Euro Area Policies: Financial Sector Assessment Program-Technical Note-Stress Testing the Banking Sector

International Monetary Fund. Monetary and Capital Markets Department

doi:10.5089/9781484369401.002.A001

Jump to Content Jump to Main Navigation

Euro Area Policies

Financial Sector Assessment Program-Technical Note-Stress Testing the Banking Sector

Author: International Monetary Fund. Monetary and Capital Markets Department

Publication Date:: 19 Jul 2018

eISBN:: 9781484369401

Language:: English

Keywords:: ISCR; CR; central bank; interest rate; bank deposit

Download PDF (6.7 MB)

The FSAP team undertook a thorough top-down stress testing analysis using end-2017 data. This note covers the methodology and results of the scenario-based solvency tests, the single factor sensitivity tests, and the liquidity tests. The stress test exercise was carried out on a sample of major euro area banks supervised by the Single Supervisory Mechanism (SSM). The analysis is heavily dependent on comprehensive and granular supervisory data on individual banks’ positions shared by the European Central Bank (ECB). While FSAP results are not directly comparable to the 2018 EU-wide stress test results due to differences in scenarios, methodologies, and objectives, they provide an assessment of the system-wide resilience of the euro area banking sector at the current juncture.

Abstract

Executive Summary

The FSAP team undertook a thorough top-down stress testing analysis using end-2017 data. This note covers the methodology and results of the scenario-based solvency tests, the single factor sensitivity tests, and the liquidity tests. The stress test exercise was carried out on a sample of major euro area banks supervised by the Single Supervisory Mechanism (SSM). The analysis is heavily dependent on comprehensive and granular supervisory data on individual banks’ positions shared by the European Central Bank (ECB). While FSAP results are not directly comparable to the 2018 EU-wide stress test results due to differences in scenarios, methodologies, and objectives, they provide an assessment of the system-wide resilience of the euro area banking sector at the current juncture.

Overall the solvency of large euro area banks has improved, but important and heterogeneous vulnerabilities are exposed in stress scenarios. In the test, adverse conditions reduce the aggregate common equity tier 1 (CET1) ratio from 13.9 percent at the end of 2017 to a low point of 10.0 percent in 2020. Comparable results are obtained using alternative solvency metrics. Aggregate capital buffers are sizeable relative to immediate threats, but some banks are especially vulnerable to credit risk while others are more exposed to market risks from risk premia decompression, basis risk, and valuation shocks. Strains related to business risk from lower-than-expected volumes or interest margins are also relevant for some banks.

Results suggest that large, but less complex, internationally-active banks are more robust to a deterioration in economic and financial conditions. Banks that have succeeded in diversifying credit risk and income sources, and cut their exposure to complex securities are more equipped to absorb adverse macrofinancial conditions. Because G-SIBs’ business models generally emphasize trading and capital markets-related activities, they are disproportionately affected by market and funding liquidity risks, especially in foreign currencies. In contrast, domestically-oriented banks are more vulnerable to deteriorations in domestic macroeconomic conditions, and in some instances, also to market-wide liquidity shocks.

The resilience of the system to liquidity shocks has increased in recent years, yet multiple banks may face liquidity challenges in extreme scenarios, assuming only marketable securities are available. The EuroSystem’s extraordinary operations have changed the composition of banks liabilities and the structure of their counterbalancing capacity (CBC) by increasing banks’ reliance on ECB support. At the same time, banks differ notably in terms of funding structures, dependence on foreign currency funding, asset encumbrance, and the amount and composition of CBC. Due to the G-SIBs’ funding structure and trading activities, they are more vulnerable to market runs, but they generally have sufficient CBC to cope with funding liquidity shocks; at the same time some banks may face challenges in replacing central bank funding with market alternatives.

The analysis suggests that solvency and liquidity risk may feed on each other. Sensitivity tests show that an abrupt and sharp increase in rates can have a material impact on bank capital, as funding costs outpace the repricing of assets and valuation effects erode buffers, particularly when fair value hedges are ineffective. Banks that are more dependent on central bank refinancing operations and those with lower quality collateral are most at risk of facing higher funding costs. Solvency-funding loops are exacerbated by non-linear interactions among credit, liquidity, and market shocks. Funding outflows triggered by solvency concerns are likely to result in some banks breaching their capital requirements and facing negative CBC positions.

The ECB’s stress testing analytical framework—already very sophisticated—could be enhanced along several dimensions. Further integrating supervisory data, especially related to market risk and cash flow based liquidity risks, into the existing stress testing infrastructure would facilitate higher frequency and more comprehensive risk monitoring. The analysis of risk interactions could also be deepened to examine their role in aggravating systemic risk. To the extent that the failure of a significant subsidiary could cause major disruption of the euro area financial system, and that sufficient granular supervisory data is available, ECB’s macroprudential stress testing framework should be expanded to include them.

The initiative to collect cash-flow data for all significant currencies is a welcome step; more data are needed to monitor market, funding, and contingent liquidity risks. Cash flow data needs to be collected by significant jurisdictions to help supervisors monitor the availability and transferability of collateral. Stepping-up efforts on data collection on securities financing transactions and derivatives would enhance the monitoring and supervision of market and contingent liquidity risks, including those associated with the flow of collateral via central clearing counterparties (CCPs) and cyber risks.

Table 1.

Euro Area: Main Recommendations on Stress Testing

^{^*}I (immediate) = within one year; ST (short term) = 1–2 years; MT (medium term) = 3–5 years

Table 1.

Euro Area: Main Recommendations on Stress Testing

Recommendations	Timing^*
Solvency Stress Testing
Make further use of supervisory data for stress testing, enhancing analysis, and feeding back into supervision	ST
Deepen analysis of banking group risks by extending the perimeter of macroprudential stress tests to cover systemic subsidiaries of EA significant institutions and significant investment firms.	MT
Continue enhancing the assessment of bank’s risk interactions across various risk categories (i.e. credit risk, market risk, basis risk, liquidity risk).	MT
Liquidity Stress Testing
Use cash-flow data for all significant currencies and significant jurisdictions to refine liquidity monitoring and stress testing	I
Enhance collection of data on, and analysis and supervision of contingent liquidity risks associated with the cross-border flow of collateral, securities funding transactions, and cyber risks	ST

^{^*}I (immediate) = within one year; ST (short term) = 1–2 years; MT (medium term) = 3–5 years

Table 1.

Euro Area: Main Recommendations on Stress Testing

Recommendations	Timing^*
Solvency Stress Testing
Make further use of supervisory data for stress testing, enhancing analysis, and feeding back into supervision	ST
Deepen analysis of banking group risks by extending the perimeter of macroprudential stress tests to cover systemic subsidiaries of EA significant institutions and significant investment firms.	MT
Continue enhancing the assessment of bank’s risk interactions across various risk categories (i.e. credit risk, market risk, basis risk, liquidity risk).	MT
Liquidity Stress Testing
Use cash-flow data for all significant currencies and significant jurisdictions to refine liquidity monitoring and stress testing	I
Enhance collection of data on, and analysis and supervision of contingent liquidity risks associated with the cross-border flow of collateral, securities funding transactions, and cyber risks	ST

^{^*}I (immediate) = within one year; ST (short term) = 1–2 years; MT (medium term) = 3–5 years

Introduction¹

A. A Complex Financial System

1. The euro area is home to a large, complex, and globally interconnected financial system. Of the thirty global systemically important banks (G-SIBs), seven are located in the euro area: BNP Paribas (FR), Deutsche Bank (DE), Credit Agricole (FR), ING (NL), Santander (ES), Societe General (FR), and UniCredit (IT). The interbank and cross-border connections of the euro area banking system are extensive, although they have contracted since the global crisis.

2. Euro area banks represent around two-thirds of total banking system assets in Europe. Banks are the most important financial intermediaries for households, nonfinancial corporates (NFC), and the public sector. Since the global financial crisis, nonbanks and financial markets are playing a greater role, particularly in funding larger nonfinancial firms.

3. In 2017, total assets of the euro area banking sector, including foreign subsidiaries and branches, stood at around €35 trillion on a consolidated basis. The financial crisis ended a period of strong growth in banking sector assets in many euro area countries. This adjustment occurred mainly through a reduction in business volumes and some progress in consolidation, particularly in countries more affected by the crisis. Large euro area banks have also become more selective in their international banking activities and have become more focused geographically in their core markets. In addition, banks have shifted towards safer and less complex assets, notwithstanding some legacy portfolios, and towards more stable funding sources.

B. Banking Sector Trends

4. Banks have benefited from a benign funding environment with stable or improving market access in recent years but this trend might reverse. Of the top euro area banks, the number of banks carrying a positive outlook outweighs the number of negative outlooks on the back of improving economic fundamentals and balance sheet strengthening.² Yet, in some cases, banks’ current outlook depends on further progress on profitability and reduction of nonperforming assets. Investors’ perception of bank risk has lagged sovereigns although it remains more volatile than corporates.³

5. Euro area bank net interest margins have been resilient in recent years. This has been supported by accommodative financial conditions which have eased funding costs helping to offset falling interest income. However, the flat yield curve has enhanced banks’ reliance on commercial margins to secure interest income which might narrow due to heightened competition, including from nonbank sources.

6. Although improving macroeconomic fundamentals should support headline income, performance might remain subdued for some banks. While asset quality has generally improved and therefore loan impairment charges have been contained on aggregate, there is business risk associated with the management of legacy assets, nonperforming loan (NPL) strategies including NPL provisioning and write-offs, fluctuations in client flows, repricing of risk in financial markets, and ongoing restructuring costs in some banks.

C. Supervisory Stress Tests in the Euro Area

7. The resilience of the euro area banking sector has recently been examined by the 2016 EBA stress test and the 2017 ECB Sensitivity Analysis of the Interest Rate Risk in the Banking Book (IRRBB). Whereas the 2016 EBA stress test was a scenario-based stress test, the 2017 IRRBB sensitivity test focused on the impact of shocks to interest rates on banks’ net interest income (NII) and economic value of equity (EVE). The 2016 EBA stress test found that the EU banking sector is resilient to shocks with a capital depletion of 380 bps on average. The 2017 IRRBB test concluded that while interest rate risk is adequately managed at most of the major banks, IRRBB is a risk difficult to assess given the reliance on modelling assumptions. Results point at the increasing trade-off between the NII and EVE approach.⁴ EBA launched its 2018 EU-wide stress test on January 2018, and results are expected to be published by November 2018.

8. While the top-down FSAP stress tests are complementary to the bottom-up stress tests coordinated by EBA, they differ in crucial ways (Box 1). While both EBA and FSAP tests provide a quantitative assessment of banks’ risk profiles across various risk categories for a large sample of large significant institutions. The approach of the EBA exercise is that of a constrained bottom-up stress test where banks are required to project the impact of the scenarios on their projected capital position and P&L but subject to strict constraints defined in the common methodology. By contrast, the FSAP stress test is a top-down exercise with projections generated by in-house models developed by the FSAP team. While the FSAP and EBA scenarios broadly share a consistent narrative of risks, the FSAP excludes conduct risk and other operational risks unlike the EBA test. They also differ with regard to the coverage of geographies and variables, and more importantly to the calibration of shocks. Comparing stress test results (once published) will provide an indication of sensitivities, non-linear effects, and interactions among risk factors.

Box 1.

The 2018 EU-Wide Stress Test and the FSAP Solvency Stress Tests

The 2018 EU-wide stress test and the FSAP solvency stress tests share key similarities:

Both adverse scenarios assume the materialization of systemic risks leading to a balance sheet recession in the euro area, with the FSAP scenario featuring long-lived shocks (U-shape) and the EBA scenario incorporating deep (V-shape) shocks. The implied real GDP growth rate in the euro area amounts to -1.6 percent (-2.2 percent) at year t+2 under the FSAP (EU-wide) stress test by 2019. While the recession is more pronounced under the EU-wide test, it’s more protracted under the FSAP scenario with a -0.3 percent growth rate posted by 2020. Both tests include a traded risk scenario, with paths for financial variables linked to the macroeconomic scenario.
Both tests incorporate a high degree of granularity to capture credit risk from domestic and foreign exposures, and valuation impact from sovereign exposures at the issuer level. Credit risk from exposures in 37 jurisdictions under IRB-A, IRB-F, and STA approaches are modeled separately to project loan impairment charges and capital requirements. All sovereign debt securities held in the ‘held for trading’, ‘designated at fair value’, and available for sale’ are repriced at fair value covering all issuers.
Both exercises exclude any management actions in response to the stress scenarios (e.g., portfolio rebalancing or liquidation of positions).

At the same time, the EBA and the FSAP stress tests differ in a number of ways:

Stress testing approach: EBA uses a constrained balance sheet approach, with the SSM challenging banks’ bottom-up submissions and synthesizing outputs of different supervisory models. The FSAP test is based on a top-down (TD) approach using in-house models. The two tests use different methodologies to calculate credit risk, market risk, pre-provisioning net revenue (PPNR), and RWA impact.
Risk weighted assets (RWAs): While the EU-wide stress test is conducted on the assumption of a static balance sheet, the FSAP test assumes constrained balance sheets whereby credit demand declines relative to the baseline but supply effects are disallowed. It also includes foreign exchange (FX) shocks to foreign claims which drive loan dynamics, as well as a credit rating downgrades on standardized exposures which contribute to stressed capital requirements. While the FSAP test applies Point-in-Time (PiT) PD shifts to compute RWAs, EBA test uses Through-the-Cycle (TTC) PD shifts. Under the FSAP approach, capital requirements for market risk evolve with balance sheet assumptions rather than with stressed market risk factors.
Interest rate risk: The FSAP stress test conducted a granular assessment of interest rate risk using SSM’s IRRBB data templates. Shocks to funding costs and lending rates were calibrated at the bank and portfolio level linked to the scenario and informed by statistical analysis to project idiosyncratic risk. The EBA approach allows banks to project net interest income which is constrained by a series of caps and floors laid out in EBA methodological note.
Market risk: The FSAP test stressed the fair valuation on marked-to-market positions using SSM’s market risk sensitivities (“greeks”) reported under the STE. Hedges on trading positions were considered inefficient under stress on concerns over basis risk. EBA stress test incorporates a bottom-up (BU) detailed assessment of market risk at the portfolio level, including hedges, and covers counterparty credit risk (CCR) losses and stressed credit valuation adjustment (CVA).
Risk coverage: By contrast with the FSAP stress test, the EBA stress test includes projections for conduct risk and other operational risks.
IFRS 9: The EBA test assumes that for banks commencing to report under IFRS 9 in the first quarter of 2018, the impact of the introduction of IFRS 9 is included in starting point data as well as in the projections of banks. By contrast, the FSAP stress test uses the accounting regime valid at end-2017 to make credit risk projections.

D. Stress Testing in the FSAP

9. The FSAP team conducted a range of top-down stress tests to assess the euro area banking system ability to withstand system-wide shocks. The FSAP stress tests contribute to the objective of ensuring that banks have the capacity to absorb losses without engaging in deleveraging or amplification of market and liquidity stress. These tests were designed to identify macroprudential concerns and complement the bottom-up EBA exercise, which serves microprudential and supervisory purposes.

10. Stress test results should be interpreted with caution. The FSAP stress test results are based on end-2017 supervisory data. These data were complemented by supervisory historical data starting in September 2014 when the SSM was established. The FSAP team had to merge post-September 2014 SSM reporting templates with data sourced from publicly available sources as it did not have access to long series of supervisory data needed to conduct econometric work. Public data sources were also required to fill data gaps (e.g., portfolio composition of the trading book). Despite the FSAP team’s best efforts to build a consistent database, the matching and reconciliation of risk data extracted from multiple data sources is a complex exercise subject to caveats.

11. Additional challenges are posed by the use of proxy variables. Some risk parameters were projected using proxy variables on the back of data availability constraints, adding challenges to the quantitative impact of the stress testing exercise. For instance, the duration of the fixed income portfolio was proxied by remaining maturity using the transparency template of the 2016 EBA exercise even though periodic coupon payments would reduce duration. Probability of default (PD) projections for selected portfolios was based on econometric analysis conducted using Moody’s expected default frequency (EDFs) series which tend to display greater volatility than regulatory PDs and are based on a corporate representative portfolio. Market risk projections were based on banks’ market risk sensitivities to risk factors calculated on the value of net exposures after hedging, which limited the estimation of basis risk under stressed conditions.

12. A broader caveat is that stress test scenarios are calibrated using historical data which identify “tail events” based on a historical distribution. Yet it is well known that the nature of crises is to have unanticipated shocks and unexpected interrelationships where the past offers limited guidance. This is particularly relevant for the effect of a possible hard Brexit, the implementation of IFRS 9 and other regulatory changes, repricing risk due to basis risk and optionality, and liquidity risk associated to contingent liabilities (e.g., margin calls or pipeline risk). While some nonlinear effects can be captured in stress tests, it is always possible that unknown patterns emerge, especially if extreme shocks materialize.

E. Sample of Banks

13. The FSAP stress test is carried out on a sample of 28 (29) large euro area banks for the solvency (liquidity) test covering broadly 65 percent (70 percent) of banking system assets.⁵ The list of 28 (29) covered banks are headquartered in nine-euro area jurisdictions, and account for over 15 (16) trillion euros. The criteria chose are similar to EBA’s criteria to select the sample of large banks participating in the EU-wide stress test (including coverage of around 70 percent of euro area banking system assets, and minimum size of 30 billion euros). The FSAP sample of banks for the solvency stress test covers around 85 percent of the 2018 EBA sample. Of the excluded banks relative to the 2018 EBA sample, three are German, two are French, and one of each is domiciled in Italy and the Netherlands.

14. The sample of banks include major euro area G-SIBs. The FSAP sample for solvency (liquidity) stress test includes six (seven) G-SIBS of the 30 G-SIBs identified by the FSB in November 2017. Banks are heterogeneous in terms of their size, business mix and international presence. Among the G-SIBs that are more internationally diversified, three remain focused mainly on lending activities while the others have a more balanced business mix, including capital market activities.

15. To track patterns, identify trends, and enable simple benchmarking, the 29 euro area banks were grouped into three broad categories. To find suitable peer groups of banks structurally similar across jurisdictions, a combination of quantitative decision rules (e.g., size, complexity, size of trading books, derivative instruments, foreign subsidiaries) and expert judgment was applied. Different business models were identified with three broad categories: G-SIBs, large, but less complex, internationally-active banks, and relatively smaller domestically-focused banks (Table 2).

Table 2.

Euro Area: Sample of Banks and Material Geographies

Source: Banks’ Pillar 3 disclosures, and IMF staff calculation.

F. FSAP Stress Testing Strategy

16. The resilience of the euro area banking system was assessed using a battery of stress tests. The assessment followed a two-pronged approach:

A battery of solvency stress tests including a fully-fledged scenario-based test, a range of sensitivity tests, and a targeted reverse stress test to assess the banking system’s capital adequacy. The macrofinancial scenario explored domestic as well as global risks. It also included a traded risk scenario to capture business risk from tighter financial conditions, changes in market valuations, and vulnerabilities from asset liability management (ALM) operations. A wide range of sensitivity tests was conducted to further explore the resilience of the euro area banking system to wider shifts to risk factors. A reverse stress test on opaque complex assets assessed the valuation shock from soft mispricing required to exhaust capital buffers of G-SIBs.
A wide range of liquidity stress tests to assess bank resilience to structural and idiosyncratic liquidity risks which lead to sudden, sizable withdrawals of funding. The analysis was based on: (i) structural liquidity analysis, that is, a Liquidity Coverage ratio (LCR), Net Stable Funding Ratio (NSFR), Asset Encumbrance, and funding concentration, and (ii) a cash-flow test applied under an array of alternative scenarios, simulating a range of stress factors over five-day, one- and three-months maturity buckets.

