardized approach (STA) and the internal models approach (IMM), projections of RWAs under the baseline and adverse scenarios, and maturity gaps for interest rate risk calculations.

11. Data submitted by banks were combined with publicly available data, drawing on a wide range of sources. To perform the stress testing analysis, the FSAP team matched banks’ data at the cut-off date with a database of publicly available data to construct time series for all relevant variables feeding the stress testing models. Bank-by-bank data were sourced from individual bank Pillar 3 disclosures, Bankscope, SNL, Bloomberg, Datastream, Markit, and the 2015 EU-wide transparency exercise. Aggregate data were drawn mainly from Haver Analytics, Moody’s KMV, International Financial Statistics (IFS), and the World Economic Outlook (WEO).

12. The assessment criteria (“hurdle rates”) include the capital standards under Basel III capital framework, implemented via the Capital Requirements Regulation (CRR) using PRA national discretions, and the PRA leverage framework. The PRA has used certain discretions in the CRR in a prudent way. For instance, the PRA has implemented an end-point definition of CET1 that does not take into account the transitional provisions for CET1, such as the phasing in of deductions.⁷ Also, 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.⁸ Under the baseline scenario, the hurdle rate applied in the FSAP stress test 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. A caveat to the hurdle rate is that it does not include the Global Systemically Important Banks’ (G-SIBs’) higher loss-absorbency requirements that began to be phased in January 1, 2016, with full implementation by January 1, 2019.⁹ The hurdle rate also includes a 3 percent Tier 1 ratio based on the U.K. leverage framework.¹⁰ The U.K. leverage framework uses the CRR end-point definition of Tier 1 capital to calculate the numerator¹¹ with limited recognition of Additional Tier 1 capital and the CRR delegated act definition for the exposure measure.

13. The stress test examined a comprehensive range of credit risk exposures, market risk positions, and funding risk channels.

View Full Size

Geographical composition of exposures

(In Percent)

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Source: Bank of EnglandData are as at end-2015. Geographical exposures are based on residence of immediate counterparty. Exposures includes loans and advances, claims undersale and repurchase agreements, longand short-term debt securities and claims

• Download Figure
• Download figure as PowerPoint slide

• For IRB exposures, a separate credit risk analysis was calibrated for five Basel asset classes and the 15 most material geographies for U.K. banks. This yielded a matrix of 75 credit risk models for IRB exposures. The Basel asset classes covered: retail unsecured exposures, including small- and medium-sized enterprises (SME) credit; retail secured exposures (that is, mortgages); corporate lending (including commercial real estate); lending to institutions; and sovereign and central bank lending. U.K. banks have significant cross-border exposures (see chart). Material geographies for IRB exposures included: the 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; and emerging economies (10 percent), including China, India, and South Africa. The credit stress test included securitization and covered bond exposures.

• The scope of the market risk stress test included all positions exposed to risks stemming from changes in market prices. This includes positions in Held for Trading (HFT), Available for Sale (AFS) and FVO, including sovereign and non-sovereign exposures. Positions in hedge accounting portfolios, economic hedges, and cash flow hedges were excluded. The treatment of sovereign exposures in the banking book follows the Basel III framework. The FSAP team used the IRB approach, which relies on the FSAP team risk assessments of the underlying issuers.

• The funding risk analysis examined shocks arising 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.

B. FSAP Scenarios

14. The FSAP solvency stress test examined two macroeconomic scenarios. The two hypothetical scenarios include a baseline and an adverse scenario.

• The baseline scenario draws from the October 2015 WEO projections for key variables,¹² expanded to generate additional variables that are relevant to project credit risk losses. These include real estate prices (residential and commercial real estate) for the U.K., the U.S., euro area core, and euro area periphery; U.K. Bank rate; U.K. credit growth; and equity prices for the U.K. and the U.S. These variables were projected by adjusting the baseline scenario specified for the BoE’s 2015 concurrent stress test to the WEO core variables and spanning the horizon through 2020.¹³ Country specific variables were complemented by global assumptions’ forecasts from the IMF Global Assumptions (Gas) database, including six-month London Interbank Offered Rate (LIBOR) by major currency, and a range of commodity prices.

• Given the geographic footprint of major U.K. banks, their presence in global financial markets, and their exposure to the U.K. real estate market, the adverse scenario explores the following risks:

  • ∘ A sharp downturn in emerging markets, leading to a substantial dampening of global growth.

  • ∘ A sizeable increase in rates and steepening of the yield curve in the United Kingdom and globally, triggering broad-based abrupt price corrections across financial markets.

  • ∘ Funding risk from rising LIBOR spreads, the dry-up of issuance in money markets, and disruptions in foreign exchange (FX) swap markets.

  • ∘ A large correction in the U.K. property markets. This is a key source of credit risk for banks, as real estate is used as collateral in secured lending.

• More specifically, the adverse scenario examines the impact on U.K. banks from a balance sheet recession in the United Kingdom and financial crises in fragile emerging economies. The scenario assumes accelerated monetary normalization in the United States with a 200 basis points policy rate hike during 2016–17, induced by a stronger private domestic demand-driven macroeconomic expansion than is projected under the baseline. This is accompanied by a drying up of liquidity in money markets and the steepening of the yield curve driven by heightened monetary policy uncertainty, internationally correlated credit risk premium shocks, and duration premium shocks. Furthermore, there is a stock market correction triggered by equity risk premium shocks that spreads to FX markets driven by currency risk premium shocks. Vulnerable emerging economies experience sudden stops associated with a tightening of financial conditions that trigger a broad-based correction of equity markets and sharp currency depreciations. The effect of the global shock in the United Kingdom is amplified by an autonomous domestic demand shock, a confidence loss in property markets, stress in funding markets, and a decompression of the term premium in debt markets. U.K. GDP contracts by 1.6 percent in 2016, and reaches a peak deviation from baseline levels in 2017 at -7.5 percent.

15. Both the FSAP scenario and the BoE’s 2015 scenario have a global focus, but triggers and transmission mechanisms across geographies in the two scenarios differ. While the 2015 BoE scenario focused on a synchronized global downturn and a correction in market risk appetite affecting mainly Asia and the euro area,¹⁴ the FSAP scenario features a disorderly monetary normalization in the United States, triggering a broad-based dislocation in financial markets and spillovers to the most vulnerable emerging market economies.¹⁵ Both scenarios have a major impact on the U.K. economy and share a similar degree of severity as measured by their effect on U.K. real GDP (Figure 2).

View Full Size

United Kingdom: GDP Growth Projections¹

(In percent)

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Sources: BoE, IMF, and IMF staff estimates.¹ The IMF (BoE) scenario constitutes a 2.13 (2.04) standard deviation move in two-year cumulative real GDP growth rate over the first two years of the horizon. The standard deviation has been calculated at the quarterly frequency over 1990–2014.

• Download Figure
• Download figure as PowerPoint slide

16. The persistence of shocks and the channels of distress are also different. The FSAP scenario incorporates a quicker recovery toward steady state. Also, the FSAP scenario features a severe dislocation in money markets and a steepening in the yield curve (Figure 3). The diversity of scenarios allows for capturing different risks and assessing the U.K. banking system’s resilience against a range of possible stresses.

View Full Size

United Kingdom: Comparison of Key Variables in BoE and IMF Adverse Scenario

(In percent)

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Sources: BoE, IMF, and IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

C. The FSAP’s Credit Risk Model Approach

17. Capital requirements are reflected in banks’ regulatory RWAs.¹⁶ After a secular decline in the risk density ratio of major U.K. firms, during which average risk weights declined from 70 percent in the mid-1990s to a trough of 29 percent in 2008:Q3, average risk weights for major U.K. banks reached 36.3 percent in December 2015. Two offsetting effects have been driving the trend in risk weights in recent years: whereas new capital rules have pushed up risk weights, including for market risk, major U.K. banks have been reducing their risky non-core asset portfolios as part of their deleveraging strategy, suggesting the possibility that a real risk reduction has taken place.

18. Credit risk accounts for the largest regulatory capital requirement faced by U.K. banks. 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). By contrast, market risk represented just over 5 percent, with CCR amounting to over 7 percent, and operational risk under 10 percent (Figure 4).

View Full Size

United Kingdom: Credit Risk and Risk Weights for the U.K. Banking System

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Sources: Pillar 3 Disclosures, Bloomberg, and IMF staff calculations.Note: The bottom two charts show asset-weighted averages for the largest five U.K. banks.

• Download Figure
• Download figure as PowerPoint slide

19. The impact of credit risk on banks’ capital ratios depends on the regulatory approach used by banks to book credit exposures. Scenario-based stress testing of credit risk requires mapping the impact of changes in macroeconomic and financial variables onto banks’ loan loss provisions and capital requirements as the level of credit risk rises. Credit risk models are used to project both actual regulatory capital and required regulatory capital (defined as 8 percent of RWA). All new or materially changed IRB capital models require PRA approval. For exposures under the IRB approach, credit risk depends on the exposure at default (EaD), the probability of default (PD), and the loss-given default (LGD). For exposures under the STA approach, risk depends on banks’ loan classification and provisioning requirements set out by the U.K. authorities.

20. The larger U.K. banks rely largely on the advanced IRB approach to book credit exposures. The aggregate credit risk EaD reported in 2014 reached GBP 4.4 trillion for major U.K. banks, of which GBP 3.3 trillion was booked under the IRB approach and GBP 1.1 trillion under the STA approach. Of that GBP 1.1 trillion, GBP 400 billion was exposures to sovereign and central banks, with an asset-weighted average risk weight of under 5 percent.¹⁷

Credit Risk Model for IRB Exposures

21. The impact of stress on regulatory capital through increased provisions was computed as the level of expected losses: $E{L}_{i,t}^j=P{D}_{i,t}^j*LG{D}_{i,t}^j*EA{D}_{i,t}^j$ where i denotes the bank, j denotes the asset class, and t is the time dimension. The FSAP team estimated separate credit risk models by Basel asset class and geography. All material geographies for U.K. banks are covered in the credit risk analysis. Although specialized lending is subject to the slotting approach under the PRA regulatory framework, for FSAP purposes the computation of unexpected losses is treated under the corporate IRB approach.

Approach to Project Probabilities of Default (PDs)

22. The FSAP team used a two-step process to project stressed PDs. The team built a time series of bank-specific PDs by Basel asset class and geography and used it to refine projections from econometric analysis based on aggregate PD proxies. In their Pillar 3 disclosures, banks disclose information on selected portfolios’ PDs calibrated using their regulatory IRB models. As regulatory capital is based on banks’ IRB models, disclosed PDs provide useful information to inform regulatory capital projections. A limitation of this data is, however, that disclosures are available at annual frequency, they have a short history starting in 2008, and do not cover all relevant portfolios and geographies. The approach of the FSAP team was twofold:

• Based on the stress test scenario, a time series of PD proxies was projected for each material asset class and geography using market-based PDs (across all geographies) and write-off data (for U.K. exposures).¹⁸ These series are available at quarterly frequency and cover all material geographies, and therefore are suited to feed the FSAP team satellite models for credit risk.

• An econometric approach was used to forecast bank-specific PDs using banks’ Pillar 3 disclosures as the dependent variable and the projected series of PD proxies as the main driver.

23. For U.K. exposures, the FSAP team applied two different PD proxies to forecast expected losses. The first PD proxy was obtained from Moody’s Analytics using the average one-year expected default frequency (EDF) for the corporate group, the financials group, and the consumer nondurables and services group. These categories were mapped to corporate exposures, exposures to institutions,¹⁹ and retail unsecured exposures, respectively. The PD proxy for exposures to sovereign and central banks was extracted from sovereign yields. The second PD proxy was sourced from the BoE using write-off rates on U.K. exposures, mainly booked in the U.K., by main asset class.²⁰ Although the write-off rate is the closest measure to estimated portfolio loss ratio, write-off rates tend to lag defaults as defaulted exposures, and loan loss provisions remain on the books for up to about 24 months. Figure 5 shows the difference in the loss distribution across EDFs and write-off rates for selected portfolios.²¹ While the write-off rate distribution shows positive skew, the EDF distribution is more neutral, with a higher spike and lower persistence, reflecting shifts to expectations at the time of deteriorating economies’ conditions. Appendix I Table 1 summarizes the key variables used to inform PD projections for U.K. exposures.

View Full Size

United Kingdom: Probability of Default and Credit Loss Rate Across Asset Classes

(In percent)

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Sources: Moody’s KMV, BoE, and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

24. For overseas exposures, the FSAP team applied a market-based PD proxy. The BoE publishes aggregate write-off rates for U.K. banks’ domestic exposures, but similar data for cross-border exposures at the portfolio level are not available. Instead, the FSAP team used Moody’s Analytics one-year EDF for the corporate group, the financials group, the consumer nondurables and services group, and the construction group for Canada, China, France, Germany, Hong Kong SAR, India, Ireland, Italy, Korea, the Netherlands, Singapore, South Africa, Spain, and the U.S.²²

25. An integrated credit risk approach drawing on bilateral and multivariate vector autoregressive (VAR) estimation techniques and nonlinear principal component analysis (PCA) was used to project conditional PDs. The simultaneous behavior of credit risk, macroeconomic conditions, financial conditions, and real estate conditions was modeled explicitly in the econometric specification. In addition, global conditions and spillovers from relevant geographies were included as a factor, given the size and prominence of major U.K. banks in international markets.

26. A battery of over 900 credit risk specifications was run to obtain PD projections. The FSAP team ran a comprehensive set of over 700 bilateral VARs for all pairs of credit risk by PD proxy, asset class, and geography against each variable forecast in the macroeconomic scenario. In addition, 150 PCA analyses were conducted for each factor, PD proxy, asset class, and geography, which fed into 50 multivariate VARs for each PD proxy, asset class, and geography.²³ Appendix I gives the details of the econometric approach used to project credit risk.

27. Credit risk in U.K. exposures is mainly driven by financial conditions. Figure 6 shows that rising spreads across money and equity markets (PC1_fin), and tight corporate credit markets (PC1_fin_c) contribute to rising corporate EDFs with a peak reached in 2017:Q2 at 5.4 percent. Whereas macro conditions are not statistically significant, an abrupt correction in the commercial real estate (CRE) market is also a driver of heightened credit risk. Write-off rates in U.K. mortgages increase with deteriorating financial conditions in secured lending markets (PC1_fin_m). Deteriorating macroeconomic conditions are also relevant to explain losses in banks’ mortgage books, whereas movements in housing prices are not statistically significant.

View Full Size

United Kingdom: Aggregate PD Proxy Projections—Selected U.K. Exposures

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Source: IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

28. The determinants of credit risk in overseas exposures differ across geographies. Figure 7 shows the key PD drivers for selected corporate exposures in France and China. While domestic macroeconomic conditions and real estate prices are drivers of corporate credit risk in France,²⁴ global financial conditions, particularly related to FX shocks and the steepening of the yield curve—as well as developments in oil markets—are the most significant drivers of corporate credit risk in China.

View Full Size

United Kingdom: Aggregate PD Proxy Projections—Selected Cross-Border Exposures

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Source: IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

29. To avoid a structural shift at the cut-off date between banks’ estimated PDs and FSAP team’s PD projections, an econometric approach drawing on banks’ Pillar 3 disclosures was used. Table 1 shows an extract from HSBC Holdings plc’s Pillar 3 disclosures. The choice of HSBC is motivated by the size of the firm’s balance sheet (HSBC Holdings plc accounts for one-third of U.K. major banks’ assets), its broad product base, and its geographic footprint. Notwithstanding caveats related to changes in the composition of the underlying portfolios, the peak of bank estimated PDs stood at 4 percent for European unsecured retail exposures, 15 percent for the U.S. mortgage portfolio,²⁵ and 3 percent for both European mortgage exposures and global corporate exposures. On the other hand, PDs for exposures to institutions (mainly financial) stood at 0.5 percent at the height of the financial crisis, whereas the average PD for sovereign exposures peaked at 0.2 percent in 2008.

Table 1.

United Kingdom: Bank-Estimated PDs for IRB Exposures

Source: HSBC Holdings plc Pillar 3 disclosures over 2008–15 and IMF staff calculations. Note: For 2013–15, the data is for Asia (no split between HK and Rest of Asia).

Table 1.

United Kingdom: Bank-Estimated PDs for IRB Exposures

			2008	2009	2010	2011	2012	2013	2014	2015
Retail (by region)
	Europe
		PD mortgages	1.51	1.67	1.53	1.00	0.92	3.08	0.98	1.43
		PD unsecured	3.91	4.01	3.47	2.57	1.98	2.06	1.65	1.56
	HK1
		PD mortgages	1.15	0.81	0.76	0.75	0.50	3.06	1.00	0.99
		PD unsecured	1.71	1.86	1.52	1.36	0.94	1.14	1.21	1.24
	Rest of Asia
		PD mortgages	2.06	2.05	1.55	1.47	1.37	3.06	1.00	0.99
		PD unsecured	0.50	0.60	0.50	0.50	0.50	1.14	1.21	1.24
	North America
		PD mortgages	10.27	13.39	12.92	14.65	15.01	8.48	11.54	9.66
		PD unsecured	10.48	9.18	6.28	4.90	5.72	3.30	5.37	4.38
Wholesale IRB Foundation
	PD gov and central banks		0.20	0.16	0.11	0.11	0.13	0.17	0.17	0.14
	PD institutions		0.47	0.49	0.36	0.46	0.39	0.46	0.36	0.18
	PD corporates		217	332	282	257	219	232	185	204

Source: HSBC Holdings plc Pillar 3 disclosures over 2008–15 and IMF staff calculations. Note: For 2013–15, the data is for Asia (no split between HK and Rest of Asia).