G. Banking Sector Risks

17. Financial vulnerabilities in the euro area remain significant,⁶ though unevenly distributed across countries and segments of their financial systems:

In certain euro area countries, households, nonfinancial firms, and government bear heavy debt burdens, making them vulnerable to a tightening of rollover conditions and limiting the buffers available to deal with contingencies.
Global rising asset prices and compressed spreads are raising increasing concerns in the context of historically low market volatility and ultra-low “safe” interest rates. Some indications suggest that credit and market risks may be being downplayed, particularly in the high-yield fixed income market and possibly equity markets.⁷ Residential real estate prices continue to rise, particularly in some regions.

18. The euro area banking system faces a possible confluence of external and domestic risk factors involving tighter financial conditions, a return to weaker growth, as well as policy and geopolitical uncertainties. These risks, which could be mutually reinforcing, are summarized in the Risk Assessment Matrix (RAM, Appendix I) and entail:

Tighter global financial conditions: A reassessment of fundamentals amid global monetary policy normalization could trigger abrupt financial market movements, which may be amplified by a resurgence of concern over sovereign, bank, and corporate creditworthiness, with adverse knock-on effects to the real economy. Higher risk premia (including a possible decompression in term premia associated with an upside inflation surprise) and lower asset prices imply valuation losses and a reduction in the value of collateral and recoveries. Given vulnerabilities and other risk factors, distress in a single major institution or group of institutions could create contagious investor uncertainty.
Weaker global growth: A relapse of growth in advanced economies or a significant slowdown in China and other larger emerging market countries could have a sustained negative impact on euro area exports, investment, and consumption, thereby bringing the recent growth surge to a halt. The impact on public and private debt sustainability may contribute to higher risk premia.
Heightened policy and geopolitical uncertainty: Political and policy uncertainty—especially over Brexit negotiations—combined with a wide-scale retreat from cross-border integration, spurred on by rising protectionism, as well as geopolitical tail risks would reduce policy collaboration, dampen the flow of capital and liquidity across countries, increase risk aversion, and could abruptly undermine the cyclical recovery.

Solvency Stress Test

A. Key Elements of the Stress Test

19. The IMF stress test was conducted on end-2017 data at the highest level of consolidation in the euro area. The perimeter of consolidation of the banking group is specified in CRD IV. While insurance activities were excluded, banking associates were included. To select the banks included in the exercise, a combination of criteria was used including the sample of banks covered in the 2018 EU-wide stress test, banks’ share in the domestic market, and banks’ role in the euro area payment system.

20. Stress tests were based on ECB confidential supervisory data post-June 2014. The ECB shared all supervisory data templates including EBA’s Implementing Technical Standards (ITS) which cover FINREP and common reporting templates (COREP) as well as data gathered via the ECB’s STE with detailed quarterly reporting on interest rate risk in the banking book and banks’ elasticities to market risk factors. The composition of bank’s debt securities portfolio was extracted from EBA’s 2016 and 2017 transparency exercise, banks’ annual reports, and Pillar 3 disclosures. Other public data sources included ECB’s MIR statistics, financial data vendors (i.e. Bloomberg, Dealogic, Haver Analytics, Moody’s KMV, Fitch), and IMF’s World Economic Outlook (WEO).

21. The assessment criteria (“hurdle rate”) includes the capital standards implemented via the Capital Requirements Regulation (CRR) and the phased-in buffers. The hurdle rate applied in the FSAP stress test was set at the CET1 regulatory minimum of a 4.5 percent Pillar 1 requirement, a fully-phased capital conservation buffer (CCB), and a phased-in buffer for G-SIBs and other systemically important institutions (O-SIIs). This led to a CET1 hurdle rate ranging between 7.0 and 7.5 percent. The hurdle rate also includes a 3 percent Tier 1 ratio based on the CRR leverage framework. Attention is also paid to other measures of soundness, such as headline income, own equity funds, and other regulatory capital definitions (i.e. Tier 1, total regulatory capital).

B. Scenarios

22. The scenario-based stress test included a baseline scenario and a stress scenario projected over a three-year horizon. The scenario starts in the first quarter of 2018 and extends through the fourth quarter of 2020 to enhance comparability with EU-wide stress test. The baseline scenario was based on the October 2017 WEO.

23. The adverse scenario was designed by the FSAP team in cooperation with the ECB to explore the impact of global and domestic risks on major euro area banks. The adverse scenario was simulated using IMF staff’s Global Macrofinancial Model, a structural macroeconometric model of the world economy, disaggregated into forty national economies.⁸

24. The scenario covers the risk factors presented in the RAM, and the calibration of shocks is in line with tests undertaken as part of past euro area country FSAPs. Accordingly, the scenario incorporates a widespread growth slowdown; heightened uncertainty reflected in trade, productivity and investment shock; higher risk premia and money market rates; and valuation losses on real estate, equity, and other assets. The severity of the economic contraction in this scenario is comparable to those in other euro area country FSAPs and the authorities’ EU-wide stress tests (Figure 1). The overall scenario is expected to have a strong effect on banks’ profits and capital because it combines a macroeconomic recession with financial market tightening, when fiscal and monetary policy buffers are largely exhausted in the euro area.

Macroeconomic Risk Components

25. The adverse scenario is characterized by a severe global recession in which the euro area experiences a balance sheet recession, concentrated in high-spread economies.⁹ Overall, output contracts by 7.1 percent relative to the baseline by 2020, while consumer price inflation falls by 3.3 percentage points. This output loss is somewhat concentrated in high spread economies, which experience a contraction of 7.6 percent, versus 6.9 percent in low-spread economies. The assumed increases in bank funding costs in the form of higher money market interest rate spreads, together with higher credit loss rates, induce bank lending interest rate to increase, which in turn induce contractions in bank credit. Total employment falls by 5.9 percent in the euro area, the government debt ratio rises by 18.4 percentage points in high spread economies and by 13.3 percentage points in low spread.

26. Each scenario includes 370 core macrofinancial variables encompassing domestic and international economic activity. Projections include paths for core macrofinancial variables in all material geographies for large euro area banks. Core variables for each geography include growth, inflation, real investment, unemployment, policy rate, yield curve, equity prices, foreign exchange, credit growth, and residential real estate prices. Global assumptions include world GDP growth, growth in the euro area, growth in emerging markets, fuel prices, non-fuel prices, euro area swap curve, and credit spreads by major index.

Market Risk Components

27. The stress test scenario features financial market stress, amplified by heightened uncertainty and confidence losses. De-globalization initiatives in Europe and the United States limit or reverse international trade and financial integration, generating a sell-off in stock markets on profitability concerns and reduced risk appetite, with the real equity price falling by 20 percent in the euro area, United Kingdom, and United States over two years.¹⁰ Elevated global capital market volatility widens money market spreads by 75 bps in the euro area, and by 50 bps in the United Kingdom and United States.

28. The impact of financial tightening conditions on bond yields is, on average, larger than that prescribed by the 2018 EU-wide stress test. Term premium decompression in Europe and the United States, together with the re-emergence of sovereign stress in high spread euro area economies, raises long term government bond yields, by 150 bps in high spread euro area economies, by 75 bps in low spread euro area economies, and by 100 bps in the United Kingdom and United States. While the re-emergence of public debt sustainability concerns leads to an increase in the level and dispersion of government bond yields in the 2018 EU-wide stress test, the impact of government yield shocks under the FSAP scenario is larger. This is consistent with other FSAPs and is due to the cumulative nature of shocks in the FSAP test in contrast with the instantaneous shock approach under the 2018 EU-wide stress test (Figure 2).

29. The yield curve was constructed using polynomial interpolation methods. The yield curve for each of the covered public debt markets was built using the polynomial of the lowest possible degree between the projected short- and long-end of the curve. The euro are swap curve was spanned based on a constrained VAR forecasting approach using the exogenous path for the policy rate and benchmark curves in the euro area and the United States. A similar methodology was used to generate the projected paths for European iTraxx indices, as well as European corporate bond yields by sector (financial, non-financial) and rating (investment grade, high-yield) based on the exogenous path for policy rates and government spreads. Missing data on some geographies was estimated using benchmark data from comparator economies.

C. Credit Risk Modelling Approach

Relevance of Credit Risk

30. Credit risk accounts for the largest regulatory capital requirement of euro area banks. At end-2017, the size of the largest 28 banks reached €15.0 trillion, with RWAs of €5.6 trillion, representing around 37 percent risk density. Over 85 percent of capital is required to cover unexpected losses from credit risk. By contrast, market risk represents under 10 percent of capital requirements which partly reflects banks’ reactions to new regulatory initiatives, including restrictions on proprietary trading by banks, as well as changes in business models.

31. The impact of credit risk on banks’ capital ratios depends on the regulatory approach used by banks to book credit exposures. As credit risk materializes, scenario-based stress testing requires assessing the impact of shocks to macroeconomic and financial variables on banks’ loan loss provisions and capital requirements. For exposures booked under the IRB approach, credit risk evolves with the exposure at default (EaD), the PD, and the loss-given default (LGD). For exposures under the STA approach, a deterioration in credit risk is reflected in higher specific and collective allowances from higher default rates and lower creditworthiness of performing loans, as well as in higher capital requirements from credit risk downgrades of the underlying exposures.

32. The reliance of euro area banks on IRB models varies across banks and portfolios. On aggregate, two-thirds of credit risk exposures were booked under the IRB approach, however, there is a wide dispersion of practices across banks and portfolios. There is a cluster of banks that shows greater reliance on standardized models for exposures to central banks and general governments than for other credit portfolios (Figure 3).¹¹

Credit Risk Model for IRB Exposures

33. The impact of the economic scenarios on risk parameters was conducted by country segment and exposure class. Overall, there are 37 material geographies for the sample of large euro area banks, including 9 home (headquarter) jurisdictions, 11 other EU countries, and 17 geographies outside the EU¹² (Table 2). Covered COREP exposure categories include seven portfolios including central governments or central banks, institutions, corporate—SME, corporate—specialized lending, corporate—other, retail—secured by real estate, retail—other. Other non-credit obligations assets and default fund contributions were treated as corporate— other. A separate credit risk model was estimated for each bank’s material geography and regulatory portfolio.

34. To calculate expected losses and capital requirements (RWAs), the FSAP team projected its own stressed risk parameters using in-house internal models. Loan loss impairment charges are given by $E L_{i, t}^{j} = P D_{i, t}^{j} * L G D_{i, t}^{j} * E A D_{i, t}^{j}$ where i denotes the bank, j denotes the asset class, and t is the time dimension. Capital requirements $c a p_{i, t}^{j}$ were calculated using the conditional expected loss laid out in Basel III’s supervisory mapping function with asset correlations and maturity adjustments prescribed in CRR/CRD IV. PiT shifts were applied to regulatory PDs for non-defaulted exposures, and risk-weighted assets were defined as $R W A_{i, t}^{j} = 12.5 * c a p_{i, t}^{j} * E A D_{i, t}^{j}$

35. Stressed projections were applied to around 10,000 initial credit risk parameters. The pool of parameters included four key risk parameters (PD, LGD, EaD, RWA) broken down by exposure class (7 portfolios), country (12 geographies), and bank (28 banks). Risk parameters assigned to the obligor pool by portfolio and geography in COREP 09.02 were complemented by worldwide parameters reported in COREP 08.02 according to banks’ modeling approach (i.e. IRB-Foundation, IRB-Advanced), asset class (e.g. institutions, large corporates...), and obligor grade (internal rating scale). Obligor grades with an implied PD=1 were excluded. For credit exposures reported with geographic breakdown, the FSAP team subtracted the defaulted exposure from the original exposure and estimated implied non-defaulted PDs and LGDs. This process required making adjustments to address the fact that not all exposures are assigned to obligor grades or pools by obligor.¹³

36. Stressed credit risk parameters were applied to non-defaulted net exposures using regulatory parameters. The FSAP team applied stressed conditions to non-defaulted exposures. Although on aggregate, it did not compute an additional capital charge for defaulted assets to cover systematic uncertainty in realized recovery rates for these exposures, it included an add-on capital surcharge on the value of the collateral for large defaulted portfolios.¹⁴ Credit risk projections were applied to projected net exposures after credit risk mitigation techniques including SME-supporting factor.

37. The FSAP team used a two-step process to project stressed PDs:

Based on the economic scenario, econometric methods were applied based on the historical relationship between economic and financial conditions and proxies on credit risk parameters for each aggregate portfolio and country of exposure;
A statistical analysis was used to adjust bank-specific projections based on portfolio proxies at the aggregate level to banks’ portfolios to capture the individual risk profile of underlying exposures, drawing on post-June 2014 supervisory data.¹⁵

38. The PD proxy for exposures to central governments was extracted from sovereign yields using a Merton-based approach.¹⁶ The FSAP team used data from FINREP returns on the breakdown of financial assets by counterparty to back out banks’ exposures to ‘general governments’ from the combined COREP category ‘general governments or central banks’. It then applied a reduced-form structural model to extract PD estimates from the sovereign spreads projected in the scenario.¹⁷ Using the credit spreads for sovereign i linked to the scenario $S_{t, T}^{i}$ , time to maturity (T-t), and assuming LGD=45 percent, the implied risk-neutral PD is backed-out as:

P D_{t, T}^{i} = \frac{1 - \exp^{- S_{t, T}^{i}} \cdot (T - t)}{L G D_{t}^{i}}

39. PDs for institutions, corporates (including specialized lending), and retail-unsecured were sourced from Moody’s KMV using the one-year expected default EDF average estimate. The following categories were used: the financials group, the corporate group, the construction and real estate development group, and the consumer nondurables and services group. These categories were mapped to major COREP portfolios, i.e. institutions,¹⁸ corporate (including SME and specialized lending), and retail unsecured (including qualifying revolving, and other than secured lending), respectively. For expected losses, the breakdown of loans and advances to non-financial corporations by NACE code in FINREP was mapped to Moody’s sectoral (e.g., mining and quarrying was mapped to steel and metal products group).

40. To address the truncated nature of the default rate distribution, a logit transformation was applied before conducting the econometric analysis. This transformation addressed biases and ensures that the projected rate is contained within the 0–1 bound once the logit forward path is applied to the forecast.

41. Given the uncertainty associated to tail risk predictions, a robust econometric framework was implemented using a variety of approaches:

A time series econometric technique using Newey-West HAC-robust standard errors to obtain consistent estimators when the error term is heteroskedastic, autocorrelated, or both, once the regressors are proved stationary;
A quantile regression approach to address the concern that the drivers of the conditional mean might be different in the higher tail of the credit risk distribution. The distribution was divided in quartiles (four segments). Robustness of the results was checked using a distribution based on deciles (ten segments) to explore cliff-effects;
A Bayesian Model Averaging (BMA) approach to address modeling uncertainty and the plausibility of different drivers of credit risk dynamics, using a Normal diffuse prior distribution.

42. Credit risk was linked to a set of local drivers, regional variables, and global factors. Arguably, credit risk, particularly for wholesale clients, is not only a country-specific type of risk but it is also driven by regional macroeconomic forces external to the country as well as by global funding conditions. To estimate credit risk, a statistical-type model was developed by country and portfolio level using the following set of drivers:

Local variables: there are a number of possible economic forces that determine the risk profile of a portfolio including trends in real GDP, a GDP-based recession indicator (to capture non-linear effects), inflation, unemployment, the yield curve, the spread over swap rate, credit growth, equity prices, and real estate developments. Local drivers were forecasted for 37 geographies under baseline and adverse conditions;
Regional variables to capture information about the state of the euro area and the growth outlook of emerging markets. Regional factors include GDP growth and inflation in the euro area and emerging economies, developments in the euro swap curve, 3-month euro repo rates, 3-month euro Libor, European corporate spreads, EMBIG spreads, Eurostoxx 50, and structural challenges from debt sustainability concerns (measured by sovereign debt spreads over the Bund);
Global financial market variables: the ability of wholesale clients to repay also depends on the state of the global economy; in addition, shifts to global liquidity are likely to trigger default risk shocks. Global variables include world GDP growth, developments in commodity prices (fuel, non-fuel), changes in 3-month U.S. dollar Libor, TED spread, changes in U.S. investment-grade (IG) and high-yield (HY) corporate bond spreads, and U.S. equity prices.

43. Real house prices were projected using a long-run price and investment equations in an error correction framework.¹⁹ The explanatory variables in the price equation (Pt) include real income (yt), the stock of residential dwellings (st), and the real interest rate (rt). Real house price indices were sourced from the OECD analytical house price database. Income was measured by real household disposable income. Interest rates are measured by the long-term real interest rate. The measure of new housing supply in the investment equation (It) was real residential investment defined as real gross fixed capital formation in dwellings (ct) sourced from the OECD. The following system of equations is estimated for each euro area country:

\begin{matrix} P_{t} = α_{0} + α_{1} y_{t} + α_{2} s_{t} + α_{3} r_{t} + ζ_{t}^{P} \\ I_{t} = β_{0} + β_{1} c_{t - 1} + β_{2} P_{t - 1} + ζ_{t}^{I} \end{matrix}

44. From the initial set of determinants, the core set of drivers was chosen using a general-to-specific selection approach. For major portfolios, the final set of core determinants in the credit risk equations included: GDP growth, the slope of the yield curve, the 10-year government spread over the swap rate, the 3-month LIBOR-U.S. dollar spread over U.S. T-bill rate, real GDP growth in the euro area, and a dummy variable that takes the value of 1 if GDP growth is below its 15^th percentile or negative, and zero otherwise; this dummy is multiplied by the size of the GDP shock. In addition, credit risk in real estate exposures was explained by real estate price developments, while credit risk equations for retail unsecured loans included credit growth projections. Credit risk in emerging markets was also explained by changes in inflation, FX developments, and EMBIG corporate spreads.

45. Regressions were run using different sample periods and lag structures. Regressions were run over two sample periods: a full sample to assess in-sample performance, and a truncated sample to assess out-of-sample performance.²⁰ Quarterly regressions included different lagged structures to test model performance.

46. Using time-series analysis, econometric results suggest that high-spread economies are particularly impacted by financial stress in the euro area. Annex Table 1 reports regressions for the PD corporate portfolio of a low-spread country where all the data is pooled in columns (1) through (3), while columns (4) through (6) report regressions based on the truncated sample. Corporate stress increases with low GDP growth, a steepening of the yield curve, and a widening of money market spreads. Annex Table 2 show regressions for the corporate PD of a high-spread economy in columns (1) and (2), and another high-spread economy in columns (3) and (4). Results suggest that sovereign stress and the growth outlook in the euro area are relatively more significant.

47. Credit risk in emerging markets is also driven by rising long-term rates, foreign exchange shocks, and negative trend in regional growth. Annex Table 3 results suggest that stress in the corporate sector of a large emerging market is heightened with rising interest rates, a depreciation of the real relative to U.S. dollars, and lower GDP growth in emerging markets. This picture is consistent across the most relevant emerging markets for euro area banks including Brazil, Chile, China, Colombia, India, Mexico, Peru, Russia, and Turkey.

48. A quantile regression approach is used to project severe credit risk in the upper quantiles of the PD distribution. The empirical specification of the PD level of portfolio j in geography i is as follows:

P D_{λ, t}^{i, j} = Z_{t - 1}^{i} \cdot β_{λ}^{i, j} + u_{λ, t}^{i, j}

where $Z_{t - 1}^{i}$ is the set of lagged determinants including the country, regional, and global factors specified above, and λ. denotes the quantile of the conditional distribution.