View Table

Table 1.

United Kingdom: Bank-Estimated PDs for IRB Exposures

			2008	2009	2010	2011	2012	2013	2014	2015
Retail (by region)
	Europe
		PD mortgages	1.51	1.67	1.53	1.00	0.92	3.08	0.98	1.43
		PD unsecured	3.91	4.01	3.47	2.57	1.98	2.06	1.65	1.56
	HK1
		PD mortgages	1.15	0.81	0.76	0.75	0.50	3.06	1.00	0.99
		PD unsecured	1.71	1.86	1.52	1.36	0.94	1.14	1.21	1.24
	Rest of Asia
		PD mortgages	2.06	2.05	1.55	1.47	1.37	3.06	1.00	0.99
		PD unsecured	0.50	0.60	0.50	0.50	0.50	1.14	1.21	1.24
	North America
		PD mortgages	10.27	13.39	12.92	14.65	15.01	8.48	11.54	9.66
		PD unsecured	10.48	9.18	6.28	4.90	5.72	3.30	5.37	4.38
Wholesale IRB Foundation
	PD gov and central banks		0.20	0.16	0.11	0.11	0.13	0.17	0.17	0.14
	PD institutions		0.47	0.49	0.36	0.46	0.39	0.46	0.36	0.18
	PD corporates		217	332	282	257	219	232	185	204

Source: HSBC Holdings plc Pillar 3 disclosures over 2008–15 and IMF staff calculations. Note: For 2013–15, the data is for Asia (no split between HK and Rest of Asia).

30. The forecast path of conditional PD proxies under the macroeconomic scenario is transformed using information embedded in bank estimated PDs, adjusting for point-in-time (PiT) parameters. There are two key reasons PD proxies and bank-specific PDs are likely to differ. First, PD proxies are calculated for the aggregate market portfolio, and hence do not necessarily reflect the underlying risk profile of banks’ portfolios. Second, forecast PD proxies are either market-based (using the Merton approach) or approximate the ex post realization of credit losses (write-off rates). By contrast, bank-estimated PDs are ex ante measures computed using banks’ approved IRB models for the calculation of capital requirements. For consistency with the PRA regulatory regime, the time series of bank-specific PDs by asset class and geography is regressed on the historical series of EDFs and write-off rates. The estimated coefficients were used to forecast bank-specific PDs over the stress test horizon. Figure 8 shows the aggregate PD proxies and bank-estimated PDs for selected U.K. portfolios. 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.

View Full Size

United Kingdom: Aggregate EDFs, Write-Offs, and PDs for U.K. Exposures—Selected Portfolios

(In percent)

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Sources: IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

$P{D}_{i,t}^j=P{D}_{i,t-1}^j+\Delta P{D}_t^j$

where $P{D}_{i,t}^j$ is bank i blended PD for portfolio j (that is, including both defaulted and non-defaulted counterparties) at time t, and $\Delta P{D}_t^j$ is the aggregate PD shift for portfolio j at time t.

Approach to LGD

Calculation of LGD for U.K. mortgages

31. The FSAP team applied a fair-value option approach to compute LGDs for mortgages. The estimated LGD on mortgages has option-like features conditional on the original LTV distribution and housing prices at default. LGD is a highly nonlinear function of the house price at the time of default V_T and the original LTVt ratio (priced at the time of origination t) of the mortgage loan (see chart). This warrants the estimation of the initial distribution of LTV by vintage repriced at stressed house prices rather than relying on original LTV ratios.

View Full Size

LGD and House Prices

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

LGD_T shows the LGD at the time of default T, LTVt reflects the original LTV distribution at time t, and VT denotes the house price at the time of default T.

• Download Figure
• Download figure as PowerPoint slide

32. LGD projections under the stress scenario 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; and

• The forced sales discount on the property’s market price under foreclosure.

33. The distribution of LTVs at origination has significantly improved in the wake of the financial crisis. Drawing on data on residential loans to individuals from the Mortgage Lenders and Administrators Return (MLAR) statistics, Figure 9 shows that, in the tail of the distribution (right bottom chart), only 0.1 percent of regulated mortgages²⁶ granted at end-2015 show LTV ratios over 95 percent relative to over 6 percent in 2007. At the same time, high-rated mortgages with LTV lower than 75 percent (left top chart) now account for over 60 percent of originated mortgages relative to just about 50 percent in 2007.

View Full Size

United Kingdom: Distribution of LTV by Vintage

(In percent)

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Sources: Mortgage Lenders and Administrators Statistics, 2015:Q4 for regulated institutions, and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

34. The volume of loans outstanding from each vintage depends on the principal rollover rate. The rollover rate was estimated for each vintage as the volume of redemptions relative to the volume of loans outstanding each period. Figure 10 (left chart) shows that the principal rollover rate has slowed down since the crisis as initial maturity has lengthened, reflecting higher reliance on fixed mortgages, as well as longer repricing periods for tracker mortgages post-crisis.

View Full Size

United Kingdom: Rollover Rate, Initial Maturity, and U.K. Residential Property Prices

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Sources: Haver Analytics, BoE, and IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

35. The property price haircut is linked to the macroeconomic scenario. The path to the U.K. residential property index draws on the 2015 BoE scenario adjusted for the time horizon under the IMF scenario. BoE’s baseline projections are assumed to realize up to end-2015, and forecast projections are extrapolated in the outer years of the macro scenario. Figure 10 (right chart) shows that the largest year-on-year (yoy) decline in house prices will be achieved at end-2016, and the peak-to-trough decline in prices reaches 20 percent by 2018:Q3.

36. A conservative forced sales discount was set at 30 percent of fair-value U.K. residential property prices. This discount was derived by comparing the realized sale price with the fitted market value for houses sold in foreclosure proceedings. There is wide evidence of price discounts relative to fair-market value in the case of fire sales during crisis periods. Empirical evidence varies across countries, LTV ratios, and home quality. A recent study in European countries found a price discount between 15 percent and 36 percent of the fair-market value.²⁷ The estimated marginal effect for losses at foreclosure in the U.S. post-crisis is 20 pps.²⁸ These loans may be associated with weaker underwriting, higher expenses, weaker markets, and longer time lines. Previous studies on U.S. foreclosure sales had documented fire sales discounts of about 25 percent. In view of the empirical evidence, the FSAP team considered that a 30 percent discount over fair-value prices would be an appropriately conservative estimate.

37. Under the adverse scenario, the average LGD shock was set at 3.5 percent. Figure 11 shows the LGD by vintage (left chart). Interestingly, the pattern is non-monotonic: LGD is higher for loans issued prior to the crisis or issued most recently. The former is driven by high LTV ratios (particularly at the right tail of the distribution), whereas the latter is driven by higher fair-market prices prevalent at the time of the pricing of the collateral. Multiplying the LGD ratio from each vintage by the volume of loans outstanding yields the average LGD shock of 3.5 percent. The right chart shows the estimated outstanding volume per vintage. Notably, the average estimated initial maturity for pre-crisis originated loans ranges between five and seven years, due to the prevalence of five-year fixed-rate mortgages subject to refinancing, yielding no material outstanding volume prior to 2008:Q3.

View Full Size

United Kingdom: LGD Projections by Vintage

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Source: IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

38. For robustness, the estimated LGD shock was cross-checked against banks’ reported LGDs and write-off rates on mortgages in stressed episodes. Drawing on Pillar 3’s disclosures, the maximum one-year increase in IRB LGDs for U.K. mortgages in the post-crisis period is about 3 pps. (Figure 12).²⁹ Using write-off rate data on secured lending to individuals in U.K., the peak was reached in 1997:Q3 and 2009:Q1 at around 0.14 percent. Using an estimated PD for U.K. mortgages of 1 percent, this is equivalent to an LGD of 14 percent, which approximates the average stressed LGD projected by the FSAP team for U.K. mortgages over the stress test horizon.

View Full Size

United Kingdom: Write-Off Rates and LGDs for U.K. Mortgages

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Sources: BoE, HSBC Holdings plc Pillar 3 disclosures, and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Calculation of LGD for Other Asset Classes

39. To derive LGD projections for non-mortgage loans, the FSAP team drew on the macroeconomic scenario, as well as on insights from the global financial crisis and the EA sovereign debt crisis. Adverse LGDs were calibrated for the first two years of the scenario, given the recovery path assumed over the outer three years of the horizon. LGD projections were derived bank by bank, using initial bank-specific LGD ratios and shocks to LGD linked to the scenario:

where $LG{D}_{i,t}^j$ is bank i post-credit risk mitigation LGD for portfolio j at time t, and $\Delta LG{D}_t^j$ is the aggregate LGD shift for portfolio j at time t.