49. Results suggest that credit risk projections are more severe using estimated coefficients from the upper quantiles of the PD distribution. Figure 4 presents a summary of quantile regression projections for corporate exposure PDs in a low-spread country under the adverse scenario. Credit risk increases from the 1.3 percent value observed in 2017 to a peak of 4.0 percent in 2019 conditional on the upper quantiles. This impact is softened to 1.8 percent at the lower quantiles.²¹

50. The quantile-based model identifies greater impact from sovereign stress and a cliff effects from GDP contraction in the upper tail of the distribution. Figure 4 reports the coefficients of credit risk determinants in the corporate portfolio of another low-spread country across quantiles. While the effect of GDP growth and the slope of the yield curve is comparable across percentiles, the effect of sovereign stress and the GDP-based recession indicator is more severe for higher PD grades.

51. A Bayesian VAR estimation (BVAR) analysis is applied to produce conditional forecasts based on the scenario. The model includes the country, regional, and global factors specified above. All the variables are endogenous except the global factors which are considered exogenous. The model relies on a normal-diffuse prior distribution. Projections were computed using the Bayesian estimation, analysis, and regression (BEAR) toolbox.²²

52. The BVAR approach suggests larger shifts to corporate PDs in the core relative to the periphery. Figure 5 shows projections for the nine home euro area countries under the adverse scenario. Results suggest three clusters of PD shifts under stress: elevated impact in countries 1, 3, 6, and 9; moderate impact in countries 2, 5, and 8; and lower impact in countries 4 and 7. Projections were performed over a 5-year horizon to test the stability of the coefficients. These projections are, however, subject to wide confidence intervals in the outer years of the scenario.

53. Final credit risk projections were based on a model selection routine which prioritized out-of-sample performance to avoid overfitting. A combination of criteria was used to inform final projections using the following order: out-of-sample forecast performance, measured by the root mean squared error over the period 2015Q1–2016Q4; in-sample forecast performance for the overall period 2005Q1–2016Q4; goodness of fit of the regression; sign of coefficients according to theory; and expert judgment applied over the projected paths benchmarked against the 2008 financial crisis and the 2012 European sovereign debt crisis.

54. Projections for credit risk in real estate were consistent with ECB staff’s projections.²³ The lack of a time series on default rates in retail mortgages motivated the use of the ratio of PD projections in mortgage loans relative to corporate PDs generated by ECB staff’s modelling tool. This ratio was applied to the FSAP team’s projected corporate PD paths to generate default rates on retail exposures secured by immovable property. LGD projections for loans collateralized by real estate were assumed not to decline under the baseline scenario despite the slight improvement in the outlook of real estate. The real estate correction assumed in the adverse scenario triggered an increase in LGD across geographies with the median LGD increasing by 50 percent relative to the starting point for residential real estate and 12 percent for commercial real estate.²⁴

55. Stressed LGD projections were informed by banks’ reported projections on defaulted exposures and ECB staff’s multipliers. LGD projections were derived using the ratio of LGDs on defaulted exposures over the LGD on non-defaulted exposures using bank supervisory data. They were also informed by aggregate shifts to LGDs provided by ECB staff broken down by portfolio and geography. The following approach was applied:

L G D_{i, t}^{c, j} = \max (\frac{L G D_{i, 2017}^{c, j} | d e f a u l t e d}{L G D_{i, 2017}^{c, j} | n o n - d e f a u l t e d}, m u l t i p l i e r_{t}^{c, j}) \cdot L G D_{i, 2017}^{c, j} |

where $L G D_{i, t}^{c, j}$ is bank’s i LGD for country c, portfolio j and time t, and multiplier is the credit risk benchmark provided by ECB staff calibrated to the FSAP scenario. The LGD for the non-defaulted portfolio was backed out from COREP 09.02 using information on the reported LGD for the aggregate portfolio and the LGD for defaulted exposures.

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

E A D_{i, t}^{c, j} = E A D_{i, t - 1}^{c, j} \cdot (1 + g_{t}^{c} + f_{t}^{c} \cdot Δ F X_{t}^{c}) \cdot (1 - P D_{i, t - 1}^{c, j}) + Δ L_{i, t}^{c, j} \cdot U C L_{i, t - 1}^{c, j}

where i denotes the bank, j denotes the portfolio, c denotes the country of exposure, $g_{t}^{c}$ is credit growth in country c (where demand effects are incorporated but supply effects are disallowed), $f_{i}^{c}$ is the fraction of foreign currency loans, $Δ F X_{t}^{c}$ is the depreciation of the foreign currency relative to the euro, $(1 - P D_{i, t - 1}^{j})$ represents the non-defaulted portfolio, $Δ L_{i, t}^{j}$ is the shock to triggered credit lines and guarantees, and $U C L_{i, t - 1}^{j}$ is the amount of undrawn guarantees. While paths for credit growth and FX shocks were generated by the scenario, stressed credit conversion factors on undrawn credit lines and guarantees were informed by historical behavior of offbalance sheet migration during stress periods drawing on banks’ pillar 3 disclosures.

57. To compute capital requirements, ‘hybrid’ regulatory risk parameters were projected and applied to Basel III formula for IRB exposures. The regulatory credit risk parameters (PD, LGD, EaD) for banks’ approved IRB models are typically “hybrid” estimates between ‘pure PiT’ and ‘pure TTC’ estimates, depending on each bank’s methodological choice (model structure, method to calibrate on historical data, incorporation of time-varying ‘central tendency’). The derivation of RWA expansion is dependent on stressed credit risk parameters, correlation assumptions, and effective maturity for each exposure. In line with the Basel III framework, RWAs were computed after applying the scaling factor of 1.06 to credit RWAs and using a 1.25 multiplier to the correlation parameter of all exposures to large regulated financial institutions and to all unregulated financial institutions. PiT shifts to PDs were used to feed stressed PDs into the IRB supervisory formula while LGDs were projected using a stressed recovery value on regulatory downturn LGDs.

Credit Risk Model for STA Exposures

58. The estimation of impairments and capital requirements for standardized exposures required the use of econometric analysis to estimate:

the starting value of the risk parameters;
the impact of the scenario on projected risk parameters;
the computation of impairment flows and provisions;
the impact on capital requirements based on credit rating assessments of claims.

59. Credit risk projections were applied to around 7,000 initial credit risk parameters. For each bank, a key segmentation is between defaulted and non-defaulted exposures. For each of these exposures, a further breakdown included the sixteen STA asset classes reported in COREP. These exposures were mapped to the seven portfolios constructed for IRB exposures: government (central government or central banks, regional governments, public sector entities, multilateral banks, and international institutions); institutions; corporate-SME, corporate-other (corporates excluding SME), retail unsecured (retail), retail secured (secured by mortgages on immovable property), and other exposures (items with high risk, covered bonds, claims with a short-term credit assessment, claims in the form of CIU, equity exposures, and other items). Credit risk parameters included credit risk adjustments (general and specific), exposures in default (by asset class), exposure value, and risk weights. Geographic coverage included all material geographies reported to EBA by each bank in the context of the 2016 EU-wide stress test.

60. Two approaches were used to construct the starting point for the default rate (PD) by portfolio and geography:

The default rate was calculated as the ratio of the stock of exposures in default to the amount of total exposures drawing on COREP 09.01.a. reporting. To convert this stock-based metric into a flow-based measure, the accumulation of defaults over time was projected using a combination of statistical analysis using the flow of new defaults and write-offs reported in COREP 09.01.b by portfolio since 2014, and expert judgment based on individual banks’ management of problem loans (drawing on information supplied by the ECB);
The 1-year expected default rate reported on IRB exposures for the same portfolio and geography was used drawing on COREP 09.02 reporting. While this definition is a forward-looking measure rather than a realized measure, it has the advantage of being reported as a flow and therefore does not require additional assumptions on the flow of write-offs needed to project default rates and loan loss impairment charges.

61. The coverage ratio for defaulted exposures was floored at the LGD rate reported for the same portfolio under the IRB approach.²⁵ Banks report the general and specific provisions computed on defaulted exposures in COREP templates. This measure captures the eligible collateral and guarantees on defaulted claims in line with credit risk mitigation techniques. Under the adverse scenario, a 65 percent estimated coverage ratio was applied subject to the following constraint:²⁶

L G D_{t}^{c, i} = \max {\max (\frac{g e n e r a l ​ p r o v i s i o n_{t}^{c, i} + s p e c i f i c p r o v i s i o n_{t}^{c, i}}{\exp o s u r e i n d e f a u l t_{t}^{c, i}}, L G D_{t}^{c, i}) \cdot m u l t i p l i e r_{t}^{c, i}, 0.65}

62. To forecast the flow of exposures in default, the default rate was multiplied by the amount of performing exposures projected each period. General and specific credit risk adjustments were projected by multiplying the flow of new impairments by the stressed LGD by geography and portfolio. The flow of new impairments was projected using the same econometric approach applied to IRB exposures. In line with balance sheet assumptions for the IRB portfolio, the stock of performing exposures was driven by the growth of the loan book, structural FX risk in foreign exposures, triggered credit lines and guarantees, and the flow of new impairments.

63. Regulatory capital requirements were computed using three steps:

The risk weight of exposures in default, excluding exposures at zero percent risk weight, was computed for each portfolio drawing on COREP 09.01.a;
The risk weight on the new flow of exposures in default was computed as:
$R W {}_{t}^{c, i}| d e f a u l t = \max {R W {}_{2017}^{c, i}| d e f a u l t, 100 %}$
For the non-defaulted portfolio, one-notch downgrade of the underlying exposure was assumed under the adverse scenario.²⁷ The initial risk-weight density reported by banks for non-defaulted exposures in COREP 09.01.a was mapped to the external rating by portfolio laid out by Basel, and one-notch downgrade was added to calculate the projected risk-weight density.²⁸

Credit Risk Impact on Profit and Loss

64. Starting regulatory risk parameters and point-in-time shifts were used to compute loan impairment charges. Expected losses were calculated on all exposures including on-balance sheet and off-balance sheet exposures, taking into account migration of off-balance sheet commitments to on-balance sheet. Coverage included all asset classes for IRB and STA exposures reported in CRR. Initial credit risk parameters were those reported for regulatory purposes in COREP templates. Point-in-time shifts were informed by models calibrated on the scenario. These shifts were applied to banks’ regulatory parameters to derive stressed expected losses and capital requirements.

65. The IMF stress test allowed the draw-down of provisions built for non-defaulted exposures to cover expected loan losses over the stress test horizon. The FSAP team constructed a measure of total provisions for non-defaulted exposures aggregating data on value adjustments and provisions on IRB exposures for obligor grades lower than 1, and data on general and specific credit risk adjustments on STA exposures excluding exposures in default. These provisions were used to mitigate the profit and loss (P&L) impact of credit losses over the 3-year stress test horizon. At the same time, provisions were re-built on the flow of new exposures using the scenario paths for credit risk drivers.

66. The ratio of provisions for non-defaulted loans over unimpaired loans varies between 0.4 percent and 2.5 percent across jurisdictions. The wide disparity in provisioning practices across major euro area banks is reflected in the differential impact of credit risk on P&L. There are no international principles that promote the adoption of specific prudential tools on (general) provisions.²⁹ While the ECB and the EC have proposed a system of prudential backstops for accounting provisions related to nonperforming loans,³⁰ provisioning practices on non-defaulted exposures vary widely across banks.³¹ Under the IMF approach, banks with a greater deterioration in the obligor pool or lower built-up provisions for credit risk were hit more severely under stress.

67. Interest payments were assumed to accrue only on performing exposures under both the baseline and adverse scenario. The interest revenue on performing exposures was calculated on the gross carrying amount. While accounting rules allow banks accrue interest income on nonperforming exposures with provision required on the more delinquent and uncollectible assets, this practice distorts banks’ coverage ratios and requires a dedicated assessment of the degree of delinquency of the loan under stress. The IMF approach is more conservative than the 2018 EU-wide stress test methodology which allows banks to project income on nonperforming exposures on a net basis, i.e. on the value of the exposure net of provisions subject to the cap provided by the interest rate accrued at the reference date. The EBA choice is coherent with the IFRS 9 treatment of those assets.

68. The effective lending rate was computed on the amount of unimpaired assets at the 2017 reference date. To be consistent with the accrual assumption on performing exposures, the interest income from loans and receivables reported in the statement of profit and loss was allocated to unimpaired loans to compute the effective interest rate. This rate was subject to rate shocks under each scenario. The projected lending rate path was applied to the forecast volume of unimpaired loans to generate interest income on loans and advances.

D. Market Risk Approach

69. The scope of the market risk methodology covered all positions under fair value measurement. This includes financial assets held for trading (HFT), financial assets designated at fair value through profit and loss (FV), trading financial assets, as well as available-for-sale financial assets (AFS) which impact regulatory capital through other comprehensive income (OCI). The scope excludes amortized cost positions held in a hedge-accounting relationship, as well as hedge accounting derivatives. The traded risk scenario had an impact on capital resources while capital requirements for market risk evolve with balance sheet assumptions. Changes in CVA and CCR are excluded.

70. All accounting categories under a full or partial fair value measurement were revalued under baseline and adverse conditions over the 3-year horizon. This contrasts with the 2018 EBA methodology which assumes no market impact under the baseline scenario. The 2018 EBA methodology also assumes the full recognition of gains and losses from instantaneous market shocks in the first year of the stress test. Under the IMF approach, hedging instruments for interest rate risk associated to positions in the trading book and AFS portfolios are assumed to be ineffective under stress. While this is a conservative approach, the traded risk losses are partially reversed as asset prices recover after a sharp correction assumed during the first year of stress. Also, the severity of the correction increases as risks build and decrease after those risks abate, penalizing buffers built on risky securities.

71. Market risk factors include interest rates and credit spreads for debt instruments, equity prices, exchange rates, and commodities. Market risk losses arise owing to adverse moves in risk factors (market prices and interest rates) and default risk. The calibration of the traded risk component of the scenario is linked to the forward-looking scenario. The path for the market risk factors that were not included in the macroeconomic scenario (e.g., credit spreads, swap rates, yield curves for specific countries) was generated using satellite models broadly aligned to the macroeconomic scenario. For credit spreads on corporate bonds, the calibration was inspired by what happened in the 2008 financial crisis.

72. Revenue and cost changes in banks’ investment banking business were included provided they were reported under interest income. Client revenues of trading assets include bid/ask spread, fees and commissions, and investment banking fees. There revenues were projected as net interest income by applying shocks to the effective implied rate at the cut-off date using the amount reported in interest income for HFT and FV assets. Client revenues reported by banks under gains on financial assets and liabilities held for trading in the FINREP statement of profit and loss were excluded as they could not be disentangled from the valuation impact of the trading position.

73. The market impact was computed using a combined approach of instantaneous and multi-year shocks. For derivative positions and equity instruments in the trading book an instantaneous shock was applied. For the valuation impact of debt instruments (repricing risk, credit spread risk) as well as the P&L impact of net open positions in market risk factors (e.g. foreign exchange, commodities), a multi-year shock was applied in line with the 3-year scenario. Losses were absorbed the same year the shock hits.

74. Banks’ securities portfolio evolved in line with the constrained balance sheet assumption for credit risk. The notional values of the securities portfolio grow according to: where $B_{i, t}^{j} = (B_{i, t - 1}^{j} - \Pr_{i, t}^{j}) * (1 + g_{i, t} + f_{i, E U R} \cdot Δ F X_{E U R} + f_{i, U S D} \cdot Δ F X_{U S D}) \Pr_{i, t}^{j}$ 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.

Sovereign Exposures

75. A full revaluation of bank’s sovereign exposure by issuer and accounting portfolio was implemented combining FINREP data with the 2016 EU-wide transparency exercise results. FINREP templates do not provide a breakdown of sovereign bonds by issuer. The most recent data at the issuer level is provided by the 2016 EU-wide transparency exercise which shows sovereign exposures by country, residual maturity, and accounting portfolio. A mapping table between end-2017 FINREP data and the 2016 EU-wide transparency exercise was built to match the aggregate amount of sovereign exposures at end-2017. Results were checked against the update provided in the 2017 EU-wide transparency exercise which shows a higher level of aggregation.³² Market losses on sovereign exposures booked under HFT and FV were booked through profit and loss while valuation impact on AFS impacted regulatory capital through OCI.

76. While the average euro area bank holds around 75 percent of the sovereign portfolio in AFS, there is a wide dispersion across banks. Table 3 shows that euro area banks are subject to repricing risk in sovereign debt spreads as they hold around 90 percent of sovereign securities at fair value (i.e. AFS and HFT portfolios). This consideration does not consider IFRS 9 implementation. There is a wide dispersion of holdings across banks with some universal banks holding over 50 percent of sovereign exposures in the trading book making them particularly exposed to market impact through profit and losses (P&L).

Table 3.

Euro Area: Bank Sovereign Exposures by Book, December 2015

Source: 2016 EU-wide stress testing Transparency Exercise, and IMF staff calculations.Note: By contract to the 2016 EU-wide stress testing Transparency Exercise, the 2017 EBA Transparency Exercise did not show the breakdown of sovereign debt holdings by accounting book.

Table 3.

Euro Area: Bank Sovereign Exposures by Book, December 2015

	(In Million euros)			(In Percent)
Banks in:	AFS	HFT	HTM	AFS	HFT	HTM
Austria	15,420	9,318	22,525	33	20	48
Belgium	17,520	2,080	28,959	36	4	60
Germany	103,015	53,954	0	66	34	0
Spain	205,159	48,880	6,255	79	19	2
Finland	3,332	645	0	84	16	0
France	243,369	24,955	21,556	84	9	7
Ireland	14,668	0	5,406	73	0	27
Italy	147,497	41,620	11,359	74	21	6
Netherlands	117,794	8,343	6,029	89	6	5
Total	867,774	189,797	102,089	75	16	9

Table 3.

Euro Area: Bank Sovereign Exposures by Book, December 2015

	(In Million euros)			(In Percent)
Banks in:	AFS	HFT	HTM	AFS	HFT	HTM
Austria	15,420	9,318	22,525	33	20	48
Belgium	17,520	2,080	28,959	36	4	60
Germany	103,015	53,954	0	66	34	0
Spain	205,159	48,880	6,255	79	19	2
Finland	3,332	645	0	84	16	0
France	243,369	24,955	21,556	84	9	7
Ireland	14,668	0	5,406	73	0	27
Italy	147,497	41,620	11,359	74	21	6
Netherlands	117,794	8,343	6,029	89	6	5
Total	867,774	189,797	102,089	75	16	9

77. Banks hold around 9 percent of total assets in sovereign debt securities of which 30 percent are to their own sovereign, as of June 2017 (Table 4). There are significant differences across banks domiciled in different countries with holdings of sovereign securities ranging between 3 percent (Finland) to 14 percent (Austria and Italy) of total assets. On average, 30 percent of exposure is to their domestic sovereign and another 30 percent to other euro area sovereigns. Own-sovereign exposure ranges between 5 percent (Finland) to just under two thirds (Ireland).

Table 4.

Euro Area: Bank Sovereign Debt Security Exposures by Country, June 2017

Source: 2017 EBA Transparency Exercise, and IMF staff calculations.

Table 4.