• To compute LGD shocks, information was drawn from banks’ disclosed IRB-based estimates for LGDs during stressed periods. For instance, the average peak LGD in HSBC Holdings plc’s Pillar 3 disclosures for wholesale IRB corporate portfolio is reached in 2011 at 39.2 percent (Table 2), which represents an increase of about 3.6 pps over 2015 LGD levels. LGD floors were applied to central bank and Government exposures in 2013 and to institutions in 2014, and were kept under the adverse scenario.

• Projected increases in corporate LGDs were cross-checked against the drop in nominal GDP forecast under adverse conditions. For retail unsecured exposures, the increase in LGD was matched with rising unemployment by IRB jurisdiction, including a rise of over 2 pps in the U.K. unemployment rate in 2017.

Table 2.

United Kingdom: LGD Calibration for Non-Mortgage Portfolios

(In Percent)

Source: HSBC Holdings plc Pillar 3 disclosures. Note: LGD floors applied to central Government and central banks since 2013 and to institutions since 2014.

Table 2.

United Kingdom: LGD Calibration for Non-Mortgage Portfolios

(In Percent)

LGD Global Portfolios
	2008	2009	2010	2011	2012	2013	2014	2015	max Δ
LGD gov and central banks	20.3	19.9	20.9	20.3	19.6	45.0	45.0	45.1	floor
LGD institutions	29.6	32.5	29.5	32.5	32.1	33.6	42.0	42.0	floor
LGD corporates	37.8	38.9	38.4	39.2	37.8	38.5	36.0	35.6	3.2

Source: HSBC Holdings plc Pillar 3 disclosures. Note: LGD floors applied to central Government and central banks since 2013 and to institutions since 2014.

View Table

Table 2.

United Kingdom: LGD Calibration for Non-Mortgage Portfolios

(In Percent)

LGD Global Portfolios
	2008	2009	2010	2011	2012	2013	2014	2015	max Δ
LGD gov and central banks	20.3	19.9	20.9	20.3	19.6	45.0	45.0	45.1	floor
LGD institutions	29.6	32.5	29.5	32.5	32.1	33.6	42.0	42.0	floor
LGD corporates	37.8	38.9	38.4	39.2	37.8	38.5	36.0	35.6	3.2

Source: HSBC Holdings plc Pillar 3 disclosures. Note: LGD floors applied to central Government and central banks since 2013 and to institutions since 2014.

40. The projection of EAD was driven by balance sheet assumptions, structural FX risk in foreign currency loans, and triggered credit lines and guarantees. Specifically, changes to EAD in the IRB portfolio are governed by:

$EA{D}_{i,t}^j=EAD{J}_{i,t-1}^j\cdot (1+g_{i,t}+f_{i,FX}^j\cdot \Delta F{X}_t)\cdot (1-P{D}_{i,t-1}^j)+\Delta L_{i,t}^j\cdot UC{L}_{i,t-1}^j$

where i denotes the bank, j denotes the asset class, and t is time, g_{i, t} is the growth rate of the IRB portfolio, $f_{i,FX}^j$ is the fraction of foreign currency loans. The FSAP team considers two major currency pairs by bank, namely EUR and USD for GBP-reporting banks; and EUR and GBP for USD-reporting banks. ΔFX_t is the shock to foreign currency under the macro scenario, $(1-P{D}_{i,t-1}^j)$ represents the non-defaulted portfolio, $\Delta L_{i,t}^j$ is the shock to triggered credit lines and guarantees, and $UC{L}_{i,t-1}^j$ is the amount of undrawn guarantees.

41. The FSAP scenario assumed constrained balance sheets by geography. Bank credit supply behavior is constrained through the activation of macroprudential policy tools to avoid credit supply rationing.³⁰ Under the FSAP scenario, credit growth is determined by credit demand shocks triggered by negative consumption and investment shocks that lead to a slowdown of credit under the adverse scenario. Given the geographic footprint of U.K. banks, the growth rate of credit varies across banks. It was computed as the EAD-weighted nominal GDP growth across relevant jurisdictions. Data on FX structural effects from foreign currency loans were provided by firms. When there were data gaps, it was assumed that loans are denominated in the domestic currency of the overseas exposure. To calibrate the shock to triggered credit lines and guarantees, the FSAP team used the maximum increase in the off-balance sheet exposures to EAD ratio for corporates reported by banks in their Pillar 3 disclosures over 2008–15, which is about 40 percent.

42. To compute capital requirements, regulatory risk parameters were considered and the Basel III formula for IRB exposures was applied. The derivation of RWAs is dependent on estimates of PD, LGD, EAD, correlation assumptions, and effective maturity for each exposure. According to the Basel III framework, RWAs were computed after applying the scaling factor of 1.06 to credit RWAs. Also, a multiplier of 1.25 was applied to the correlation parameter of all exposures to large regulated financial institutions and to all unregulated financial institutions.

Credit Risk Model for STA Exposures

43. Expected losses for STA banking book exposures rise due to the impact of migrations from performing to nonperforming loans (NPLs), as well as to migration effects across rating grades within NPLs. The first effect was estimated by running a panel regression of NPLs on key macrofinancial variables. A caveat is that NPL ratios are not disaggregated between IRB and STA exposures. For large international banks, NPLs can be larger among STA exposures if the STA approach is used to book exposures warehoused in overseas subsidiaries subject to higher credit risk. The second effect was calibrated using Basel findings on NPL coverage ratios during the global financial crisis, which stood at about 65 percent on average.

44. To compute the impact of migration on capital requirements, the FSAP team used a two-prong approach:

• The risk weight of NPL was set at an average 130 percent, partly informed by banks’ reported average risk weight of corporate exposures under special management. The difference between 130 percent and the average risk weight of each bank STA exposures (excluding sovereign and central bank exposures) multiplied by the nominal amount of NPLs under the scenario (driven by the balance sheet dynamics and the forecast of the NPL ratio) represents the increase in risk weights attributed to the STA exposures in default.

• The non-defaulting portfolio was assumed to downgrade two notches under the adverse scenario. The asset-weighted risk weight of major U.K. banks’ STA exposures (net of sovereign and central bank claims) stands at about 75 percent, corresponding to a Basel III credit assessment for corporate claims between A+ to A- (50 percent) and BBB+ to BB- (100 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 shows the increase in RWAs due to migration effects.

Table 3.

United Kingdom: EAD and Risk Weights of STA Exposures

(In millions of British Pounds)

Sources: Pillar 3 disclosures and IMF staff calculations. Notes: The figures for HSBC Holdings plc and Standard Chartered plc were originally reported in USD. The risk-weight density for Santander U.K. plc was estimated drawing on inference from Nationwide Building Society.

Table 3.

United Kingdom: EAD and Risk Weights of STA Exposures

(In millions of British Pounds)

		HSBC	Barclays	RBS	SC	Lloyds	San UK	Nationwide
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 gov&cb	12,622	2,828	84	1,180	11	…	0
	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%
Δ RWAs due to Δ NPL		46.0%	60.3%	52.2%	61.6%	60.0%	90.0%	91.8%
Δ RWAs due to rating effects on PL		16.7%	16.7%	16.7%	16.7%	16.7%	16.7%	16.7%

Sources: Pillar 3 disclosures and IMF staff calculations. Notes: The figures for HSBC Holdings plc and Standard Chartered plc were originally reported in USD. The risk-weight density for Santander U.K. plc was estimated drawing on inference from Nationwide Building Society.

View Table

Table 3.

United Kingdom: EAD and Risk Weights of STA Exposures

(In millions of British Pounds)

		HSBC	Barclays	RBS	SC	Lloyds	San UK	Nationwide
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 gov&cb	12,622	2,828	84	1,180	11	…	0
	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%
Δ RWAs due to Δ NPL		46.0%	60.3%	52.2%	61.6%	60.0%	90.0%	91.8%
Δ RWAs due to rating effects on PL		16.7%	16.7%	16.7%	16.7%	16.7%	16.7%	16.7%

Sources: Pillar 3 disclosures and IMF staff calculations. Notes: The figures for HSBC Holdings plc and Standard Chartered plc were originally reported in USD. The risk-weight density for Santander U.K. plc was estimated drawing on inference from Nationwide Building Society.