Euro Area: Bank Sovereign Debt Security Exposures by Country, June 2017

	(In Million euros)
Banks in:	All Countries	Own Country	Other Euro Area	Of Which	Spain	Italy	Portugal	Non-Euro Area
Austria	47,412	8,372	11,592		972	484	55	27,448
Belgium	52,139	17,460	20,611		2,660	4,480	258	14,069
Germany	211,995	62,767	52,131		5,869	18,217	2,319	97,097
Spain	305,471	85,928	40,730		85,928	28,751	10,385	125,128
Finland	3,897	197	2,953		0	0	1	747
France	425,011	123,732	138,062		15,733	38,055	2,293	163,217
Ireland	21,219	13,323	5,516		1,665	1,189	126	2,379
Italy	232,416	97,842	75,208		28,909	97,842	143	30,719
Netherlands	124,465	31,734	55,084		2,811	1,438	0	37,647
Total	1,424,025	441,355	401,887		144,547	190,457	15,579	498,450
	(In Percent of Total)
Banks in:	All Countries	Own Country	Other Euro Area	Of Which	Spain	Italy	Portugal	Non-Euro Area
Austria	100	18	24		2	1	0	58
Belgium	100	33	40		5	9	0	27
Germany	100	30	25		3	9	1	46
Spain	100	28	13		28	9	3	41
Finland	100	5	76		0	0	0	19
France	100	29	32		4	9	1	38
Ireland	100	63	26		8	6	1	11
Italy	100	42	32		12	42	0	13
Netherlands	100	25	44		2	1	0	30
Total	100	31	28		10	13	1	35

Source: 2017 EBA Transparency Exercise, and IMF staff calculations.

Table 4.

Euro Area: Bank Sovereign Debt Security Exposures by Country, June 2017

	(In Million euros)
Banks in:	All Countries	Own Country	Other Euro Area	Of Which	Spain	Italy	Portugal	Non-Euro Area
Austria	47,412	8,372	11,592		972	484	55	27,448
Belgium	52,139	17,460	20,611		2,660	4,480	258	14,069
Germany	211,995	62,767	52,131		5,869	18,217	2,319	97,097
Spain	305,471	85,928	40,730		85,928	28,751	10,385	125,128
Finland	3,897	197	2,953		0	0	1	747
France	425,011	123,732	138,062		15,733	38,055	2,293	163,217
Ireland	21,219	13,323	5,516		1,665	1,189	126	2,379
Italy	232,416	97,842	75,208		28,909	97,842	143	30,719
Netherlands	124,465	31,734	55,084		2,811	1,438	0	37,647
Total	1,424,025	441,355	401,887		144,547	190,457	15,579	498,450
	(In Percent of Total)
Banks in:	All Countries	Own Country	Other Euro Area	Of Which	Spain	Italy	Portugal	Non-Euro Area
Austria	100	18	24		2	1	0	58
Belgium	100	33	40		5	9	0	27
Germany	100	30	25		3	9	1	46
Spain	100	28	13		28	9	3	41
Finland	100	5	76		0	0	0	19
France	100	29	32		4	9	1	38
Ireland	100	63	26		8	6	1	11
Italy	100	42	32		12	42	0	13
Netherlands	100	25	44		2	1	0	30
Total	100	31	28		10	13	1	35

Source: 2017 EBA Transparency Exercise, and IMF staff calculations.

78. Own-sovereign exposure has come down in 2015–17 and reallocated mainly outside the euro area. Table 5 shows a synchronized reduction in own-sovereign exposures for most euro area banks over the last two years. While some banks have increased their sovereign holdings in 2015–17 (notably France, Germany, Italy, and Spain) they have reduced the exposure to their domestic sovereign (particularly Spain) by taking positions mainly outside the euro area.

Table 5.

Euro Area: Change in Bank Holdings of Sovereign Debt Securities over 2015–2017

Source: 2016 EU-wide stress testing Transparency Exercise, 2017 EBA Transparency Exercise, and IMF staff calculations.

Table 5.

Euro Area: Change in Bank Holdings of Sovereign Debt Securities over 2015–2017

Banks in:	(In Percent of the portfolio)
Banks in:	All Countries	Own Country	Other Euro Area	Of Which	Spain	Italy	Portugal	Non-Euro Area
Austria	0	-8	11		-1	-58	-73	-1
Belgium	7	-12	27		6	-9	0	13
Germany	33	15	25		44	102	237	53
Spain	17	-32	12		-32	17	17	28
Finland	-2	4	-20					841
France	39	15	30		31	46	-4	78
Ireland	6	-2	-1		11	-20	-14	146
Italy	16	-5	8		35	-5	33	8
Netherlands	-6	-18	-12		-20	-63	-100	21
Total	21	-6	14		-16	9	23	41

Source: 2016 EU-wide stress testing Transparency Exercise, 2017 EBA Transparency Exercise, and IMF staff calculations.

Table 5.

Euro Area: Change in Bank Holdings of Sovereign Debt Securities over 2015–2017

Banks in:	(In Percent of the portfolio)
Banks in:	All Countries	Own Country	Other Euro Area	Of Which	Spain	Italy	Portugal	Non-Euro Area
Austria	0	-8	11		-1	-58	-73	-1
Belgium	7	-12	27		6	-9	0	13
Germany	33	15	25		44	102	237	53
Spain	17	-32	12		-32	17	17	28
Finland	-2	4	-20					841
France	39	15	30		31	46	-4	78
Ireland	6	-2	-1		11	-20	-14	146
Italy	16	-5	8		35	-5	33	8
Netherlands	-6	-18	-12		-20	-63	-100	21
Total	21	-6	14		-16	9	23	41

Source: 2016 EU-wide stress testing Transparency Exercise, 2017 EBA Transparency Exercise, and IMF staff calculations.

79. The valuation impact on the sovereign securities portfolio is disaggregated into repricing risk and credit spread risk. The impact on P&L and capital is recognized over the stress test horizon. Although the notional value of exposures evolves with the balance sheet, banks are assumed to maintain the same structure of the portfolio throughout the stress testing horizon.

80. The impact of repricing risk depends on the shock to the risk-free curve and bond duration. Interest rate sensitivity is measured by the effective maturity of the bond proxied by duration. Duration is approximated by the residual maturity of the bond at the cut-off date. Under the full valuation approach, the change in the risk-free bond price is the result of multiplying modified duration by the shock to the risk-free rate:

Δ P_{t}^{i} = \frac{D^{i}}{(1 + r_{t}^{i})} \cdot B_{t} \cdot Δ r_{t}^{i}

where Dⁱ denotes average duration of bond i,B_t denotes the carrying value of the debt security at time t, $r_{t}^{i}$ captures the risk-free rate level and $Δ r_{t}^{i}$ the shock to the risk-free rate.

81. The risk-free curve is proxied by the Bund yield term structure. Figure 6 shows the projected yield curve in 2018 for the Bund. Term premium decompression in Europe and the United States together with the absence of flight-to-quality effects contribute to steepen the Bund yield curve with the 10-year rate rising by 75 bps in 2018 relative to the starting point. A polynomial interpolation method is applied to span the term structure between the 3-month and the 10-year rate generated in the macroeconomic scenario.

82. Term premium decompression together with the re-emergence of sovereign stress raises long term government bond yields further in high spread economies. Figure 6 shows the projected term structure under the baseline and adverse scenario for selected issuers against the yield curve in 2017. The 10-year yield on high-spread sovereign issuers reaches over 3 percent in 2018. For countries with monetary policy space at the starting point, the expansionary policy response assumed in the adverse scenario makes the yield curve pivot downwards mitigating the pricing impact at the short-end (e.g., Czech Republic, United States).

83. The full valuation impact for other than Bund holdings depends on shocks to credit spreads and bond duration. Using the modified duration approach, the FSAP team calculated the haircut for each fixed income instrument under each scenario by multiplying modified duration by the shocks to credit spreads:

Δ P_{t}^{i} = - \frac{D^{j}}{(1 + r_{t}^{i} + c s_{t}^{j})} \cdot B_{t}^{j} \cdot Δ c s_{t}^{j}

where D^j denotes average duration of bond j, $B_{t}^{j}$ denotes the carrying value of security j at time t, $r_{t}^{i}$ shows the risk-free rate, $c s_{t}^{j}$ denotes the credit spread, and $Δ c s_{t}^{j}$ the shock to credit spreads.

84. Although proxied duration is typically low across banks, particularly for traded securities, the relatively larger duration of some banks’ debt holdings and the constrained monetary policy response in some countries exacerbates valuation impact in a few instances. Duration affects valuation through two channels. First, as the curve pivots downwards and steepens under stress, long-duration assets are hit by larger shocks. Second, the sensitivity of the portfolio to shocks increases with the duration of the portfolio. Relatively less monetary policy space in some jurisdictions (e.g., euro area) limits the downward shift of the yield curve increasing the relative impact of term premium decompression and sovereign stress.

Corporate Exposures

85. Some euro area banks hold large positions in corporate debt instruments at fair value in their trading book and AFS portfolio. Corporate exposures include debt securities issued by credit institutions, other financial corporations, and non-financial corporations.³³ Exposures at fair value also include loans and advances held in the trading book. For some banks, the size of corporate exposures at fair value is significant reaching up to 15 percent of the balance sheet.

86. Corporate spread shocks to major corporate yield indices were projected by rating and counterparty. Banks’ corporate exposures at fair value were split according to the credit rating (i.e. investment grade and high-yield), type of issuer (i.e. financial corporations and non-financial corporations), and geography (i.e. Europe, United States). Figure 7 illustrates the projection for the Moody’s seasoned Baa corporate bond yield for U.S. corporate exposures based on statistical analysis during the financial crisis.

87. The average duration was extracted from the discounted expected cash-flow of outstanding bonds in the euro area. The modified duration (mid quote) of all active corporate bonds at end-2017 was computed for the nine home jurisdictions of the 28 banks covered in the exercise. The average duration, weighted by the face value of the bond, was used as a proxy of the duration of banks’ corporate portfolio. Across euro area countries, the average duration of corporate bonds reached 3.6 years in 2017. This duration was adjusted downwards for exposures in the trading book in line with the residual maturity of the sovereign portfolio.

Equity Exposures

88. The FSAP team explored the impact from an equity price correction using complementary metrics. The scenario assumed a broad-based sell-off in stock markets, with equity prices falling by around 20 percent in the euro area (Figure 8). The loss of value on equity instruments arising from changes in market parameters was assessed on trading book transactions and AFS portfolios. Performance metrics included nominal positions (without netting effects), sensitivity (showing local risks), concentration, and holding period. Stressed market losses were benchmarked against the value-at-risk (VaR) and stressed VaR (SVaR) reported in COREP.³⁴ The confidence interval varied between 99.0 and 99.9 percent.

89. For equities held with a trading intent, the fair value impact was floored using the 2018 EBA methodology. The market impact from full revaluation of equity holdings was floored using the following constraint:

Δ E q_{t}^{j} = 1.5 \cdot (- 0.20 % \cdot (E q u i t y_{t}^{l o n g, j} + E q u i t y_{t}^{s h o r t, j}))

where the VaR scaling factor has been set to the upper bound of 1.5, and the trading position includes the fair value of equity instruments (assets) and the short positions in equity instruments (liabilities).

90. The FSAP team used banks’ risk sensitivities to market risk factors to assess valuation impact under stress. The data is not part of regular supervisory reporting but is collected by the SSM in the scope of the STE for market risk benchmarking. The template included the option sensitivities to the price of the underlying (delta), first derivative of delta (gamma), and to changes in volatility (vega).

91. The assessment of equity impact included the hedges used by banks to decrease sensitivities to equity shocks. Hedges for equity risk were assumed effective under both the baseline and adverse scenario. Given data limitations from the STE, the FSAP team did not have access to market risk sensitives for cash and hedging positions separately. Banks that hold large equity trading books, tend to hedge their positions (i.e. they do not typically take directional trades), limiting the extent of the equity correction on asset valuation.

Derivatives

92. Despite the decrease of euro area banks’ wholesale activities, some large euro area banks still hold large derivative books. In line with the post-global financial crisis (GFC) reduction of trading activities, the size of the derivative portfolio has shrunk across euro area banks. However, some euro area banks still hold large derivative books. In terms of market value, the size of the derivative portfolio in some banks reaches over 15 percent with an upper bound of 30 percent of on-balance sheet assets. Most of the derivatives are interest rate derivatives traded over-the-counter (OTC). Some banks active in non-euro area countries also hold FX-derivatives. It is worth noting that whereas some derivatives are classified in the IFRS category ‘held for trading’ some of them are economic hedges used to hedge a specific portfolio (e.g., in countries with fixed rate mortgages).

93. Some existing derivative positions are back-to-back contracts, and other derivatives are position hedging with residual risk. Some euro area banks agree new contracts with reverse features to neutralize their existing derivative positions (back-to-back) therefore the asset and liability side of the balance sheet develops in parallel. Other banks enter into derivative contracts to hedge a position against the risk of adverse market moves. While most of these positions do not reflect directional trading strategies there is some residual risk if an extreme market event materializes. Also, the ability to rollover some of the hedges under stressed conditions is untested.

94. A large number of derivatives traded by large banks do not have quoted prices in active markets. This is because most derivatives incorporate bespoke features or are traded OTC. For these products, fair value is determined using models based on valuation techniques including the Black-Sholes formula. Furthermore, some inputs used in the valuation methods are derived from observable market data while others are based on unobservable inputs (e.g., volatilities, correlations, recovery rate variance).

95. The strategy to assess valuation risk in derivatives included two approaches:

The use of fair-value sensitivities of derivative contracts to market risk parameters including the greeks on changes in equity prices, interest rates, FX rates, and commodity prices. The scenarios included standardized moves to the underlying parameters, including 20 percent shift to equity prices, 100 bps increase in interest rates, 20 percent EUR/USD move, 10 percent increase in FX and equities volatility, and 20 percent increase in rates volatility
The implementation of a reverse stress test on G-SIBs on hard-to-value assets which include many complex or long-dated derivatives (see section on sensitivity tests).

96. Hard-to-value assets include Level 3 and to some extent Level 2 assets. While euro area banks have de-risked their hard-to-value portfolios including through portfolio compression, some euro area G-SIBs still hold material portfolios in Level 3 assets carried at fair value on unobservable inputs. Level 3 instruments include derivatives with longer maturities than those usually trade, with tailored return profiles, or complex derivatives including equity, interest rate, and credit derivatives. The reverse stress test assessed the size of valuation shock on Level 2 and Level 3 assets that would deplete G-SIBs’ capital buffers over SREP CET1 regulatory minimum.³⁵

Capital Requirements Using the Standard Approach

97. The risk assessment included the capital impact from market risk measured under the standard approach. While most of large euro area banks’ capital requirements related to market risk are determined using an internal model approach, the standard approach is also used for some structural FX risk.³⁶ Market shocks on global currency pairs (e.g., GBP, JPY, CHF, CAD, and USD), as well as local currencies in banks’ material geographies were applied to net positions. The traded risk component of the scenario also included market risk from sharp corrections in equity positions and commodity prices for fuel and non-fuel commodities. Euro area banks, however, do not carry material short or long net positions under the standard approach for market risk, mitigating the impact.

E. Interest Rate Risk in the Banking Book Setting the Stage

98. Euro area banks’ net interest income has been compressed relative to peers. While, on aggregate, the net interest margin (NIM) in the euro area has remained steady under the low interest rate environment at around 120 bps, this is low compared to the 200 bps in Australia, 230 bps margin in the United Kingdom, and 315 bps in the United States, although it outperforms the 85 bps in Japan (Figure 9).³⁷ At the same time, deposit rates are close to the zero lower bound, eroding the advantage of banks to tap cheap deposit funding as this is only partly linked to market rates which have become negative at the short-end of the swap curve.

99. In addition, the flattening of the swap curve has constrained banks’ ability to widen the intermediation margin. The net interest margin has two components: the structural margin and the commercial margin.³⁸ While the level of interest rates constrains banks’ ability to apply commercial margins, the slope of the euro swap curve is associated to banks’ intermediation margin.³⁹ The swap curve has flattened from a slope of 180 bps in 2009 to 100 bps at end-2017. A similar pattern is observed with the Bund curve with a peak-to-trough flattening of 210 bps to 100 bps in 2017 (Figure 9).⁴⁰

100. While a scenario in which interest rates rise gradually is likely to support profitability in the medium term, and abrupt and sharp increase in rates can erode margins further. The impact of gradual rising rates in bank profitability was explored in the baseline scenario. Under baseline conditions, the 6-month Euribor is expected to increase by 100 bps by 2021 from a negative rate of 32 bps at end 2017. Whereas an increase in rates might support margins, the overall impact on regulatory capital will depend on the offsetting valuation effects from assets at fair value. On the other hand, an unfavorable adverse scenario in which increasing funding costs outpace the repricing of assets would put pressure on net interest income, if banks are not adequately hedged against higher market rates.

101. Results from the 2017 ECB Sensitivity Analysis of IRRBB suggest that banks are well equipped to deal with rising rates (Box 2).⁴¹ However, the test shows that two key assumptions for the positive impact of rising rates on NIM are the stability of deposits, and the reliability of banks’ loan pre-payment models. Also, while, on average, net interest income would progressively recover with an increase in interest rates, EVE would be hit by a sharp rise of rates. In addition, hedges might become ineffective under stress on the back of basis risk. Basis risk arises typically from the discrepancy between interbank rates and sight and deposit rates. Another source of basis risk is linked to the fact that the reference rate for repricing variable rate mortgages is linked to the Euribor or government yields while the hedging is based on swap contracts.

Box 2.

The 2017 ECB Sensitivity Analysis of IRRBB vs FSAP approach to IRRBB

Both the ECB sensitivity test of IRRBB and the FSAP assessment of IRRBB are based on the supervisory template collected for the short-term exercise conducted in 2017. This is a very rich template showing the repricing gaps for all assets and liabilities, including both on- and off-balance sheet items, with granular maturity buckets ranging from ‘without specified maturity’, ‘overnight’, ‘≪1-month’, through ‘>20-year’. Liabilities are split by type of contract (fixed rate with remaining contractual maturity; floating rate with repricing date), modeling approach, and currency. Instruments include deposits from corporates, retail deposits, deposits from other counterparties, debt instruments, and derivatives.

Results from the 2017 ECB Sensitivity Analysis of Interest Rate in the Banking Book (IRRBB) suggest that banks are well equipped to deal with rising rates. In particular, while the average decline in the EVE is severe under a hypothetical 200 bps parallel upward shift, with a -2.7 percent CET1 impact, on average, net interest income would progressively recover with an increase in interest rates. Under the 200 basis points parallel upward shift, aggregate net interest income would rise by 10.5 percent while under current conditions (the “year-end 2016” scenario), banks’ aggregate net interest income would decrease by 7.5 percent by 2020. While, on aggregate, the use of hedges has shortened banking book duration thus mitigating EVE impact, hedges are costly. The ECB estimated that, on average, the impact of hedges contributes to a +1.7 percent EVE uplift with a negative average contribution to net interest income of -1.1 percent in 2016.¹

The ECB and the FSAP assessment of IRRBB share the same approach:

Focus: Both tests assess the impact of interest rate shocks on net interest income. The aim is to assess the risk to banks’ earnings from adverse movements in interest rates that affect banks’ banking book positions.
Time horizon: Both tests used a constant portfolio structure over the same 3-year time horizon and drew on bank models of customer behavior.

On the other hand, the 2017 ECB sensitivity test and the FSAP exercise differ in a number of ways:

Coverage: ECB’s IRRBB exercise is conducted over a sample of 111 significant institutions while the FSAP test is implemented over the sample of 28 large banks in the euro area.
Balance sheet: While the ECB’s test assumes constant balance sheets as of end-2016, the IMF test assumes dynamic balance sheets with positions evolving in line with macrofinancial conditions.
Impact: the ECB uses the same interest rate scenarios to assess the effect on banks’ capital and earnings. The FSAP test uses the macro scenario-based assumptions to assess the fair value impact from adverse movements in risk-free curves and credit spreads on debt securities. This metric does not reflect the effect of a rate move on the next value of liabilities. At the same time, it uses a separately calibrated interest rate shocks to assess impact on P&L.
Calibration: The ECB test was conducted on six hypothetical interest rate shocks including: ‘the end-2016” curve (low-rates for long); two regulatory shocks: +-200 parallel moves; a steepener scenario with higher long-term rates than short-term rates; a flattener scenario and “the end-2010” curve. By contrast, the FSAP team calibrated separately the funding shock and the lending shock in line with the traded risk component of the scenario. Rate shocks incorporated systematic risk factors (from market moves) as well as idiosyncratic shocks based on banks’ specific asset and liability structure and the historical behavior of rates using econometric analysis.