45. Capital requirements for STA exposures were driven by changes in provisioning rates, growth of EAD, structural FX risk, triggered credit lines and guarantees, and migration effects. The equation below shows four main components driving RWAs in the STA portfolio. The first component reflects the motion of RWAs generated by the flow of provisions, the growth rate of the portfolio, and FX effects. The second component shows the increase in risk weights resulting from triggered off-balance sheet credit lines and guarantees. The third component reflects the increase in risk density from the transition of loans from the performing to nonperforming category. Finally, the fourth component denotes the change in risk density from the transition matrix estimated for performing exposures:

$\begin{array}{l}RW{A}_{i,t}=\left(RW{A}_{i,t-1}-\Delta {\mathrm{Pr}}_{i,t}\right)*\left(1+g_{i,t}+f_{i,EUR}\cdot \Delta F{X}_{EUR}+f_{i,USD}\cdot \Delta F{X}_{USD}\right)+\\ +\left(RW{D}_{i,t-1}^{STA}-RW{D}_{i,t-1}^{OBS}\right)\cdot \mathrm{\Delta }{L}_{i,t}^{STA,OBS}+\left(RW{D}_{i,t-1}^{NPL}-RW{D}_{i,t-1}^{PL}\right)\cdot \mathrm{\Delta }NP{L}_{i,t}+\left(RW{D}_{i,t-1}^s-RW{D}_{i,t-1}^b\right)\cdot P{L}_{i,t}\end{array}$

where Δ Pr_i,t is the increase in provisions in the STA book, g_{i, t} is the growth of STA exposures, f_{i, EUR} (f_i,USD) is the fraction of RWAs in STA exposures denominated in EUR (USD), ΔFX_EUR (ΔFX_USD) is the FX shock to EUR (USD), $RW{D}_{i,t-1}^{STA}$ is the average risk-weight density of on-balance sheet STA exposures, $RW{D}_{i,t-1}^{OBS}$ is the average risk weight density of off-balance sheet STA exposures, $\Delta L_{i,t}^{STA,OBS}$ is the increase in loans triggered by unused credit lines and guarantees, $RW{D}_{i,t-1}^{NPL}$ is the average risk weight density of the nonperforming portfolio, $RW{D}_{i,t-1}^{PL}$ is the average risk-weight density of the performing portfolio, ΔNPL_{i, t} is the increase of NPLs in the STA portfolio, $RW{D}_{i,t-1}^s\left(RW{D}_{i,t-1}^b\right)$ is the average risk-weight density of the performing portfolio under stressed (baseline) conditions due to credit risk migration effects.

D. The FSAP Team’s Approach to Market Risk

46. Market liquidity risks on the grounds of higher fragility of liquidity in some fixed income markets have been recently exposed in several advanced markets. Market liquidity using different measures such as bid–ask spreads, imputed round-trip costs, and Corwin and Schultz’s high-low spreads among others, is not low in most markets when compared to historical averages.³¹ But there are reasons to believe that the current level has been supported by benign cyclical conditions, including accommodative monetary policy in most advanced economies. In addition, the resilience of market liquidity has been challenged by recent structural changes in financial markets, including the increased footprint of and concentration among asset managers, a retrenchment of banks from trading activities and a reduction of their balance sheet space to support market making in credit markets, and the proliferation of smaller issues or issues by unseasoned or higher credit risk issuers.

47. The FSAP solvency stress test explores in some detail market risks from valuation losses in bond markets. The analysis covers shocks to sovereign debt securities in 22 jurisdictions and market stress in 4 non-sovereign corporate bond indices, that is, 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 to cover the whole debt securities portfolio. Exposures arise from an immediate borrower basis. They do not include exposures to other counterparties with government guarantees.

48. Shocks to the securities portfolio are consistent with the macroeconomic scenario and hit banks throughout the five-year stress test horizon (Figure 13). This is a particularly severe assumption, especially for the trading book, as rebalancing of the portfolio is disallowed. Shocks to risk factors impact the fair valuation of securities under both the baseline and the adverse scenario. This contrasts with the methodology applied in the 2016 EU-wide stress test, whereby no changes for the baseline scenario were required. This also differs from the BoE stress test, whereby market risk in the trading book (along with CVA movements, PVA movements, and counterparty credit defaults) was incurred in the first year of the stress test horizon. In the FSAP stress test, the market shock was applied as an instantaneous shock to all the positions covered by the market risk analysis each year of the horizon, with losses fully recognized each year of the stress test:

View Full Size

United Kingdom: Projections of Five-Year Yields for Selected Sovereigns—IMF Adverse Scenario

(In percent)

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Source: IMF staff calculations.Note: The chart shows yield to maturity (YTM) projections for five-year government securities in domestic currency under the baseline scenario (blue bars) and the adverse scenario (red bars).

• Download Figure
• Download figure as PowerPoint slide

• In line with the constrained balance sheet assumption for credit risk, the notional values of the securities portfolio grow according to:

where ${\mathrm{Pr}}_{i,t}^j$ is the level of provisions for asset class j, by bank i, at time t; g_{i, t} is the growth of interest-bearing assets for bank I; f_{i, EUR} (f_{i, USD}) is the fraction of bank’s i portfolio denominated in EUR (USD), and ΔFX_EUR (ΔFX_USD) is the FX shock to EUR (USD). On the other hand, no portfolio rebalancing or liquidation of positions was allowed throughout the stress test horizon.

• The calibration of market risk factors was consistent with the macroeconomic scenario. Changes to baseline sovereign yield to maturity (YTM) rates were extracted from WEO projections for 10-year bond yields. Levels of projected yields were adjusted using Bloomberg generic bond rates for the relevant residual maturity and jurisdiction. Shocks to yields under the adverse scenario were calibrated using the IMF in-house structural model. Adverse shifts to risk factors (including risk-free rates and credit spreads) stem from monetary policy shocks and internationally correlated term premium shocks.

• The model assumes no flight-to-quality effects, with all sovereign yields edging up relative to December 2015 rates across sovereign curves. Compared to the distribution of bond yields observed during the global financial crisis and the EA sovereign debt crisis, the projected adverse shocks look quite severe. The largest yoy hikes in yields under the adverse scenario are for the United States (310 bps), Ireland (205 bps), Hong Kong SAR (192 bps), and the U.K. (168 bps) by 2017. The largest two-year cumulative shocks over 2015–20 range between 133 and 164 bps.

49. Shocks to corporate bonds were calibrated using BoE’s projections for corporate spreads, adjusted for market yields observed in 2015 (Table 4). Under the baseline scenario, a decompression of credit spreads across U.K. investment grade, U.K. high yield, U.S. investment grade, and U.S. high yield corporates was assumed as the credit cycle turns. 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.

Table 4.

United Kingdom: Shocks to Corporate Spreads

(In percent)

Sources: BoE, Bloomberg, and IMF staff calculations.

Table 4.

United Kingdom: Shocks to Corporate Spreads

(In percent)

			Baseline Yields					Adverse Yields
	Dec-15	2016	2017	2018	2019	2020	2016	2017	2018	2019	2020
UK IG	3.26	3.48	3.63	3.78	3.92	4.04	7.80	6.05	4.59	4.33	4.32
UK HY	6.15	6.75	7.21	7.67	8.11	8.56	20.78	12.86	8.48	8.53	8.83
US IG	3.50	4.22	4.94	5.56	6.12	6.37	7.03	7.13	6.56	6.49	6.44
US HY	8.33	9.39	10.21	10.94	11.60	11.95	17.97	14.33	11.94	11.97	12.02

Sources: BoE, Bloomberg, and IMF staff calculations.

View Table

Table 4.

United Kingdom: Shocks to Corporate Spreads

(In percent)

			Baseline Yields					Adverse Yields
	Dec-15	2016	2017	2018	2019	2020	2016	2017	2018	2019	2020
UK IG	3.26	3.48	3.63	3.78	3.92	4.04	7.80	6.05	4.59	4.33	4.32
UK HY	6.15	6.75	7.21	7.67	8.11	8.56	20.78	12.86	8.48	8.53	8.83
US IG	3.50	4.22	4.94	5.56	6.12	6.37	7.03	7.13	6.56	6.49	6.44
US HY	8.33	9.39	10.21	10.94	11.60	11.95	17.97	14.33	11.94	11.97	12.02

Sources: BoE, Bloomberg, and IMF staff calculations.

50. The impact of market risk on HFT, AFS, and FVO was assessed through a full revaluation of securities. The FSAP team applied a granular approach for market risk, which relies on full revaluation of securities to all firms, including those whose market risk capital requirement is lower than 5 percent of the total capital requirement. Banks are not allowed to rebalance their portfolios, and the mitigating effect of hedge accounting portfolios, designed to hedge positions at fair value, is excluded. Also, the offsetting impact from derivatives classified as economic hedges on related positions is excluded. This is a particularly severe assumption, as banks rely on fair-value hedges to hedge interest rate risk on AFS. Also, prudential filters are excluded, which implies that where a position has a prudential filter that eliminates its impact from capital, such position has nevertheless been included.³²

51. Figure 14 shows the distribution of fixed income securities in the banking system by regulatory treatment and asset class. As of December 2015, fixed income securities booked in the AFS portfolio accounted for 67 percent of the aggregate debt securities portfolio for U.K. major banks, with HFT securities representing 26 percent of the portfolio, and held to maturity (HTM) securities accounting for 7 percent. Relative to bank total assets, the average size of the AFS portfolio reached 9.7 percent of total assets, compared to 3.8 percent of assets for the trading book, and 1.0 percent for HTM securities. By asset class, U.K. sovereign securities accounted for 2.6 percent of assets, while other sovereign securities accounted for 8.5 percent and corporate securities for 3.4 percent.

View Full Size

United Kingdom: Distribution of Debt Securities Portfolio

(In percent of Assets)

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Source: IMF staff calculations.Note: Boxplots include the 25^th and 75^th percentiles of the distribution (blue shaded area) and the 10^th and 90^th percentiles (whiskers).

• Download Figure
• Download figure as PowerPoint slide

52. Other risk factors include rates risk, FX risk, and commodity risk. Table 5 reports shocks to other risk factors, including interest rates for currencies, exchange rates for nine major currency pairs showing a sharp appreciation of U.S. dollar in 2016–17 across all currencies, and shocks to commodity prices for metals and fuel plunging by 2017. Shocks to risk factors other than sovereign yields were also linked to the macroeconomic scenario and calculated for every year of the scenario. Other risk factors, including volatility risk, correlation risk, inflation risk, and basis risk, were excluded from the P&L sensitivity of the trading book. Credit counterparty risk and credit valuation adjustments were also excluded to compute expected losses.

Table 5.

United Kingdom: Shocks to Other Risk Factors

(In percent)

Source: IMF staff estimates.

Table 5.

United Kingdom: Shocks to Other Risk Factors

(In percent)

FX shocks (- denotes appreciation of USD relative to FX)
Country	Currency			Baseline						Adverse
		2016	2017	2018	2019	2020		2016	2017	2018	2019	2020
Hong Kong	HKD	−0.09	0.56	0.69	0.51	1.26		−0.43	0.19	0.45	−0.22	−0.19
Canada	CAD	0.49	0.74	0.41	0.67	0.95		−7.04	−7.33	−4.21	−2.46	−2.51
China	CNY	0.18	0.30	0.56	0.96	0.87		−4.65	−4.59	−2.06	−0.48	−0.24
France	EUR	2.33	1.11	0.80	1.17	1.47		−7.28	−6.19	−2.30	0.59	1.25
Singapore	SGD	0.56	0.22	−0.03	0.28	0.55		−5.10	−4.79	−1.37	0.94	1.42
South Africa	ZAR	−0.94	−3.38	−3.14	−3.19	−2.72		−16.26	−15.10	−4.83	2.48	3.89
Korea	KRW	1.33	1.38	1.04	1.31	1.57		−7.29	−7.08	−2.92	−0.60	−0.63
India	INR	−2.06	−3.35	−3.02	−2.71	−1.39		−9.99	−9.85	−4.46	−1.39	−1.43
UK	GBP	1.17	1.65	1.31	1.58	1.85		−7.64	−6.57	−0.08	5.05	6.75
		Shock to U.K. Risk-Free Interest Rate							Shock to Gold
		2016	2017	2018	2019	2020		2016	2017	2018	2019	2020
	baseline	0.33	0.25	0.16	0.10	0.14		−1.52	1.19	1.40	1.70	2.18
	adverse	0.33	−1.16	−0.22	0.99	0.95		−8.65	−14.15	−9.07	−2.89	−0.98
							Shock to Oil
					2016	2017	2018	2019	2020
				baseline	15.55	11.52	6.76	2.75	0.33
				adverse	−10.43	−20.54	−12.56	−1.93	1.33

Source: IMF staff estimates.

View Table

Table 5.

United Kingdom: Shocks to Other Risk Factors

(In percent)

FX shocks (- denotes appreciation of USD relative to FX)
Country	Currency			Baseline						Adverse
		2016	2017	2018	2019	2020		2016	2017	2018	2019	2020
Hong Kong	HKD	−0.09	0.56	0.69	0.51	1.26		−0.43	0.19	0.45	−0.22	−0.19
Canada	CAD	0.49	0.74	0.41	0.67	0.95		−7.04	−7.33	−4.21	−2.46	−2.51
China	CNY	0.18	0.30	0.56	0.96	0.87		−4.65	−4.59	−2.06	−0.48	−0.24
France	EUR	2.33	1.11	0.80	1.17	1.47		−7.28	−6.19	−2.30	0.59	1.25
Singapore	SGD	0.56	0.22	−0.03	0.28	0.55		−5.10	−4.79	−1.37	0.94	1.42
South Africa	ZAR	−0.94	−3.38	−3.14	−3.19	−2.72		−16.26	−15.10	−4.83	2.48	3.89
Korea	KRW	1.33	1.38	1.04	1.31	1.57		−7.29	−7.08	−2.92	−0.60	−0.63
India	INR	−2.06	−3.35	−3.02	−2.71	−1.39		−9.99	−9.85	−4.46	−1.39	−1.43
UK	GBP	1.17	1.65	1.31	1.58	1.85		−7.64	−6.57	−0.08	5.05	6.75
		Shock to U.K. Risk-Free Interest Rate							Shock to Gold
		2016	2017	2018	2019	2020		2016	2017	2018	2019	2020
	baseline	0.33	0.25	0.16	0.10	0.14		−1.52	1.19	1.40	1.70	2.18
	adverse	0.33	−1.16	−0.22	0.99	0.95		−8.65	−14.15	−9.07	−2.89	−0.98
							Shock to Oil
					2016	2017	2018	2019	2020
				baseline	15.55	11.52	6.76	2.75	0.33
				adverse	−10.43	−20.54	−12.56	−1.93	1.33

Source: IMF staff estimates.

53. The impact of traded risk stress test on P&L differentiates between the general interest rate impact and the credit spread impact. The FSAP team calculated a haircut for each fixed income instrument under stressed conditions as the result of multiplying the maximum modified duration for each residual maturity by a change in the yield to maturity at stress test levels. Shocks to market yields stem from (1) the general interest rate impact assumed under the scenario; and (2) the credit spread impact for each individual security. In addition to the impact of market stress on the fair-value of the securities portfolio, the impact of the traded risk scenario was applied by relevant risk factors on banks’ reported net open positions.

54. The calculation of RWAs for market risk drew on firms’ projections for the 2015 BoE stress test. These projections were 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.³³

E. The FSAP Team’s Approach to P&L

55. U.K. banks’ profitability is driven partly by structural factors. While U.K. banks have continued to deleverage their balance sheet in an effort to mitigate underlying risks, profits of major U.K. banks have been notably lower than they were before the crisis.³⁴ Unquestionably, part of the pre-crisis profitability was driven by excessive risk-taking and was therefore unsustainable. To improve profitability, some banks have been exiting businesses with lower returns, including their global investment banking activities and their non-core operations in foreign jurisdictions.³⁵ Also, they have reduced their exposures to other financial institutions through repo lending and securities lending transactions, reducing their interconnectedness relative to 2008. Accumulated charges relating to past misconduct have also weighed on profitability. Structural factors are likely to be reflected in banks’ interest income base, non-interest income, and non-interest expense.

56. Cyclical factors are also key contributors to U.K. banks’ profitability. Until recently, U.K. banks were facing pressures from the sluggish economic recovery and a sustained period of low interest rates. Whereas the impact of low policy rates on major U.K. banks’ net interest margins (NIMs) has not been material, due in part to banks’ management of interest rate risk through hedging practices, banks’ low returns are reflected in market-based indicators with below par price-book value ratios. Most U.K. banks’ shares continue to trade below their book value, indicating investors’ expectations of subdued future profitability. At end-2015 the asset-weighted average price-to-book ratio (P/B) for major U.K. banks stood at 0.8 from a peak of 1.9 in July 2007 (Figure 15). At the same time, five-year credit default spreads (CDS) stood at 69 basis points.

View Full Size

United Kingdom: Banking System—Market-Based Indicators

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Source: Bloomberg.Note: The sample of U.K. banks includes HSBC Holdings plc, Barclays plc, the Royal Bank of Scotland Group plc, Lloyds Banking Group plc, and Standard Chartered plc.

• Download Figure
• Download figure as PowerPoint slide

57. Adverse macroeconomic conditions and stress in money markets were used to project net interest income. Banks that rely on net interest income to generate profits are particularly exposed to low NIMs. The ability of banks to pass-through higher funding costs to consumers is likely to depend on macroeconomic conditions. Under stressed conditions, deteriorating creditworthiness among borrowers is likely to limit the extent of pass-through to keep credit risk at bay. Also, the income base is expected to shrink under stress as performing loans migrate into exposures at default. At the same time, the decline in asset prices and the sharp increase in loan delinquencies are likely to increase charge-offs and loan-loss provisions, reducing banks’ profits and generating solvency concerns among bank creditors. This may lead to a spike in bank funding costs. Banks with unstable funding structures are particularly exposed to credit sensitive investors.