¹ For the sample of banks, 55 percent of interest derivatives are net payer fixed positions to shorten banking book duration due for instance to long-term fixed rate mortgage loans. Under a payer swap, fixed interest payments are swapped into variable interest receipts. However, 45 percent of interest rate derivatives are net received-fixed swaps.

102. By contrast to the 2017 ECB test, the FSAP assessment of IRRBB followed a four-pronged approach (Box 2):

The ‘economic value approach’, which measures the valuation effect of an interest rate shock on the fair value of assets, and impacts EVE due to long duration (e.g., long-dated fixed rate bonds), was assessed using a market risk approach (see section on market risk). By contrast to the EVE measure, only marked-to-market assets were re-valued rather than all balance sheet positions;
The ‘earnings approach’, which measures the impact of an interest rate scenario on banks’ future net interest income over a 3-year horizon was assessed against a range of interest rate shocks calibrated at the bank/instrument level. This contrasts with ECB’s multiple heuristic approach using six hypothetical scenarios;
The scenario includes a commercial margin and an intermediation margin. In particular, two components are included in the calibration of the interest rate shocks: a systematic risk factor (broadly aligned with the scenario) and an idiosyncratic component (linked to banks rates’ historical behavior);
As in the 2017 ECB test, shocks were applied to the repricing structure of assets and liabilities using the IRRBB template as of June 2017. Impact is measured separately for the stock of loans/liabilities and for the flow of new lending/funding instruments.

Structure of Assets and Liabilities

103. The exposure of banks’ interest earnings to interest rate risk depends on changes in market rates as well as on banks’ structure of assets and liabilities. The impact of a change in market rates on funding costs depends on the liability margin which varies across funding instruments (e.g., sight deposits vs debt securities). The impact of a rate shock on interest income depends on the commercial margin which differs across loan products (e.g., mortgage vs consumer loans). Margins are a function of banks’ business model as well as the level of market competition in product markets.

104. For the sample of banks included in the stress test, exposures to the private non-financial sector (i.e. non-financial corporation and households) represent over two thirds of total loans (Figure 10). Half of this amount is exposure to households, mainly mortgage loans, and half to corporates. Around 15 percent of loans are exposures to financial corporations, 10 percent is loans to central banks, and over 5 percent loans to general governments. This hides however, important differences in the composition of the loan portfolio across banks, in line with their business model.

105. The evolution of banks’ funding structure shows that euro area banks have tended to increase their reliance on customer deposit funding (Figure 11). In line with banks domiciled in other advanced economies, euro area banks have re-oriented their business away from trading and more complex activities, towards less capital-intensive activities, including commercial banking.⁴² This pattern is evident in the changes in banks’ asset portfolios, revenue mix as well as their increased reliance on customer deposit funding.

106. Specifically, in the stress testing sample, retail funding represents over half of interest-bearing liabilities (Figure 12). By contrast, reliance on less stable sources of funding (‘wholesale funding’) is moderate. Customer deposits from financial institutions and debt instruments contribute equally to overall funding at 20 percent. While average bank reliance on central bank funding is contained at 4 percent, this masks wide variability across banks.

107. While the market sensitivity of central bank reserves and retail funding is low, they are subject to considerable uncertainty regarding behavioral assumptions. This is because they are partly motivated by the ultra-low interest rate environment, and therefore, are subject to change if rates start to raise.⁴³ Moreover, the decoupling of rates paid to depositors from market rates has increased the uncertainty of the pricing of these deposits if rates were to rise.⁴⁴

108. Key dynamic effects were captured by the FSAP’s IRRB assessment. Given that banks’ capacity to generate substantial interest income depends significantly on dynamic effects, the FSAP team attempted to model explicitly changes in commercial margins, liability margins, and pass-through effects. Also, non-maturing deposits (e.g., current accounts) where modeled as overnight deposits. On the other hand, the team did not change the maturity profile of maturity positions at renewal, or customer behavior related to shifts between different types of deposits, in line with the constant portfolio nature of banks’ balance sheets.⁴⁵

Modelling Strategy

109. The FSAP team calculated net interest income by estimating line-by-line interest income and expense generated from interest-bearing assets and liabilities. Such an approach is heavily dependent on comprehensive data on individual banks’ repricing maturities and hedging practices. This approach was feasible given the granular data provided in the IRRBB template, covering all banking book items including derivatives. The IRRBB repricing data were mapped to banks’ asset and liability structure, as well as to the implicit interest rate received (paid) on interest-bearing assets (liabilities) using FINREP templates.

110. Two complementary modelling strategies were used to project bank net interest income conditional on the macrofinancial scenarios:

Strategy 1: Projection of front-book interest rates at the individual country and portfolio level;
Strategy 2: Projection of interest rates by loan product and funding instrument at the bank level.

Strategy 1

111. The projection of funding rates was mapped to banks’ breakdown of financial liabilities by product and counterparty using FINREP templates. The following eight portfolio segments were constructed: deposits from central banks, deposits from general governments, deposits from financial institutions, non-financial corporate deposits (overnight; with agreed maturity); household deposits (overnight; with agreed maturity); and, debt securities issued.⁴⁶ These portfolios were mapped to the following pricing paths:

The pricing of central banks deposits from central banks and general governments followed changes to the monetary policy rate. Given the initial zero lower bound, the deflationary environment assumed in the scenario, and the endogenous monetary policy response, policy rates in the euro area remained unchanged under stress;⁴⁷
Rates paid on deposits from financial institutions (including credit and other financial institutions) followed the LIBOR rate path assumed in the scenario;
The pricing of retail deposits was linked to the scenario. The econometric approach included thirty-six regressions for the nine euro area countries and four retail categories (i.e. corporate overnight, corporate term, household overnight, household term);
Bond pricing of debt securities were benchmarked against the euro swap rate curve and a bank-specific spread over the reference rate for new debt issuance.

112. Country-level front-book interest rates on retail deposits were projected using ECB interest rates’ statistics (MIR). Econometric modelling was developed for each euro area country and each portfolio linking the front-book rates to the macrofinancial variables projected in the scenario. Figure 13 shows the evolution of deposit rates for households and corporates in a low-spread country. The chart suggests a structural break in 2012. While corporate rates on deposits with agreed maturity comoved with household rates over 2005–11, there is evidence of decoupling starting in 2012 Q1 with corporate rates declining below overnight household rates entering into negative territory in 2016 Q3 and reaching -4 bps in end-2017.

113. The main drivers of interest rates on retail deposits include macroeconomic variables and market rates. The benchmark empirical strategy relied on a Newey-West time series approach to obtain consistent estimators when the error term is heteroskedastic and autocorrelated. The time series specification takes the general form:

i_{t}^{c, j} = α + δ \cdot i_{t - 1}^{c, j} + β_{1, 1}^{c, j} \cdot m a c r o_{t - 1}^{c} + \dots + β_{1, k}^{c, j} \cdot m a c r o_{t - k}^{c} + β_{2, 1}^{c, j} \cdot f i n_{t - k}^{c} + \dots + β_{2, k}^{c, j} \cdot f i n_{t - k}^{c} + ε_{t}^{c, j}

where each funding cost rate is indexed by country c and instrument j, and the equation includes: an autoregressive component; a vector of macro variables, such as growth, inflation, and credit growth; and a vector of financial variables, such as the swap rate, the sovereign yield spread, the policy rate, the euro LIBOR rate, and the U.S. dollar LIBOR rate. ⁴⁸ Regressors included 1 to k lags to address endogeneity.

114. Results suggest that corporate deposit rates are more sensitive to changes in macro conditions and money market stress while household deposit rates are more closely linked to the policy rate. Annex Table 4 shows regressions for two dependent variables, namely the front-book deposit rate in corporate and household deposit rates with agreed maturity. Results show that corporate rates are more responsive to GDP growth, credit developments, and euro and dollar interbank rates, while household rates are more sensitive to hikes in ECB rates. This result suggests a higher degree of pass-through from policy rates to household deposit rates.

115. Figure 14 shows the projections of front-book rates in a low-rated country based on the adverse scenario. The impact of adverse conditions is more severe for corporate deposit rates. Rates are projected to rise by up to 140 bps by end-2021. The delta for household rates is lower at 110 bps. The 30-bps gap in their starting point matches exactly the differential impact of stress, leading to the same end-point of 140 bps by the end of the horizon.

116. The projection of lending rates was mapped to banks’ breakdown of loans and advances by product using FINREP data. Six portfolio segments were constructed: loans to central banks, loans to general governments, loans to non-financial corporates (including credit institutions and other financial institutions), loans to non-financial corporations, loans to households (mortgages), and other loans to households (consumer loans, other).⁴⁹ These portfolios were mapped to the following lending rates:

Loans to the central banks followed the path for the policy rate. Similarly to the assumption made for ECB’s main refinancing operations, the rate on the deposit facility is assumed to remain unchanged in the adverse scenario;
The lending rate on loans to general governments was based on statistical analysis using MIR (country) statistics and FINREP (bank) returns First, the implicit rate on government loans received by the 28 banks of the sample was extracted using FINREP data. Second, this rate was proxied by a benchmark rate using MIR statistics. The rate that tracked most closely the government rate across euro area banks was the mortgage rate (Figure 14).
The lending rate on loans to financial corporation, non-financial corporations, mortgage loans, and other household loans was projected using MIR data linked to the scenario.

117. The evolution of lending rates across the euro area suggests different lending practices across products. Figure 14 shows the level and dispersion of core lending rates across the nine euro area countries. The chart reveals the following insights: Figure 14: Use min-max or median for the time series—the individual lines are not distinguishable in any case.

Consumer loans carry a large component of credit risk premium. For the same market rate developments, the average rate on consumer loans stood at 5.6 percent in June 2017 relative to a mortgage rate of 2.0 percent and a corporate rate of 1.6 percent.
The dispersion in lending rates for consumer loans is higher than for other loan products. This suggests lower competition in consumer credit relative to the mortgage and corporate segments. While the standard deviation in 2017Q3 was 37 percent for corporate rates and 61 percent for mortgage rates, it reached 150 percent for consumer loans.
Lending rates diverge significantly from risk-free rates. First, the cyclical variability of lending rates is quite limited, particularly for consumer loans. Second, while the average lending rate across product categories reached 320 bps in 2017, the Bund rate was negative at -75 bps.

118. Econometric modelling was developed to forecast lending rates in line with the macrofinancial variables projected in the scenario. Similarly to the modeling strategy for deposit rates, a Newey-West time series approach was followed using lagged regressors under the general specification:

\begin{array}{l} l_{t}^{c, j} = μ + γ \cdot l_{t - 1}^{c, j} + λ_{1, 1}^{c, j} \cdot m a c r o_{t - 1}^{c} + \dots + λ_{1, k}^{c, j} \cdot m a c r o_{t - k}^{c} + λ_{2, 1}^{c, j} \cdot f i n a n c i a l_{t - 1}^{c} + \dots + λ_{2, k}^{c, j} \cdot f i n a n c i a l_{t - k}^{c} + \\ + λ_{3, 1}^{c, j} \cdot r i s k p r e m i u m_{t - 1}^{c} + \dots + λ_{3, k}^{c, j} \cdot r i s k p r e m i u m_{t - k}^{c} + ζ_{t}^{c, j} \end{array}

where the dependent variable is the lending rate by country c and product j. The set of regressors includes: an autoregressive variable; macro factors, such as the unemployment rate and credit growth; financial variables, such as the swap rate, the sovereign yield spread, the policy rate, the euro LIBOR rate, and the U.S. dollar LIBOR rate; and a risk premium proxied by Moody’s EDF for corporates (for the corporate rate), EDF for consumer firms (for the consumer rate), and EDF for real estate firms (for mortgage rates). Lending rate regressions also included developments in house prices. Regressors included 1 to k lags to address endogeneity.

119. Results suggest that lending rates for corporates are more closely linked to wholesale funding costs while consumer rates are driven by risk premia and mortgage rates rise with tight credit conditions. Annex Table 5 presents regressions for core front-book lending rates in a high-spread country, including corporate, consumer, and mortgage loans. Results suggest that lending rates for corporate and mortgage loans are more responsive to cyclical economic conditions, whereas lending rates on consumer credit show higher persistence. Also, they are more sensitive to market conditions suggesting a higher pass-through of bank funding costs to customers. Both factors point at lower competition in consumer credit relative to other product segments. Mortgage rates’ exposure to interbank rates is explained by the prevalence of floating-rate mortgages linked to the 12-month EURIBOR. This effect outweighs house price developments as a determinant of lending rates. The increase of corporate rates with a tightening of credit conditions suggest credit rationing under stress.

120. Figure 15 shows the projections of lending rates in a high-spread country under adverse conditions. During the first year of the stress, the impact is more severe for consumer loans with rates increasing by 110 bps, outpacing the increase in mortgage rates which rise by 60 bps and corporate rates which increase by 30 bps. However, as conditions deteriorate further, stress is transmitted to the corporate and mortgage segments which experience higher increases by the end of the horizon.

121. Funding cost and lending rates projections were applied to banks’ asset and liability structure to forecast shifts to banks’ net interest margins. Figure 16 shows the year-on-year impact of shocks to funding costs, lending rates, and margins. The results show that while funding costs rise sharply the first year of stress, lending rates follow with a lag. While, on average, funding costs rise by 40 bps in 2018, and lending rates adjust partially by 8 bps, banks hit by higher funding costs are on average less able than their peers to pass them onto customers. As a result, at the 90^th percentile, the margin contracts by 75 bps.

Strategy 2

122. This modelling strategy addresses two major developments since the global financial crisis. First, funding market conditions have become heterogeneous across banks reversing a pre-crisis trend of lower and more similar rates. Second, there have been concerns over the policy rate pass-through to lending rates. Possible sources in the inefficient transmission of monetary policy have been attributed to the misalignment of policy rates and funding costs and to banks’ internal transfer policies to set lending rates. Both aspects are examined in this section.

Funding Costs Modeling

123. A key contribution of this approach is to carefully construct an effective measure of funding costs. Interest rates paid on liabilities vary across banks depending on their funding patterns and liquidity management practices.⁵⁰ Given that core funding markets vary across the sample of banks, we expect different rates paid on deposits across banks. Another important determinant of interest expense is banks’ liability margin which depends, among other factors, on the proximity to the zero lower bound.

124. The effective funding cost measure is built by matching the whole range of liabilities that banks use to fund their balance sheet with an interest rate path and a repricing structure:

The pricing of deposits from central banks and general governments follow monetary policy rates which are unchanged under stress in the euro area (similarly to strategy 1);
Funding from deposit liabilities (excluding central bank and general government deposits) is split between financial institutions and customers. Deposit from financial institutions are equivalent to interbank deposits. Interbank rates follow EURIBOR trends;
Customer deposits include deposits from households and non-financial corporations. Deposit rates vary across institutions according to their deposit structure and liability margin. Rates evolve with economic conditions and benchmark rates;
Debt securities for instruments at floating rate are repriced according to the swap rate. They include unsecured bonds (i.e. investment grade, high-yield, medium term notes), secured instruments (i.e. mortgage-backed securities, asset-backed securities, covered bonds), and preferred securities;
The cost of debt securities at fixed rate is split into a reference rate and a spread at issuance. The coverage of instruments is the same as for floating rate instruments. That is, they include unsecured bonds, secured instruments, and preferred securities.

125. The starting point for interest expenses is matched with historical data sourced from Bloomberg, Fitch, and Dealogic. Implied rates by instrument and counterparty at the cutoff date is computed using FINREP returns and mapping banks’ liability structure (F08.01.a) with banks’ interest expense by counterparty (F16.01.a). The time series for EURIBOR rates at different tenors is sourced from Bloomberg. For customer deposits, historical rates are computed using Fitch data. For debt securities, the cost of issuance is obtained from Dealogic. Bonds are issued at the swap rate plus a mark-up over swap rate (spread).

126. It is essential to consider all sources of market funding given the behavior of issuance and yield differential between secured and unsecured debt. Figure 17 shows gross issuance of bonds by the 28 banks of the sample broken down by instrument, region, and bank from 2005 to 2018 Q1. The data reveal the following patterns:

Over the sample period, unsecured issuance accounts for around 60 percent, mainly investment grade bonds, while 40 percent represents secured issuance split between ABS, MBS, and covered bonds. Issuance of preferred equity is insignificant;
Aggregate issuance masks a marked cyclical pattern with quarterly issuance drying out during both the financial crisis (2008) and the sovereign crisis (2011Q3–2012Q2) to a quarterly average of €50 billion from a pre-crisis level of €130 billion;
As the crisis attests, when bond spreads rose, gross issuance of senior unsecured bonds fell to near zero in the peripheral countries being replaced by covered bond issuance;⁵¹
Alternative methods of funding carry very different cost of issuance, with the average spread of unsecured bonds at 130 bps more than doubling the 60-bps spread of secured bonds;⁵²
There is a wide dispersion of cost of issuance due to bank- and country-specific factors. For unsecured bonds, the spread at issuance ranges between 15 and 90 bps at the first and third quartile, respectively, while secured instruments’ spread fluctuates between 105 and 170 bps.

127. Crucially, this approach uses banks’ actual funding spreads, which allows accounting for banks’ behavioral response to stress. Spreads at issuance represent more closely the actual cost that banks face when they raise wholesale funding than the proxies used by the empirical literature. Common proxies include the 5-year CDS spreads and the secondary market spread for 5-year euro senior unsecured bonds. As Figure 18 illustrates, CDS spreads are not necessarily representative of actual funding costs, as banks tend to refrain from unsecured issue when market spreads widen significantly. Therefore, an analysis of funding costs based on market proxies is likely to generate off-equilibrium path results

128. The modeling strategy is based on estimating bank funding costs by instrument using two econometric approaches:

A panel based regression using fixed effects. To make sure that the estimation if robust to this choice we also run regressions at the country level using fixed effects.
A quantile regression approach to assess the impact of underlying drivers at different percentiles of the funding cost distribution.

129. The basic form of the relationship between funding cost and the explanatory variables is as follows:

i_{t}^{i, j} = α + δ_{1}^{j} \cdot i_{t - 1}^{i, j} + δ_{2}^{j} \cdot m a c r o_{t - 1}^{c, j} + δ_{3}^{j} \cdot f i n_{t - 1}^{c, j} + δ_{4}^{j} \cdot g l o b a l_{t - 1}^{j} + ε_{t}^{i, j}

where $i_{t}^{i, j}$ denotes the funding cost of instrument j for bank i. ⁵³ To capture broad changes in the state of the economy and shifts to benchmark rates, the set of regressors includes macro variables (growth, inflation, unemployment, recession indicator), financial variables (EURIBOR, swap rates, spreads over swap), and regional/global factors (growth in the EA, ted spread).⁵⁴

130. Results suggest a high degree of pass-through from changes in policy rates to customer deposit rates. For the pooled regression, the elasticity of deposit rates to changes in the ECB’s main refinancing rate reaches 0.2 (Annex Table 6). There is also evidence of some cyclical behavior of rates with an economic expansion in the euro area and tighter money market spreads contributing to higher deposit rates. On the other hand, sovereign stress is not associated with moves in deposit rates. This analysis does not include the effect of the potential downgrade to banks’ credit ratings from changes to the potential uplift from government support which was observed during the European sovereign debt crisis.

131. Debt spreads on bank issuance are mainly driven by sovereign spreads, ted spread, and bank solvency. Econometric results shown in Annex Table 7 are consistent with the sovereign-bank nexus whereby higher domestic sovereign stress pushes up the cost of raising market funding by reducing the sovereign rating uplift. A shortage of global liquidity, proxied by a widening of the TED spread, also raises the cost of debt issuance. The elasticity of bank equity to debt spreads is negative and statistically significant.

132. Quantile regression results show that the negative contribution of sovereign stress to the cost of debt issuance rises with sovereign stress. Figure 19 shows the coefficients of the main drivers of wholesale funding costs by quartile. At higher quartiles, the coefficient of sovereign spreads on funding costs is economically and statistically more significant than at lower quartiles pointing at an amplification mechanism from sovereign distress to stress in the banking sector through market issuance.

133. The adverse scenario is significantly more severe on the cost of debt issuance than on shifts to deposit rates. Figure 20 shows the box plots that illustrate the distribution of multipliers for projections of deposit rates and debt spreads over the stress test horizon relative to the starting point. While deposit rates rise gradually by 15 percent, 42 percent, and 56 percent over 2018–2020, debt spreads increase by up to 240 percent by 2020. This is consistent with the higher risk sensitivity of wholesale funding relative to core retail deposits.

Pass-through Modeling

134. By contrast with the empirical literature, the pass-through to lending rates was estimated using banks’ actual funding rates as well as their asset and liability structure. Loan and deposit pricing relationships are often estimated averaging variables and estimating aggregate time series at the country level. Other approaches include the estimation of net interest income at the individual bank level (both approaches are reflected in STAMP€, 2017). The relationship between bank lending rates and funding costs are typically estimated as a mark-up over a proxy rate rather than using banks’ actual funding costs.⁵⁵ A notable exception is BIS (2015) where the relationship between lending rate and market rates are based on banks’ effective funding costs taking into account banks’ liability structure.⁵⁶ However, this approach compares funding costs against aggregate lending rates on mortgage and corporate loans rather than on bank’s actual lending rates on their lending portfolio.

135. The relationship between lending rates and funding costs is as follows:

l_{t}^{i} = α + γ_{1} \cdot l_{t - 1}^{i} + γ_{2} \cdot s t a t e {var}_{t - 1}^{c} + γ_{3} \cdot f u n d i n g_{t - 1}^{i} + γ_{4} \cdot b a n k_{t}^{i} + γ_{5} \cdot f u n d i n g_{t - 1}^{i} \cdot b a n k_{t}^{i} + ε_{t}^{i}

where $l_{t}^{i}$ denotes the implicit lending rate for bank I. The equation includes an autoregressive coefficient and controls for the state of the economy reflected in the scenario. The coefficient γ₃ shows the pass-through of funding costs to lending rates.

136. The empirical strategy includes panel-based regressions as well as quantile regressions. Analogous to the modeling approach used to project funding costs, a panel based regression is performed using fixed effects. Results are shown at the aggregate level for the entire pool of banks as well as broken down at the home country level. A quantile regression approach is used to assess the impact of underlying drivers at different percentiles of the lending rate distribution.

137. There is mixed evidence of pass-through effects across countries. The relationship between funding costs and lending rates is summarized in Annex Table 8. First, the autoregressive coefficient suggests the relevance of banks’ back book of loans made in earlier periods. Second, results show that the state of the local economy affects banks’ lending rates with stronger GDP growth and wider sovereign yields contributing to higher lending rates. There is mixed evidence on pass-through of bank-specific funding costs to customers. While this effect is significant for countries 1 and 2, there is no evidence on interest pass-through in other countries after controlling for financial conditions.

Impact on Net Interest Margin

138. Using ECB’s IRRBB data template, maturity gaps by product, maturity bucket, and currency were constructed. For funding costs, the maturity buckets include instruments without specified maturity, overnight, ≪1-month, 1–3-month, 3–6-month, 6–12-month, ...,1-2-year,...,>20-years. Maturity buckets were constructed separately for deposit and debt securities as deposits tend to reprice at shorter tenors than debt securities. The coverage of securities includes fixed- and floating-rate instruments denominated in euro and non-euro. Maturity gaps are computed according to banks’ modelling assumptions and to contractual conditions. Debt instruments at floating rate were repriced at the market-based swap rate. Debt securities issued with remaining contractual maturity within the maturity bucket were re-issued at the projected swap rate and spread over swap, capturing both systematic and idiosyncratic risk.

139. The impact of interest rate shocks on banks’ net interest income was calculated using Strategy 2 as:

N I I_{i, t} = \sum_{b} g a p_{i, t}^{b} \cdot m i d_{i, t}^{b} \cdot Δ n i m_{i, t}^{b}

where $g a p_{i, t}^{b}$ is the gap of bank i in bucket b and time t, mid^b is the mid-point in bucket b, and $Δ n i m_{i, t}^{b}$ is the net interest margin shock for bank i at time t.

140. On average, margins contract by over 20 bps in 2018 with banks at the bottom quartile seeing their margins decrease by 35 bps. Figure 21 shows the year-on-year impact of adverse conditions on, funding costs, lending rates, and net interest margins funding costs, lending rates, and margins. By contrast with the results shown for Strategy 1, accounting for the empirical estimation of pass-through effects alleviates the negative impact of stress on banks’ net interest margins.

F. Other Profit and Loss Items

141. To compensate for compressed net interest margins some banks in the euro area have adapted their business models, moving towards more fee and commission-(F&C) generating activities. To offset the expected decline in net interest income related to the current low interest rate environment, some euro area banks have started to gradually shift income generation towards business generating fees and commissions. Net fees and commission income includes income from fiduciary activity, deposit fees, securities transfer orders and issuances, asset management services, and sales of third parties’ investment products.

142. At the same time, there are concerns over the financial stability implications from greater reliance on F&C.⁵⁷ Some banks have expanded their asset management activities and sales of shares of collective investment funds. However, the competition on markets such as those for investment products, where nonbanks are active too, is strong and growing. Even if banks were able to markedly increase fee and commission income, the higher degree of income diversification might not improve the stability and resilience of banks’ overall net income. This would hinge on how resilient F&C income is to adverse developments in financial markets.

143. Banks’ reliance on fees and commissions varies across business models. There are substantial differences in the degree to which the 28 banks rely on F&C income. Within the stress testing sample, the share of F&C to total assets of the average bank reached over 50 basis point in 2016. There is a wide dispersion in the distribution with universal banks relying on F&C to a greater extent. The range varies between one- and two-thirds of a percentage point in the first and third quartile of the distribution.

144. The empirical strategy relied on the estimation of a dynamic panel data model over the pool of banks:

f c_{t}^{i} = α + μ_{1} \cdot f c_{t - 1}^{i} + μ_{2} \cdot m a c r o_{t}^{c} + μ_{3} \cdot f i n_{t}^{c} + ε_{t}^{i}

where $f c_{t}^{i}$ is the share of fees and commissions to total assets for bank i at time t and the vector of explanatory variables. The model is estimated using the system Generalized Method of Moments (GMM) estimator developed by Arellano and Bond to address endogeneity and dynamic panel bias.

145. On the back of the linkages between F&C income and financial markets, the set of predictors include a wide range of financial conditions.⁵⁸ The set of variables comprises the lagged F&C over assets, contemporaneous and lagged GDP growth, the inflation rate, house price growth, stock market returns, the first difference of the short- and the long-term rate, equity volatility, government spread over swap, and TED spread.⁵⁹

146. Results show that the share of fees and commission increases with a flattening of the yield curve and equity volatility. Annex Table 9 shows that the shape of the yield curve definitely affects banks’ reliance on fees and commission income. This is reflected in the positive coefficient of the short-term rate difference and the negative coefficient of the long-term rate difference. This result suggests that to compensate for the decline in net interest income associated with the flattening of the yield curve, banks have started to gradually shift income generation towards business generating fees and commissions, instead of interest flows.

147. Other P&L items were projected drawing on a panel-based estimation using fixed effects. The time series data was sourced from Fitch and applied to bank supervisory returns drawing on FINREP data. The main variables included:

Dividend income;
Other operating income;
Non-interest expenses including administrative expenses, depreciation, negative goodwill and other operating expenses;
Tax expenses over net profits.

148. The stress test analysis does not take into account one-off adjustments. One-off adjustments reported in banks’ income statement are considered to be non-recurrent events. Therefore, their P&L impact is zero under both baseline or adverse conditions. The list of one-off events includes profit and loss from non-current assets, discontinued operations, and disposal groups, as well as extraordinary profit.

149. Dividend payout behavior, which is crucial to a bank’s ability to recover from shocks, was linked to its profits and bank capital ratio. The dividend payout ratio was subject to the following constraints: (i) A floor set at 30 percent (positive profits); (ii) the capital conservation buffer schedule for CET1; and (iii) statistical analysis on historical dividend payments. If profits exceeded 50 percent of income at the cut-off date, the dividend payout ratio was set at 50 percent.⁶⁰ If banks were making losses, a zero-dividend payout was assumed.

G. Solvency Results

150. Under the baseline scenario, there is a modest increase in capitalization ratios, reflecting supportive cyclical conditions. Baseline projections imply a weighted-average 40 basis point increase in banks’ CET1 ratio at the end of the 3-year horizon as banks continue to build capital through retained earnings and as capital requirements for credit risk decrease (Figure 22). The results differ, however, across business models. Large, but less complex, internationally-active banks benefiting most from improved macroeconomic conditions. These banks experience larger improvements to asset quality which contributes to RWA contraction. In addition, they benefit from an increase in their lending margin in core markets as well from larger expansion in their loan books boosted by credit demand across geographies.

151. Under the adverse scenario, although minimum capital requirements are met in aggregate, a few banks are significantly more vulnerable. Macroeconomic shocks and trading book risk stress are the main determinants of outcomes, but with some degree of variation across banks due to differences in business models, risk exposures, legacy portfolios, and banks’ efforts to build up provisions for non-defaulted exposures (Figure 23).⁶¹

152. Large, but less complex, internationally-active banks are relatively more resilient on average.⁶² The comparatively greater impact on G-SIBs partly stems from their large trading books and wholesale funding structures. Domestically-oriented banks are less able to diversify credit risk owing to more limited cross-border activities. Even within the broad business model groupings, bank results differ substantially across G-SIBs and domestically-oriented banks due to differences in their geographic exposures and sensitivity to market shocks. Comparable results are obtained using alternative solvency and profitability metrics (Figure 24).

153. The impact of the adverse scenario on banks’ initial capital positions mainly reflects large impairment charges, increased risk-weighted assets, and traded risk stress (Figure 25):

Weaker global growth, falling real estate prices, and tighter financial conditions increase credit risk and erode collateral values, reducing the aggregate CET1 ratio by 3.0 percentage points relative to end-2016.
Heightened credit risk (partly driven by PiT shifts to risk parameters), valuation effects in FX exposures, and the projected paths for aggregate lending in the stress contribute to an expansion of RWAs, thereby lowering the CET1 ratio by around 2.1 percentage points across banks.⁶³
Sharp movements in market prices result in substantial losses for banks with large trading books and AFS holdings. These losses are concentrated in 2018 before partially unwinding as asset prices recover. By the end of the stress test horizon, valuation losses decrease the CET1 ratio by 1.0 percentage points on average.
The aggregate effect of interest rate risk on net income is moderate for most banks, resulting in CET1 ratio which declines by around ½ percentage point. This effect reflects the large stock of customer deposits, which are less sensitive to money market stress; hedging practices to manage duration risk; the repricing structure of debt instruments with longer tenors; and banks’ ability to partly pass-on higher funding costs to customers.
Pre-provision net revenue including aggregate NII, non-interest income, and non-interest expenses, increases the aggregate CET1 ratio by 3.1 percentage points relative to the starting point.⁶⁴

154. The stress test results suggest that balance sheet metrics of strength and market-based indicators are insufficient to predict banks’ financial strength under stress. The stress testing exercise yields additional insights into banks’ future financial strength under stress conditions which are not captured by current financials.⁶⁵ Figure 26 shows the lack of co-movement between banks’ balance sheet metrics, including regulatory capital, NIM, NPLs, with capital depletion under stress. While higher initial capital buffers contribute to stronger post-stressed capital ratios, a lack of co-movement prevails for alternative metrics of balance sheet strength. Whereas CDS spreads do not discriminate banks under current benign market conditions, they incorporate some forward-looking information under stress in some cases.

H. Sensitivity Tests

Single Factor Tests

155. Scenario-based analysis is accompanied by a range of sensitivity tests to address the diverse risk factors identified above (Figure 27). It is important to underscore that these sensitivity tests are distinct, but complementary to the scenario-based approach. These tests explore (1) the additional losses resulting from replacing selected adverse scenario inputs with hypothetical shocks capturing extreme changes in risk parameters, and (2) banks’ sensitivities to a low-for-long interest rate environment over the 3-year horizon:

Interest rate risk: Replacing the shift to risk-free curves and the model-based funding and lending rate paths under the adverse scenario with a hypothetical 200 basis points parallel upward shift reduces the aggregate CET1 by 200 bps. This test incorporates the impact of hedges to mitigate IRRBB but excludes fair-value hedges on financial assets. G-SIBs are relatively more impacted given the size of their trading portfolio;⁶⁶
Further decompression of risk premia: A 200 bps widening on own (not all) sovereign spreads over the swap curve lowers the post-stress aggregate CET1 ratio by an additional 30 bps; some less international banks are affected relatively strongly. ⁶⁷ A widening of spreads on corporate bonds (including bonds issued by financial institutions) similar to that observed during the 2008 leads to a further 175-bps reduction in CET1 capital; the valuation losses more than offset higher earnings.⁶⁸ G-SIBs are more severely affected given their larger corporate bond exposures;
Credit rating downgrade: A one notch-credit downgrade (on standardized exposures) erodes bank capital ratios by 60 bps. Domestically-oriented banks are impacted more severely on the back of their exposure to credit risk corrections and higher reliance on external ratings;
Tighter LGD floor on mortgage loans: An LGD floor of 30 percent on residential mortgages in their home (headquarter) jurisdiction reduces the aggregate CET1 ratio by around 40 bps. The effect is larger for smaller banks due to their larger share of domestic mortgages;
Low-for-long: Under the current low interest rate environment, most banks would suffer a hit to their net interest income, resulting in an average CET1 reduction of 85 bps at the end of the three-year horizon.

Hard-to-value Assets

156. A sensitivity-based reverse stress test on hard-to-value assets was performed. There are concerns that the book valuation of some hard-to-value assets can be kept artificially high, leading to possible excessive risk exposure. At the same time, supervisors have turned a spotlight on banks’ riskiest assets and induced banks to reduce their holdings of hard-to-value investments.

157. The calibration of the shock was linked to the opacity of the instrument and the fair value category. Instruments that are more opaque (e.g., complex derivatives) were hit harder than other bespoke trades (e.g. reverse repos). Also, Level 2 assets, which are valued using inputs other than quoted prices due to insufficiently market liquidity or traded over-the-counter were subject to less severe shocks than Level 3 assets which use inputs that are not based on observable market data (referred to as unobservable inputs).⁶⁹

158. The results show that even a modest valuation shock on Level 2 and Level 3 assets would have a significant impact on some G-SIBs capital buffers. A combination of 10 percent valuation shock in Level 3 assets and 5 percent in Level 2 assets could deplete the capital buffers of some large banks over the SREP regulatory minimum (including the phased-in G-SIB capital surcharge, Pillar 2 requirement, and SRC buffer). This is a conservative estimate as stress test impact has not been adjusted for valuation reserves.

Liquidity Risk Analysis and Stress Tests

A. Overview

159. The liquidity risk management and liquidity conditions have significantly improved in recent years. The GFC raised awareness for liquidity risk in the banking system and reduced bank managements’ and bank shareholders’ liquidity risk tolerance. Regulators and supervisors also became more sensitized to liquidity risks. The introduction of the LCR and the NSFR (likely as of 2019) has improved liquidity regulation and the availability of reporting data across the euro area (EA). The SSM has introduced a comprehensive set of standardized analysis tools, including liquidity stress tests based on supervisory data. At the same time, accommodative euro area monetary policies have led to high central bank excess reserves.

160. To assess current banking system liquidity risks, a comprehensive analysis of large banks’ structural liquidity ratios is complemented with a variety of liquidity stress tests. The structural analysis considers the Basel III LCR and the NSFR. While the former measures short-term liquidity risks, the latter ratio gauge more structural longer-term refinancing and funding risks. The FSAP team did not stress structural LCR/NSFR ratios as it is unlikely that during the times of stress these will be binding (i.e. bank must meet average LCR requirement over the calendar month time horizon, rather than daily) and focused on cash flow based stress tests instead. Cash flow-based liquidity stress tests were conducted using supervisory data on contractual cash flows for different maturity buckets. This approach employs multiple scenarios of increasing severity covering several horizons (for example, 4 weeks, 3 months) with varying assumptions regarding liquidity buffers and shocks to cash inflows and outflows. Also, a “collateral freeze” scenario was simulated, wherein the collateral held at CCPs and available for rehypothecation remains inaccessible for five business days due to a cyber risk event.⁷⁰

161. To deal with parameter uncertainty, the cash flow tests were conducted over a wide range of scenarios featuring different degrees of severity and central bank (CB) support.

The calibration of the liquidity stress test drew on the assumptions built into the solvency stress test to ensure consistency among both tests. For example, stressed market values of securities or markets’ reaction towards banks’ ability to raise funding after drop in capital ratios. The tests also incorporated assumptions about gradual tightening of monetary conditions, such as changes in eligible collateral used to obtain liquidity from the CB, and changes in interest rates.⁷¹

162. Liquidity risk analysis used multiple data points to highlight structural changes in banks’ funding structure. Multiple data points (from Q1 2014 to Q3 2017) from supervisory data templates allowed the FSAP team to deep dive into structural liquidity risks of sample banks, in particular to assess how liquidity risk bearing capacity of banks changed through the four-year time period.

B. Structural Liquidity Risks

Liquidity Coverage Ratio (LCR)

163. Liquidity analysis reveals that all banks in the sample meet the 100 percent minimum LCR requirement. LCR ratios are comfortably above minimum requirements, highlighting large liquidity buffers banks accumulated since the GFC, including long-term funding from the ECB (Figure 28). Shorter maturity transformation by banks also played a role. Across business models, domestically-oriented banks have higher LCRs, partly due to lower reliance on short-term wholesale market funding but higher reliance on long-term CB funding.

164. Short term overnight funding is a source of liquidity risk during times of market turbulences. Overall overnight funding is close to 22 percent (Figure 28), but some G-SIBs and some large, but less complex, internationally-active banks have even higher share of short-term unsecured funding from financial corporations. This makes those banks more vulnerable to sudden market-wide liquidity dry-outs.

165. Retail funding, treated as more stable from the LCRs perspective, in the sample of banks is 37 percent. While retail deposits are the core and stable base of funding for banks, it is typically very short term, especially as customers do not want to lock-in to low or zero interest rates. 74 percent of those retail deposits are insured, bringing the total share of insured deposits to 28 percent of total liabilities. From a funding risk perspective, deposit insurance provides additional stability by lengthening behavioral terms of these deposits, but is an implied contingent liability for the sovereigns or the banking system as a whole, which, in our sample of banks, amounts to €3.6 trillion.

166. Smaller domestically-oriented banks in the sample are very diverse, with some of them having less than 10 percent of retail deposits as a source of funding. At the same time, those banks typically have higher proportion of secured long-term funding (long-term repos with CB, covered bonds, and Asset Backed Securities, ABS). This group of smaller domestically-oriented banks would be vulnerable to liquidity risks if the market liquidity of secured bonds were to freeze.