58. Net interest income is driven by cyclical shocks to funding costs, including dislocations in money markets, shocks to NIMs, interest income from securities, and interest rate risk in the banking book. The FSAP team’s approach to net interest income is summarized here:

• Funding costs were proxied by changes to deposit rates that are determined by macrofinancial conditions (that is, policy rates and LIBOR rates), as well as by firm specific conditions, including bank asset quality, bank capital adequacy, and contagion effects from funding stress in peer U.K. banks. As the capital adequacy of the U.K. banking system improves over the outer years of the scenario, bank funding costs edge down. An additional add-on funding cost reflecting dislocations in money markets was calibrated to the funding stress observed during the Eurozone crisis in H1 2012.

• NIMs were projected on the basis of the policy rate, money market rates, and the term structure assumed under the macroeconomic scenario. Given banks’ traditional role of maturity transformation, when the yield curve steepens, banks’ NIMs are expected to rise. Conversely, when the yield curve flattens, banks’ NIMs are likely to fall. Yet, if the yield curve steepens during crisis periods due to a spike in the term premium, as discussed earlier, banks may be reluctant to increase lending rates to rein in credit risk.

• Interest income that banks earn on other interest-earning assets, including assets held for trading purposes and liquid buffers, was projected by regulatory book.

• The impact of shocks to short-term interest rate risk in the banking book was computed using a gap risk approach to capture the risk arising from the timing on instruments’ rate changes as bank assets and liabilities turn over or reprice at different times.³⁶ Although banks can hedge their interest rate risk or alter their operations in other ways so that interest rate changes may have little effect on overall bank profitability, bank hedges were excluded for stress test calculations.

59. The key driver of deposit rates is the money market rate. Deposit rates were projected at the bank level as:

where D_{i, 2015} is the implicit deposit rate of bank i at the cut-off date December 2015 and ${\displaystyle \sum _{s=1}^{T-t}\Delta D_{t+s}}$ is the sum of accumulated funding shocks throughout the stress test horizon. Banks’ initial funding costs differ across U.K. banks owing to their different funding models, liquidity management, and bank soundness. Changes to deposit rates are determined by two key drivers. First, changes to LIBOR 6m rates, with elasticity of 0.5.³⁷ Figure 16 plots the projection of deposit rates with their confidence interval. Second, by bank aggregate capital adequacy reflecting the linkages between solvency and funding risk (see Section F).

View Full Size

United Kingdom: Projected Changes to Deposit Rates

(In basis points)

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Source: IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

60. NIMs were regressed on money market rates and the slope of the yield curve. NIMs were projected at the bank level as:

where NIM_{i, 2015} is bank i’s reported NIM in December 2015 and ${\displaystyle \sum _{s=1}^{T-t}\Delta NI{M}_{t+s}}$ is the sum of accumulated NIM shocks. There is cross-sectional variation of NIMs across banks. The variation in business models across the U.K. banks is the biggest driver. The weighted average NIM of U.K. major banks was around 2.8 percent in December 2015. Changes to NIMs are explained mainly by changes to LIBOR 6m rates (with elasticity 0.2) and the slope of the yield curve (with elasticity -0.1).³⁸ Again, both series are found to be non-stationary I (1), but there is no cointegrating vector. Figure 17 plots the projection of NIMs with their confidence interval.

View Full Size

United Kingdom: Projected Changes to NIM

(In basis points)

Citation: IMF Staff Country Reports 2016, 163; 10.5089/9781484394120.002.A001

Source: IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

61. Lending rates are driven by shocks to funding costs and shocks to NIMs.

During the first year of the adverse scenario, lending rates were projected to outpace baseline projections due to a rise in funding costs and increase in NIMs. This effect is mitigated in the second year of stress, as NIMs compress due to borrowers’ creditworthiness downgrade, reflected in changes to the slope of the yield curve (Table 6). Interest accrued on defaulted loans is not recognized under the baseline and adverse scenario.³⁹

Table 6.

United Kingdom: Nominal Shocks to Rates

(In basis points)

Source: IMF staff estimates. Note: The chart shows nominal shocks to rates, in the form of deviations from their end-2015 value.

Table 6.

United Kingdom: Nominal Shocks to Rates

(In basis points)

	2016	2017	2018	2019	2020
ΔDt-2015
Baseline	14	16	13	12	8
Adverse	51	17	42	51	44
ΔNIMt-2015
Baseline	3	4	3	2	0
Adverse	15	−1	13	17	13
ΔLt-2015
Baseline	17	20	16	14	8
Adverse	67	16	55	68	58

Source: IMF staff estimates. Note: The chart shows nominal shocks to rates, in the form of deviations from their end-2015 value.

View Table

Table 6.

United Kingdom: Nominal Shocks to Rates

(In basis points)

	2016	2017	2018	2019	2020
ΔDt-2015
Baseline	14	16	13	12	8
Adverse	51	17	42	51	44
ΔNIMt-2015
Baseline	3	4	3	2	0
Adverse	15	−1	13	17	13
ΔLt-2015
Baseline	17	20	16	14	8
Adverse	67	16	55	68	58

Source: IMF staff estimates. Note: The chart shows nominal shocks to rates, in the form of deviations from their end-2015 value.

62. To compute interest income from the securities portfolio, banks’ implicit interest rate was calculated. Using banks’ reported information, yields were computed for the HFT, AFS, and HTM regulatory books by asset class, that is, own sovereign bonds, foreign sovereign bonds, corporate bonds, and other securities. The implicit interest rate was projected forward, assuming no changes in portfolio composition during the stress test period.⁴⁰ This is a conservative assumption, given the extent to which the low interest rate environment prevailing at end-2015 was pushing down market yields.

63. Interest rate in the banking book is driven by policy rate shocks and banks’ repricing gaps in assets and liabilities. Interest rate gaps by time to repricing across six maturity buckets were computed for each bank, that is, <1m, 1m-2m, 2m-3m, 3m-6m, 6m-12m, and >12m, and net income is calculated as:

where $ga{p}_{i,t}^b$ is the gap of bank i in bucket b and time t, mid^b is the mid-point in bucket b (in days), and $\Delta i_t^f$ is the shock to risk free rates at t.

64. The bank tax rate is driven by the macroeconomic scenario. Econometric analysis suggests that the effective bank tax rate fluctuates with changes in the macroeconomic environment, in particular with changes to GDP growth and money market rates.⁴¹ The projected tax rate declines by 4 pps in 2017 under the adverse scenario (relative to baseline) before picking up by 2019, driven by the quick recovery of the U.K. real economy.

65. Other P&L items were projected drawing on the 2016 EU-wide stress test methodology:

• The non-interest income ratio to total assets (which includes dividend income and net fee and commission income) was kept constant at its 2015 value under the baseline. Under the adverse scenario, the minimum between the 2015 ratio and the average of the two years with the smallest value over the past five years $r_i^{\min }$ was considered as the projected ratio. The weighted average ratio for U.K. banks in December 2015 was 0.82 percent of total assets, down from 0.91 percent in June 2015. The growth rate of non-interest income $g_{i,t}^{non-ii}$ was thus adjusted on the basis of the projected balance sheet growth $g_t^{\mathit{bs}}$ and the lower bound ratio $r_i^{\min }$ according to:

• Non-interest expenses, including administrative expenses and operating expenses, were kept constant relative to total assets under the baseline scenario. This constraint is binding for four U.K. banks in 2016. Under the adverse scenario, their value could not be below the value reported at 2015. This is a conservative assumption. The 2015 value for non-interest expense contains material misconduct costs and, in addition, a number of U.K. banks incurred significant restructuring costs during 2015. This constraint is binding for all U.K. banks over 2016–17.

66. The dividend payout rule was fixed throughout the stress testing horizon. The FSAP team considered three options to set payout ratio projections:

• Econometric approach: A panel of bank payout ratios was regressed on key macrofinancial variables projected under the macro scenario. Econometric results suggest that, on average, the dividend payout ratio increases when the U.K. FTSEE index softens and LIBOR rates increase.⁴² This suggests that U.K. banks might be targeting a ROE ratio that calls for an increase of the payout ratio when net profits are eroded under stressed conditions. Under this approach, the payout ratio would peak at about 80 percent in 2017 in the adverse scenario before converging to baseline levels at about 50 percent by 2020. However, the feasibility of this approach is limited by the Capital Requirement Directive (CRD) restrictions on dividend payments under thin capital buffers, and banks’ declared dividend policies under stress.