167. Foreign exchange funding, particularly for the U.S. dollar, remains a source of vulnerability for selected banks. The U.S. dollar and the British pound are the most important foreign currencies that the sample of 29 banks are exposed to. There is no formal requirement to maintain 100 percent LCR for these currencies. In fact, LCRs for these currencies are volatile and fall substantially below 100 percent for many of the banks (Figure 28). Banks appear to rely on amply U.S. dollar liquidity and the backstop of CB swap lines in the case of more turbulent market conditions.

168. The use of collateral swaps does not reveal significant liquidity risk in the sample of banks. Collateral swaps may be used to meet the regulatory liquidity requirements, including the LCR, at the end of reporting periods. For example, such a transaction would allow a bank to borrow liquid assets in exchange for a less liquid assets (from another financial institution such as insurance company that would receive a fee). Almost half of the banks in our sample engage in such transactions, however the amounts were not material relative to total LCR requirements.

NSFR and Core Funding

169. Banks are well placed for the introduction of 100 percent NSFR requirement (Figure 29). Most of the banks report NSFR ratios above 100 percent, and aggregate stable funding needs are just short of €14 billion (a surplus in banks which report NSFR above 100 percent reaching €950 billion), but this is mainly driven not by the stock of long term market funding (exceeding one year), but by the large amounts of highly liquid short-term assets (sovereign bonds, deposits within central banks) and long-term repos with central banks.⁷²

170. GSIBs rely on short-term wholesale funding to a greater extent than do other banks. The core funding ratio (CFR) measures structure of liabilities in terms of stable relative to total funding.⁷³ CFRs tend to be high for banks engaged in traditional banking activities (like many of the large, but less complex, internationally-active and relatively smaller domestically-oriented banks in our sample) and low in banks which engage in complex trading and market making activities. G-Sibs which have complex balance sheets have lower CFR compared to their peers. While short-term funding obtained from financial counterparties is typically invested in short-term and liquid assets, funding gaps in terms of maturity mismatches may not be fully covered by available CBC. Recent episodes of banking distress highlighted the loss of commercial funding (retail, wholesale deposits) of more than 40 percent in one year.

Funding Concentration and Dependence on Central Bank Funding

171. Overreliance on one or few funding sources is risky. Diversification and lengthening of funding entails increasing share of longer-term market funding (ABS, long-term covered bonds) with diversified refinancing maturities would be more expensive than short-term wholesale and retail funding, but would greatly reduce wholesale deposit roll-over risks.

172. Funding concentration has increased in the last few years. The domestically-oriented banks display the largest relative dependence on a few funding providers (Figure 30). The share of the ten largest funding providers has gradually risen since 2015, reaching 10 percent on average in 2017Q3. This uptrend is mostly attributed to the increase in CB funding, with domestically-oriented banks in the lead. In contrast, G-SIBs have more diversified funding sources. The prevalence of FX funding concentration, especially U.S. dollar, adds an additional layer of risk.

173. Smaller domestically-oriented banks’ dependence on CB funding reflects their business model and challenges in obtaining cheap funding in domestic markets. To some extent large borrowing from CB can also be assumed as a way to build potential precautionary liquidity buffers (for example, possibly motivated by the recent bank resolution cases). Large G-SIBs as well as less complex internationally-active banks have more options to diversify funding as well as obtain cheaper market funding from abroad. Funding from CB also helps banks to maintain higher average interest margin, and last, but not least, risk premiums for long-term debt issuance for banks might differs due to differences in sovereign risk premiums among sample countries; at the same time funding from CB is provided under uniform policy rates while liquidity obtained depends on the quality of collateral.

174. Central bank funding has helped banks meet their regulatory liquidity requirements. While repos with private counterparties are typically short-term, CB funding provides banks with funding which exceeds one year (and typically goes up to five years). This increases LCR and NSFR ratios and allows banks to manage short and long-term liquidity risks. Moreover, ability to encumber assets with the central banks (especially under the full allotment auctions) provides banks with ability to obtain cheap funding amid accommodative financial conditions.

175. Tightening of financial conditions would affect sample of banks unevenly, and mainly via two channels:

Banks with higher share of central bank funding would experience relatively higher increase in funding costs. For domestic banks, replacing the respective amount of long-term CB funding with the 2-year senior unsecured bond would lead, on average, to 10 percent higher interest expenses.⁷⁴ G-SIBs and internationally active banks would be much less affected by the changes.
Long-term funding availability may have an effect on LCR and NSFR ratios, provided that banks would not be able to use less liquid assets (such as credit claims) to obtain market liquidity. On average, the effect in volume would not be large (Box 3). It should be noted however, that change in CB collateral framework and monetary policy operations may also affect market liquidity of the various types of securities, and some wider divergence among sovereign and corporate CDS spreads could be observed, which could exacerbate the direct funding cost effect.

176. For some banks, funding costs will increase. For weak banks, this could lead to a spread shock: the introduction of the BRRD might reinforce the interaction between capitalization and funding costs. An illustrative exercise suggests that the impact on funding costs from a greater reliance on market funding differs: with some smaller, less complex domestic banks affected much more than G-SIBs or internationally active banks (Figure 30). It was assumed, that banks will replace CB funding with market funding (issue 2-year senior unsecured bonds). The interest rate is determined as follows: baseline scenario forecast of 2-year swap rate plus current individual risk premiums for each bank over the respective 2-year swap rate.

Box 3.

Potential Impact of the Eventual Normalization of the of Quantitative Easing (QE) and the Discontinuation of the TLTROs on System-wide Funding Liquidity Risks

A hypothetical normalization of monetary policy would have multiple consequences for the system-wide funding liquidity risk of the EA banks at the system level. The direct and indirect effects on bank funding costs are more difficult to assess and warrant more supervisory attention, as they depend on the future balance-sheet structure of banks (asset quality and capitalization) and on market sentiment.

Some possible consequences can be identified, albeit based on strong assumptions and neglecting many factors, such as the speed of normalization. We apply comparative statics at the system level on the purely heuristic assumption that QE is unwound completely and the TLTROs are discontinued. In practice, this could take many years providing banks with time to adjust. Another strong assumption is that, in the course of the unwind, the same institutions that had sold securities to the EuroSystem under QE, would replenish holdings after QE. They do not buy directly from the EuroSystem, though, but buy the securities that the ultimate debtors of the EuroSystem, such as sovereigns.

The analysis of the impact of monetary policy normalization on the liquidity situation of the banking system needs to consider the effects on both sides of the balance sheet, i.e. assets and liabilities. We divide our analysis in three parts: impact of changes in QE on assets (i.e. counterbalancing capacity of banks) and liabilities (i.e. deposits banks received from nonbank financial institutions), and impact of unwinding of TLTROs.

In December 2017, QE amounted to about €2,300 billion and the outstanding TLTROs to around €750 billion. The QE purchase amounts can be clustered into the following components:

i) The Public-sector purchase program (PSPP) account for the lion’s share with about €1,900 billion;

ii) the Covered bonds purchase program (CBPP) of €240 billion;

iii) the Corporate sector purchase program (CSPP) for €131 billion and the

iv) ABS purchase program (ABSPP) for €25 billion.

Table 1.

The Direct Impact of the Unwind of QE on Banks’ Counterbalancing Capacity (CBC) Under Current Market Prices

Sources: http://www.ecb.europa.eu/mopo/implement/omt/html/index.en.html. An ECB analysis (ECB Economic Bulletin, Issue 4/2017, pp. 61) found that euro area banks sold 25 percent of the government securities bought by the EuroSystem. The banks in the sample account for about 60 percent of the EA banking system, so that their share in the QE purchases is about 15 percent. Euro area nonbank financial corporates (insurance companies, investment funds, and pension funds) account for about 30 percent and non-euro area residents the remaining 45 percent. We assume the similar proportions apply to the CBPP, CSPP, and the ABSPP. The average haircuts in the BCC are identical to the haircuts in the liquidity stress test (section C.).

Table 1.

The Direct Impact of the Unwind of QE on Banks’ Counterbalancing Capacity (CBC) Under Current Market Prices

Program	Volume (in billion euro)	Share of banks in the sample (in billion euro)	Assumed average haircut in CBC	Impact on banks’ CBC
PSPP	1900	285	4%	0.2%
CPBB/CSPP/ABSPP	396	60	7%	0.07%

Table 1.

The Direct Impact of the Unwind of QE on Banks’ Counterbalancing Capacity (CBC) Under Current Market Prices

Program	Volume (in billion euro)	Share of banks in the sample (in billion euro)	Assumed average haircut in CBC	Impact on banks’ CBC
PSPP	1900	285	4%	0.2%
CPBB/CSPP/ABSPP	396	60	7%	0.07%

Unwinding QE would change the composition of balance sheets but also affect the market liquidity of securities and the availability and composition of CBC in some smaller, less complex domestic banks.^1/ In that the ECB decreases its stock of securities and banks run down cash balances to buy more securities, the overall decrease in CBC capacity would be relatively small and depend on two effects: (i) haircut on swapped assets and (ii) market illiquidity of instruments which ECB accepts as a collateral.

From the accounting perspective, asset reallocation reduces the CBC of the system by the average haircut on highly liquid securities. In the case of 29 banks, data available from prudential returns suggests that the haircut is about 3 percent on HQLA (mostly sovereign bonds). Assuming that the banks account for 60 percent of the €360 billion volume proportionate to their share in the EA banking system, the reduction of their liquid CBC (under today’s market prices of EA sovereign bonds) would amount to approximately 6 billion euro, or 0.2 percent of the total CBC of the sample.

The haircut effect on less liquid instruments is larger, but the amounts of those securities pledged to the ECB are relatively small. In case of covered bonds (CBPP): the share of the banks in sample is about 60 percent of the €35 billion sold to ECB by EA banks. The average haircuts are about 4 percent, so that the impact of the asset substitution would be negligible (0.04 percent of the CBC). The ABSPP and CSPP of the sample banks amounts to about €23 billion; even though the haircuts are higher (6 to 7 percent), the impact on the CBC is also negligible.

From the market liquidity perspective, absence of ECB as a significant buyer of sovereign and high quality corporate bonds would have an impact on trading volumes and liquidity risk premiums. Based on GFC and sovereign debt crisis episodes, bid-ask spreads as well as banks’ ability to liquidate lower quality, less liquid assets may be severely hampered (this assumption was utilized in the cash flow liquidity stress tests under marketable only CBC).

Nonbank financial institutions (NBFIs) accessed QE program via banks. The EuroSystem purchased 16 percent of government securities from investment funds and a further 14 percent from insurance companies and pension funds. It bought the remaining 45 percent from non-EA residents. In general, these counterparties do not have access to the EuroSystem books and hold the cash they receive as deposits with banks, which, in turn, have to hold it as deposit at the EuroSystem.

Scaling down the QE would lead to reduction of NBFIs’ deposits at banks. If they shrink their balance sheets in line with their lower holdings of QE related securities, households’ deposits in banks increase. Consequently, unwinding QE would not only reduce the CBC of the banks but also largely reduce the €1,400 billion of deposits of NBFIs (and/or the short-term deposits of households) in the sample by a similar amount. In the LCR and in the liquidity stress test these deposits from nonbank financial institutions feature outflow rates under stress of up to 100 percent.

Banks’ deposits placed at the ECB would shrink. On the other side of the balance sheet is EA banks’ exposure to the ECB (i.e. current account holdings and deposit facility holdings) which adds up to about €2,000 billion (The banks in our sample hold about 62 percent of it). All in all, the net exposure is about €300 billion towards ECB.

The unwind the of the TLTRO would result in a reduction in CB reserves and an increase in less liquid assets, but at correspondingly higher volumes due the haircuts applied, which would largely cancel out; the major effect may be on the cost of funding, especially for weaker banks.^2/ TLTROs amount to roughly €750 billion in the EA; if the banks in the sample participated in proportion to their share of total assets in the EA banking system, their share would be €450 billion. In exchange, the CBC would increase by the encumbered assets. EA banks have posted about €1,700 billion of collateral at the EuroSystem (after haircuts). The EuroSystem operates a collateral pool, but banks are allowed to assume that least liquid assets are encumbered first. These are credit claims, unsecured banks bonds, and ABS/RMBS which in sum amount to about €1,000 billion. This is more than the outstanding TLTROs of €750 billion, such that the liquid collateral can still be reported as unencumbered by banks: government bonds (€340 billion) and covered bonds (€345 billion). The unwind of the TLTROs would lead to a substitution of CB reserves in the CBC with less liquid collateral (i.e. credit claims and ABS/RMBS). However, since the €1,000 billion posted already account for haircuts, the nominal value would be correspondingly higher and the overall impact on aggregate also would be small, except for those banks which have large dependence on EuroSystem funding, such as smaller, less complex domestic banks.

The reversal of the effect of QE on the dispersion of spreads across EA government bonds and other securities and/or a general increase of interest rates would constitute a further effect on the CBC via market liquidity risks of the securities included in it. We estimated the scenario based effects on the liquidity risk of the banks in the sample by incorporating a third approach to the CBC (see section 2.4, “market CBC”). It adds potential interest rate increases and spread dispersion across EA government bonds (i.e. takes into account flight to quality effects but not changes in liquidity risk premia) to the haircuts applied in the other approaches to the CBC. We use the estimated yield effects of the solvency stress test under the adverse scenario and translate them into valuation changes based on the data reported in the asset encumbrance and the counterbalancing concentration templates. These effects are more important for banks that will have to replace larger amounts of cheap EuroSystem funding (e.g. TLTROs) by market funding; for these banks it is important to improve capitalization and asset quality in a timely manner to avoid a spread shock.

However, without access to the Security Holdings Statistics on the bank level, the estimates are very rough and the results should be interpreted as “order of magnitude” effects. Banks’s dependence on CB funding is uneven: some smaller, less complex domestic banks received liquidity at the expense of less liquid assets, such as credit claims. Assessing detailed securities holding statistics and trading data would allow to estimate the stressed impact of discontinuation of QE and vTLTRO on market liquidity of CBC securities. Historical trading data suggest that impact may be substantial after markets enter into different liquidity regime.^3/

Careful planning is important for phase-out of exceptional measures over the medium term as this may have multiple effects on the market liquidity of securities as well as on banks with limited ability to raise market funding. TLTRO has effectively squeezed out some use of term funding markets, thus banks with high dependence on ECB funding may face problems in replacing cheap CB debt with more expensive market funding. Large scale asset purchases also affected market liquidity of securities. As monetary policy normalizes, the ECB may need to intervene in systemically important markets to maintain liquidity and price discovery.

¹/ Technically it is not an asset swap. Governments would issue bonds to repay the EuroSystem on the maturity of the bonds bought under the PSPP. Banks (and the other institutions which sold government bonds under the PSPP) would buy these newly issued government bonds. The comparative statics assume that banks would run down their excess reserves with the EuroSystem to buy these newly issued government bonds; governments would then transfer the cash to the ECB and the aggregate amount of central bank reserves would drop.²/ It is uncertain whether the banks would be able to substitute such cheap funding with market-based wholesale funding at reasonable costs.³/ See Global Financial Stability Review, October 2015, p.p. 49–82; also Han F., Leika M. An Empirical Fire-Sale Model for Stress Testing: Use of Markov Regime-Switching Models (IMF Working paper, forthcoming).

Counterbalancing Capacity and Asset Encumbrance

177. High AE ratios among some of the banks in the sample hinders their ability to further tap unsecured funding markets. In some banks the AE ratio remains high (Figure 31), reflecting dominant business model, funding constraints as well as shift of risks towards unsecured creditors. High encumbrance ratio is common to banks which issue covered bonds to finance their mortgage portfolio, for example, relatively smaller domestically-oriented banks in our sample. Banks with high level of asset encumbrance may not only face higher outflows from short-term market and deposit funding during idiosyncratic and systemic liquidity events, but also be unable to obtain additional liquidity in the markets or central banks (as central banks do typically require collateral for funding operations).⁷⁵

178. Relative asset encumbrance ratio reveals further liquidity risks. Asset encumbrance analysis is based on two ratios: absolute (encumbered assets over total assets) and relative, i.e. encumbered assets over total assets which might be encumbered (i.e. only liquid assets are counted in the denominator). The relative encumbrance ratio is much higher (average in our sample is 65 percent) and reflects available liquidity, i.e. how much assets a bank can quickly encumber in a case of liquidity stress. While average ability to encumber ratio across 29 banks is 35 percent (i.e. 100–65 percent), it does not reflect risks related to individual banks.⁷⁶

179. Contingent liquidity risks may be a source of concern in case of severe market-wide stress. Contingent liquidity risks arise due to additional margin calls because of deterioration of value of collateral (see Box 4). Based on bottom-up sensitivity analysis regularly reported by banks, €738 billion of additional collateral would need to be posted were collateral value to decline by 30 percent. The scenarios are based on gross numbers and do not reflect the value of collateral on the receiving leg (thus might overestimate total effect). In general, this represents a 23 percent increase in aggregate level of encumbrance and some of the 29 banks would face challenges in meeting all margin call requirements.

180. CBC is the first line of defense in case of liquidity shocks, and individual banks are heterogenous in terms of both, amount and quality of CBC. Some wide dispersion is observed among less complex, internationally-active and smaller, domestic oriented banks. This CBC diversity has multiple explanations, notably some relatively smaller domestically-oriented banks have higher share of long-term market based or CB based funding, thus may need lower CBC. Dispersion in CBC quality reflects asset encumbrance with the CB, i.e. banks with higher CB AE ratios tend to have higher quality CBC because the ratio measures quality of remaining unencumbered liquid assets.

Box 4.

Contingent Liquidity Risks

Contingent liquidity risks arise from financial contracts, such as repos (or securities financing transactions (SFT) in broader terms) and derivatives. Compared to behavioral liquidity flows, contingent ones are contractual, however contingent on market events, such as change in value of securities or a derivative being in or out of money. The importance of contingent liquidity risks increased due to migration of OTC SFTs and derivatives into centralized clearinghouses as well as bilateral contract margining requirements. Figure 1 below provides a high-level overview how contingent liquidity risks arise and how they impact liquidity position of a bank. In the nutshell, a market wide shock could lead to a decline in the value of collateral a bank posted on its own behalf. At the same time, a bank may also receive collateral on the opposite transaction, such as reverse repo. Collateral is typically placed and received in high liquid assets, such as cash or sovereign bonds (although there are instances when collateral is posted in less liquid assets, such as equities).

The net effect is the most important in mitigating contingent liquidity risks, i.e. if a bank has balanced own trading position and has no large unidirectional bets in the market, it would post and receive collateral, thus liquidity needs would be netted, especially if contracts are cleared via CCPs (see Figure 2).

If a bank needs to post more collateral than it receives, its AE ratio goes up, which dries out CBC and may lead to further idiosyncratic shocks, such as inability to rollover unsecured wholesale funding.

181. Banks with low RWA density and high AE may be subject to heightened liquidity risks. Banks which have high share of low risk weighted assets, such as mortgages, and issued asset backed securities or covered bonds, may be subject to heightened liquidity risks if losses in mortgage segment would lead to a sharp drop in CAR at the time of significant refinancing needs, for example, when banks have to redeem own debt securities. Banks with high LCR ratios tend to have higher AE ratios, which shows relative amount of liquidity transformation and support obtained from CB.

C. Cash Flow Liquidity Stress Tests

Introduction and Objectives

182. The cash flow based liquidity stress test (CFLST) analyses the liquidity risk exposure and risk bearing capacity of the sample of 29 banks in the EA. CFLSTs incorporate a set of embedded scenarios that allowed the FSAP team to estimate the order of magnitude of potential liquidity needs of individual banks and the banking system (comprising the sample of 29 banks) under a baseline and multiple stress scenarios. CFLST reveal levels of liquidity risk tolerance, i.e., under which circumstances banks would need additional liquidity support because of the mismatch of cash flows and absence of available CBC for which level of stress banks hold an adequate CBC. In addition, the CFLST contribute to the assessment of common liquidity risk exposures of banks in the banking system, such as reliance on unsecured short- term funding, also holdings of similar less liquid assets in the CBC. The CFSTs does not consider potential redistribution of liquidity within the banking system, i.e. migration of deposits from banks which experience capital shortfall to the banks which have strong capital buffers. This redistribution of flows was observed during the GFC.