• The 2016 EU-wide stress test approach: Under the 2016 EU-wide stress test methodology, if no publicly declared dividend policy is available, the payout ratio is the maximum of 30 percent and the median of the observed payout ratio in profitable years over the past five years. This rule has limited applicability in the U.K., as the Royal Bank of Scotland plc has not paid a dividend on its ordinary shares since it received a capital injection from the U.K. Government in 2008.⁴³ While in May 2015, Lloyds Banking Group plc paid its first dividend on its ordinary shares since its capital injection from the U.K. Government. Also, Nationwide Building Society is a mutual society that operates as a member-owned business model and profits are typically returned to members through better savings, loan, or mortgage rates.⁴⁴ The implementation of this approach would generate substantial variability in dividend policies across banks (with 17 percent standard deviation in payout ratios).

• A fixed dividend payout rule of 30 percent: This rule reflects the lower bound of the 2016 EU-wide stress test methodology, and ensures a level playing field for all U.K. banks.

F. The FSAP Team’s Approach to Funding Costs

67. While U.K. banks have shifted their funding mix away from wholesale funding sources toward deposits, a market liquidity shock could lead to strains in funding markets. After the crisis, banks have shown more stable funding structures. Major U.K. banks’ funding from customer deposits has increased by about GBP 250 billion since 2008, while wholesale funding has declined by over GBP 1.3 billion over the same period. In December 2015, the asset-weighted average loan-to-deposit (LTD) ratio stood at 92 percent. However, a generalized sell-off in fixed-income securities spurred by uncertainty over bank solvency valuation could put pressure on funding schemes as risk sentiment among investors’ turns, leading to escalating bank funding costs and amplified stress across markets.

68. The FSAP team used an econometric approach to drill down on the key drivers of U.K. banks’ funding costs. The risk factors examined fall in three main categories (Appendix II, Table 1): (1) bank-specific variables, including asset quality, leverage, regulatory capital, funding structure, business model, and earning capacity; (2) country-specific variables, including macro variables, real estate prices, equity prices, corporate spreads, and credit growth; and (3) global variables, including world GDP growth, commodity prices, volatility index (VIX), U.S. equity prices, and emerging market exchange rates.

69. The analysis incorporates explicitly contagion from peer banks’ funding pressures. For the same set of fundamentals, contagion can occur if funding stress in one U.K. bank is a signal to investors that other banks in the same U.K. banking system are likely to be in financial trouble. Contagion can result in the restriction of liquidity to other U.K. banks as possible counterparties shy away. To capture contagion, a two-prong approach was followed:

• For each U.K. bank, a peer group was defined as the U.K. banking system, excluding each bank in turn. The average funding costs for the peer group was regressed against the set of explanatory variables for the individual U.K. bank, that is, the U.K. bank-specific variables, country-specific variables, and global variables.

• The orthogonal residuals of the aforementioned regression were identified as a proxy of contagion from funding pressures in other U.K. banks.⁴⁵ Notably, the value of the contagion variable differs across banks.

70. The main model defines funding costs as the implicit interest rate paid in interest-bearing liabilities, and uses a panel model approach. A key challenge was to identify a proxy for bank funding costs. The key reference variable used for the main model is effective interest paid on interest-bearing liabilities. The effective interest rate reflects the P&L impact of funding stress, taking into account banks’ funding structure. Alternatively, bank funding costs can be proxied by five-year senior single name CDS spreads. While this is a reasonable proxy for unsecured term wholesale funding costs, as CDS market liquidity on referenced major U.K. banks’ is usually high, CDS liquidity for specific banks (for example, Nationwide Building Society) and over stressed periods might be limited. Also, it is unlikely to be representative of U.K. bank funding costs, as their main funding base is retail deposits.⁴⁶ For robustness, the analysis was replicated using five-year CDS as a proxy for funding costs. The data was sourced from Bloomberg over 2000:Q1 through 2015:Q2.⁴⁷ The econometric analysis was based on a panel model with fixed effects and robust standard errors estimates. Interest payments were computed on a biannual basis.

71. The results of regressing U.K. banks’ funding costs on a broad range of determinants suggest that (Table 7):

Table 7.

United Kingdom—Determinants of U.K. Banks’ Funding Costs

Robust standard errors *** p<0.01, ** p<0.05, * p<0.1 Estimation period: 2000:Q1–2015:Q2

Table 7.

United Kingdom—Determinants of U.K. Banks’ Funding Costs

Dependent Variable	Interest	Interest	Interest	Interest	Interest	Interest	Interest	Interest	CDS	CDS	CDS	CDS	CDS	CDS
Regressors	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)
ctl	−0.065***	0.067***	0.005	0.004	−0.004	0.001	−0.006	−0.002	0.202	0.184	9.805**	6.525	5.260	4.666
	(0.012)	(0.013)	(0.015)	(0.016)	(0.017)	(0.017)	(0.017)	(0.017)	(2.237)	(2.274)	(3.452)	(4.165)	(4.528)	(4.841)
lev					−0.010	−0.009	−0.011	−0.010			−0.710	−0.857	−0.578	−0.567
					(0.010)	(0.009)	(0.010)	(0.010)			(0.596)	(0.581)	(0.699)	(0.685)
npl	−0.014	−0.013	−0.003	−0.003	−0.000	−0.000	−0.000	−0.001	11.558**	11.356*	7.099**	7.255**	6.463**	6.524**
	(0.010)	(0.010)	(0.016)	(0.016)	(0.013)	(0.014)	(0.014)	(0.015)	(3.520)	(3.709)	(2.359)	(1.904)	(1.919)	(1.789)
stb_tf	0.001	0.001	−0.002	−0.002	−0.003	−0.003	−0.003	−0.003	0.036	0.037	0.649	0.858	0.303	0.386
	(0.004)	(0.004)	(0.003)	(0.003)	(0.002)	(0.002)	(0.002)	(0.002)	(0.215)	(0.213)	(0.471)	(0.536)	(0.574)	(0.575)
unem	−0.084*	−0.096	−0.241***	−0.223**	0.083**	0.128*	0.068*	0.123*	33.951*	27.702*	39.346**	26.736*	39.349**	26.197*
	(0.040)	(0.049)	(0.056)	(0.061)	(0.031)	(0.051)	(0.029)	(0.051)	(11.864)	(10.049)	(9.764)	(10.678)	(10.144)	(10.839)
housing	−0.003	−0.002	−0.008	−0.010	−0.005	−0.008*	−0.003	−0.006	0.383	0.488	−2.482**	−1.849**	−2.184**	−1.427*
	(0.004)	(0.003)	(0.006)	(0.006)	(0.004)	(0.003)	(0.003)	(0.003)	(0.395)	(0.363)	(0.724)	(0.572)	(0.741)	(0.603)
equity	−0.001	−0.000	0.001	0.001	−0.001	−0.002*	−0.001	−0.002*	−1.302	−1.020	−0.649**	−0.415	−0.628*	−0.384
	(0.001)	(0.001)	(0.001)	(0.002)	(0.001)	(0.001)	(0.001)	(0.001)	(0.561)	(0.496)	(0.251)	(0.249)	(0.288)	(0.268)
vix	−0.003	−0.001	0.006*	0.004	−0.001	−0.006	−0.003	−0.008*	−0.343	0.468	0.175	1.344***	0.047	1.392**
	(0.003)	(0.003)	(0.002)	(0.003)	(0.003)	(0.004)	(0.003)	(0.004)	(0.298)	(0.484)	(0.214)	(0.323)	(0.227)	(0.404)
pol					0.220***	0.230***					13.108	7.287	8.195	4.722
				(0.033)	(0.033)						(7.788)	(7.657)	(6.431)	(7.204)
libor							0.204***	0.216***
							(0.030)	(0.031)
oil		0.001		−0.001		−0.003		−0.003		0.492**		0.715***		0.699***
		(0.001)		(0.002)		(0.002)		(0.002)		(0.152)		(0.146)		(0.171)
cont_spec j			0.950***	0.947***	0.458	0.368	0.563	0.493	0.603*	0.563			0.680**	0.635**
			(0.209)	(0.212)	(0.258)	(0.203)	(0.285)	(0.249)	(0.250)	(0.256)			(0.188)	(0.209)
Constant	2.103***	2.056***	2.217***	2.249***	0.075	0.009	0.223	0.148	−168.313*	185.252*	−305.483**	−258.988*	−239.077*	−227.685
	(0.342)	(0.316)	(0.367)	(0.343)	(0.367)	(0.374)	(0.346)	(0.358)	(51.810)	(56.581)	(104.911)	(113.107)	(110.619)	(117.599)
Fixed effects	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes
Observations	292	292	292	292	292	292	292	292	172	172	276	276	273	273
R-squared	0.506	0.509	0.691	0.692	0.731	0.735	0.724	0.727	0.849	0.850	0.727	0.751	0.800	0.801
Banks	6	6	6	6	6	6	6	6	4	4	6