Method and Data

183. Cash-flow based liquidity stress tests transform reported cash-flow data into stressed cash-flows and security flow data based on a matrix of scenario dependent stress factors.⁷⁷ They focus on two key indicators, namely, liquidity risk exposure and liquidity risk bearing capacity of banks. The first indicator is defined as the difference between cash-inflows and cash-outflows in each time bucket (the net-funding gap, NFG) and the sum of these differences across buckets (i.e. the cumulated net-funding gap, CNFG). The second indicator is the CBC, defined as the sum of cash inflows banks can generate under stress at reasonable prices in the respective bucket after taking into account securities flows. The CCBC is the sum of the counterbalancing capacities across time buckets and the current one. The analysis builds on data collected within the Short-Term Exercise (STE): especially, but not limited to, the Additional Maturity Mismatch Template (Additional Maturity Ladder, C66.00) from 4Q 2014 to 3Q 2017.

184. SSM data quality assurance for the STE is deemed broadly adequate. SSM data quality assurance proceeds in three steps: (i) first the statistics departments of the NCAs run plausibility tests and apply a set of validation rules, (iii) then the SSM statistics department runs validation rules, (iii) then the JSTs use the data in their regular off-site as well as on-site analysis and include any reporting errors and inconsistencies they encounter in their finding and request banks to improve liquidity risk reporting. Nonetheless, the IMF conducted additional data quality checks, comparing the cash flow data with balance sheet data as well as data from additional liquidity monitoring metrics templates.

185. Contractual liquidity risk exposure of banks in the sample is high. Contractual outflows within the first 4 weeks amounts to about 31 percent of total assets (weighted average; excluding open maturity and overnight retail deposits (23 percent of TA) and open maturity and overnight corporate deposits (7.5 percent of TA)) (see Figure 32). The contractual inflows amount to about 21 percent of TA (excluding inflows from central bank deposits (5.9 percent of TA) due to reporting conventions). Thus, the cumulated net funding gap over the first 4 weeks reaches about 10 percent of total asset or €1,700 billion. The main drivers of the net outflows are:

(i) outflows from deposits of financial institutions (-4 percent of TA, net of inflows from financial institutions);
(ii) other deposit outflows (-2.6 percent of TA, net of other deposit inflows);
(iii) repos collateralized with 0percent risk-weight bonds (-1.4 percent of TA, net of similar reverse repos); and
(iv) The composition of repo collateral (ex 0 percent risk-weight bonds) is similar to the composition of collateral for reverse repos within the first four weeks. Gross encumbrance due to repos across all asset categories accounts for about 14 percent of TA; gross reverse repos account for about 10 percent of TA. The composition of repo collateral (ex 0 percent risk-weight bonds) is similar to the composition of collateral for reverse repos within the first four weeks.

186. Banks in the sample are also susceptible to contingent liquidity risks stemming from changes in net value of repos and derivatives positions. Due to the lack of data, the adverse scenario based impact on the value of collateral posted for derivative transactions could not be estimated, although the value of additional collateral posted/received due to repo and reverse repo transactions (up to three months) was considered in the cash flow liquidity stress tests. The value of collateral was reduced by the haircut obtained from market risk scenario.

Counterbalancing Capacity (CBC)

187. The CBC in the first month fully covers the system-wide cash flow gap. The total sum of assets in the counterbalancing capacity amounts to 20 percent of TA. It is higher than the net funding gap in the unstressed reported data (excl. retail and corporate deposits), which amounts to about 10 percent of TA over the first four weeks. While banks as a group have enough CBC to cover the gap, its distribution is uneven and some banks face shortfalls under stress.

188. Cash and deposits at the CB dominate the composition of the CBC (Table 6). Central bank deposits and 0 percent risk-weight securities account for about two third of the CBC. Credit claims and other non-HQLA are the third largest position with roughly 25 percent of the CBC.

189. Most of the securities included in CBC have low to very low credit risk. Analysis of unencumbered assets by credit quality steps (CQS) reveals that most of the unencumbered assets (i.e. excluding cash, CB reserves, credit claims and other non-marketable securities) fall into the lowest credit risk category (CQS1) thus can be quickly converted into cash and/or used for collateral management purposes. In terms of U.S. dollar liquidity, banks that are active in FX funding market accumulated large amounts of U.S. sovereign debt securities and U.S. government-sponsored enterprise issued papers, which also fall under CQS 1 category.

Table 6.

Composition of the CBC, 2017 Q3

(Weighted average across all banks, in percent of TA)

Table 6.

Composition of the CBC, 2017 Q3

(Weighted average across all banks, in percent of TA)

Compared to HQLA	Compared to ST assumptions	Asset category in the CBC	Value in percent of TA
		Cash	0.3
HQLA I		Central bank deposits	7.3
		0 percent risk-weight securities	5.8
	Marketable liquid assets	20 percent risk-weight securities	0.2
		Covered bonds	0.8
HQLA II		Corporate bonds	0.2
		RMBS	0.5
		Other CB eligible assets (i.e. Credit Claims)	3.4
Non-HQLA	Non-Marketable liquid	Non-CB eligible equity	0.7
assets	assets		0.8
		Other non-CB eligible assets

Table 6.

Composition of the CBC, 2017 Q3

(Weighted average across all banks, in percent of TA)

Compared to HQLA	Compared to ST assumptions	Asset category in the CBC	Value in percent of TA
		Cash	0.3
HQLA I		Central bank deposits	7.3
		0 percent risk-weight securities	5.8
	Marketable liquid assets	20 percent risk-weight securities	0.2
		Covered bonds	0.8
HQLA II		Corporate bonds	0.2
		RMBS	0.5
		Other CB eligible assets (i.e. Credit Claims)	3.4
Non-HQLA	Non-Marketable liquid	Non-CB eligible equity	0.7
assets	assets		0.8
		Other non-CB eligible assets

190. Euro area and U.S. sovereign securities are the most prevalent debt assets included in the CBC. Five countries (Italy, Spain, Germany, United States, and France) cover roughly 65 percent of sovereign exposures in the sample of banks. As most of the banks are exposed to these securities, in the event of a large market shock (such as sovereign crisis in 2012), market liquidity of some of these assets may be severely affected by the typical flight to quality effect with investors selling bonds with lower credit rating but buying higher credit rating securities. At the same time, risk mitigating factor is that average remaining maturity of these sovereign securities is close to two years, hence the impact of changes in interest rates and sovereign risk premiums is relatively lower compared to impact on holdings of long-term debt securities.

Scenarios

191. Parameter uncertainty is an integral component of any liquidity stress test.

Regardless of the sophistication of parametrization, liquidity stress situations can be unexpected and last for various time periods (from a few days to the slow drain of liquidity over many months). The best way to address the inherent parameter uncertainty is to run a large set of embedded scenarios of increasing severity.

192. The analysis employs a set of 20 embedded scenarios for a 4-week, 3 months as well as five scenarios for a 5-day time horizon, respectively: a baseline scenario and 4 stress scenarios with increasing severity (mild market stress, medium market stress, severe market stress and a combined severe market-wide and idiosyncratic stress. Each of the stress scenarios is combined with four different approaches to the counterbalancing capacity:

(i) full CBC without haircuts (HCs): fully endogenous liquidity supply by the central bank as long as banks have unencumbered eligible collateral;
(ii) full CBC with bank specific HCs: fully endogenous liquidity supply by the central bank as long as banks have unencumbered eligible collateral;
(iii) marketable CBC: non-marketable components of the counterbalancing capacity (i.e. credit claims and committed lines provided to the banks) are disregarded;
(iv) liquid CBC: bank specific haircuts and bank specific market price effects derived from the solvency stress test for assets that are liquid in private markets.

193. Embedded scenarios allow to simulate different degree of banks’ dependence on central bank’s support under liquidity stress (Figure 33). To ensure consistency with solvency stress tests, the calibration of haircuts under the liquid CBC approach draws on the asset prices (sovereign and corporate bonds) under the adverse scenario of the solvency stress tests.

194. This approach to parameter uncertainty shifts discussions from parameter calibration to liquidity risk tolerance of banks. The ultimate goal of the liquidity tests is to determine a banks’ risk tolerance for liquidity risk, that is, determine the maximum degree of risk that the bank is willing to accept under stress conditions. The choice of “threshold” itself is based on multiple parameters, for example, percentage of funding outflow, haircuts on CBC assets. If many banks fail under a very mild outflows or haircuts scenario, this reveals a high liquidity risk tolerance and vice versa. In addition, the broad set of scenarios allows for effective comparisons of liquidity risk exposure and liquidity risk tolerance across banks beyond the simple pass/fail dichotomy.

195. Five days cyber-risk related scenario assumed unavailability of collateral which is used for rehypothecation. Rehypothecation is a common liquidity management approach, when a financial institution (lender) uses collateral received from another financial institution (a borrower) to cover its own liquidity needs. Unencumbered collateral available for rehypothecation is included into CBC for the life of the loan contract. The scenario assumed severe liquidity event lasting for five working days; with daily simulation of contractual as well behavioral cash flows according to the scenarios. The choice of five days was based on assumption that access to the collateral would be restored at the end of this time horizon. Due to data constraints we were not able to separate bilateral and tri-party repos (the former less susceptible to a cyber risk event as such contracts may be managed bilaterally and not via central clearing counterparty or central depository institution).

196. Scenario calibration builds on event studies of system-wide and idiosyncratic liquidity stress events and is broadly consistent with the literature. For example:

(i) retail deposit outflows within one week reached 11 percent in the banking system of for Banesto (ES, 1994) 8 percent, and for IndyMac (USA, June 2008) 7.5 percent. For Washington Mutual (USA, September 2008) retail deposit outflows amounted to 8.5 percent in 10 days and for DSB Bank (NL, 2009) they reached 30 percent in 12 days (Schmieder et al. 2012, Table 3). In their severe scenario (comparable to the Lehman crisis), the outflow rates for retail term deposits amount to 10 percent over the 30-day horizon and for 20 percent for demand deposits;
(ii) unsecured short-term wholesale funding run-off rates amounted to 100 percent;
(iii) For secured wholesale funding the outflow rate is 20 percent;
(iv) The EBA exercise (Severe Market Scenario) applied run-off rates of 5 percent (retail deposits), 10 percent (NFC deposits), 20 percent (nonbank financial institutions), 100 percent (financial institutions), and 0 percent (government/public entities);
(v) In Halal, Laliotis (2017, Severely adverse scenario) the run-off rates amount to 10 percent for stable deposits, 20 percent non-stable deposits, 100 percent (net unsecured interbank funding), 50 percent (net secured interbank funding), and 100 percent (other wholesale funding, except ABS 50 percent).

197. Run-off rates are higher for unsecured than for secured wholesale funding, as well as for non-insured deposits than for insured ones. Table 7 summarizes the calibration of the inflow and outflow parameters.⁷⁸ The inflows parameters are in principle 100 percent of the contractual inflows, except for inflows from loans to retail and for corporate customers (inflows 0 percent). This is in line with the objective of the CFLST to assume that banks will continue business as normal, i.e. analyze the ability of banks to cope with liquidity stress while maintaining their ability to lend to the real economy. In fact, when bank cuts credit lines and/or stops granting loans, it may send the negative signal to the markets about its financial situation, which leads to further outflows from that bank.

198. The haircuts applied for the various components of the counterbalancing capacity capture two different effects: (i) the haircuts applied by the repo counterparty and (ii) the value upon which the haircut is applied (or market price effect). In the “full CBC without haircuts”, no haircuts are applied at all. In the “full CBC approach and the “marketable CBC” approach, scenario independent, bank specific ECB haircuts as of December 31, 2017 are applied.⁷⁹ Table 8 displays the averages across banks as well as the standard deviations and the range. The first three approaches to the CBC do not consider market price changes of the underlying assets in the CBC. These are added in the fourth approach to the CBC (“liquid CBC”). Moreover, the first three approaches to the CBC assume that banks have unlimited access to central bank funding, as long as they can provide eligible collateral.

Table 7.

Scenario Parameters: Run-off Rates for the Major Components of In- and Outflows

Note: * Stressed outflow rates for repos are captured by increasing haircuts for the underlying collateral class. Calibration based on: BCBS (2014), ..Liquidity stress-testing: a survey of theory, empirics and current industry and supervisory practices”, BCBS Working Paper 24. BCBS (2014), .Literature review of factors relating to liquidity stress – extended version” and the literature cited therein. Schmieder et al. (2012). Hataj, G., D. Laliotis (2017), “A top-down liquidity stress test framework”, In: Dees, S., J. Henry, R. Martin (eds.), “Stress-Test Analytics for Macroprudential Purposes in the euro area”, ECB, 168–191. LCR Delegated Act - Commission Implementing Regulation (EU) 2016/322 of 10 February 2016 amending Implementing Regulation (EU) No 680/2014.

Table 7.

Scenario Parameters: Run-off Rates for the Major Components of In- and Outflows

Outflows/Inflow	Range of run-off factors (in percent) across mild, medium, severe and severe/idiosyncratic scenarios
Unsecured LT/ST Issuances and financial deposits	30–70
Secured issuances	10–30
Stable retail deposits	2–7
Unstable retail deposits	3–12
Operational corporate deposits (NFCs)	10–25
Non-operational corporate deposits (NFCs) & other deposit outflows	20–50
Repo across all collateral classes*	0
Deposits FI	30–100
FX-Swaps in-/outflows	15–100
Derivative in-/outflows	25–100
Retail / corporate inflows	0
Central bank inflows	100
Other entities inflows	0–30
FI inflows	30–100
Other inflows	100
Committed lines provides by the bank (FI)	25–75
Committed lines provides by the bank (non-FI)	3–10
Margin calls (derivatives)	2–10
Adverse market outflows (derivatives, HLBA)	0–100
Outflows due to rating downgrades	0–100

Table 7.

Scenario Parameters: Run-off Rates for the Major Components of In- and Outflows

Outflows/Inflow	Range of run-off factors (in percent) across mild, medium, severe and severe/idiosyncratic scenarios
Unsecured LT/ST Issuances and financial deposits	30–70
Secured issuances	10–30
Stable retail deposits	2–7
Unstable retail deposits	3–12
Operational corporate deposits (NFCs)	10–25
Non-operational corporate deposits (NFCs) & other deposit outflows	20–50
Repo across all collateral classes*	0
Deposits FI	30–100
FX-Swaps in-/outflows	15–100
Derivative in-/outflows	25–100
Retail / corporate inflows	0
Central bank inflows	100
Other entities inflows	0–30
FI inflows	30–100
Other inflows	100
Committed lines provides by the bank (FI)	25–75
Committed lines provides by the bank (non-FI)	3–10

The sample of banks covered in the stress test are grouped into three broad categories for peer benchmarking.
G-SIB (7)			Large internationally active (10)			Smaller, more domestic-focused (12)
Country	Bank	Assets (billions of euros)	Country	Bank	Assets (billions of euros)	Country	Bank	Assets (billions of euros)
FR	BNP Paribas	1,750	FR	Groupe BPCE	1,161	DE	DZ Bank	417
DE	Deutsche Bank	1,480	ES	BBVA	673	ES	CaixaBank	336
FR	Credit Agricole	1,402	FR	Credit Mutuel	664	DE	Baden-Württemberg	235
ES	Banco Santander	1,431	IT	Intesa Sanpaolo	648	ES	Banco de Sabadell	221
FR	Société Générale	1,138	NL	Rabobank	603	ES	BFA Tenedora de Acciones	218
IT	UniCredit	862	DE	Commerzbank	452	DE	Bayerische Landesbank	215
NL	ING	846	NL	ABN AMRO	390	DE	Norddeutsche Landesbank	168
			BE	KBC Group	260	IT	Banco BPM	161
			AT	Erste Group	220	BE	Belfius Banque	149
			AT	Raiffeisen Bank Int	135	FI	OP Financial Group	118
						IE	Bank of Ireland Group	106
						IE	Allied Irish Bank	90
	Sub-total	8,910		Sub-total	5,207		Sub-total	2,433
Note: Actual data end-2017. There is one G-SIB in FSB designated bucket 3 (2.0 percent capital surcharge), one G-SIB in (1.5 percent capital surcharge), and 4 G-SIB in bucket 1 (1.0 percent capital surcharge).
Material geographies for the sample of banks include nine home euro area countries, eleven other EU countries, and seventeen countries outside the EU.

The sample of banks covered in the stress test are grouped into three broad categories for peer benchmarking.
G-SIB (7)			Large internationally active (10)			Smaller, more domestic-focused (12)
Country	Bank	Assets (billions of euros)	Country	Bank	Assets (billions of euros)	Country	Bank	Assets (billions of euros)
FR	BNP Paribas	1,750	FR	Groupe BPCE	1,161	DE	DZ Bank	417
DE	Deutsche Bank	1,480	ES	BBVA	673	ES	CaixaBank	336
FR	Credit Agricole	1,402	FR	Credit Mutuel	664	DE	Baden-Württemberg	235
ES	Banco Santander	1,431	IT	Intesa Sanpaolo	648	ES	Banco de Sabadell	221
FR	Société Générale	1,138	NL	Rabobank	603	ES	BFA Tenedora de Acciones	218
IT	UniCredit	862	DE	Commerzbank	452	DE	Bayerische Landesbank	215
NL	ING	846	NL	ABN AMRO	390	DE	Norddeutsche Landesbank	168
			BE	KBC Group	260	IT	Banco BPM	161
			AT	Erste Group	220	BE	Belfius Banque	149
			AT	Raiffeisen Bank Int	135	FI	OP Financial Group	118
						IE	Bank of Ireland Group	106
						IE	Allied Irish Bank	90
	Sub-total	8,910		Sub-total	5,207		Sub-total	2,433
Note: Actual data end-2017. There is one G-SIB in FSB designated bucket 3 (2.0 percent capital surcharge), one G-SIB in (1.5 percent capital surcharge), and 4 G-SIB in bucket 1 (1.0 percent capital surcharge).
Material geographies for the sample of banks include nine home euro area countries, eleven other EU countries, and seventeen countries outside the EU.

37 Material geographies
Core EA (9)		Other EU (11)		Outside	EU (17)
Austria	AT	Bulgaria	BG	Australia	AU
Belgium	BE	Croatia	HR	Brazil	BR
Finland	FI	CzechRepublic	CZ	Canada	CA
France	FR	Hungary	HU	Chile	CH
Germany	DE	Luxembourg	LU	China	CN
Ireland	IE	Poland	PL	Colombia	CO
Italy	IT	Portugal	PT	Egypt	EG
Netherlands	NL	Romania	RO	India	IN
Spain	ES	Slovakia	SK	Japan	JP
		Sweden	SE	Mexico	MX
		United Kingdom	GB	New Zealand	NZ
				Peru	PE
				Russia	RU
				Singapore	SG
				Switzerland	CH
				Turkey	TR
				United States	US

37 Material geographies
Core EA (9)		Other EU (11)		Outside	EU (17)
Austria	AT	Bulgaria	BG	Australia	AU
Belgium	BE	Croatia	HR	Brazil	BR
Finland	FI	CzechRepublic	CZ	Canada	CA
France	FR	Hungary	HU	Chile	CH
Germany	DE	Luxembourg	LU	China	CN
Ireland	IE	Poland	PL	Colombia	CO
Italy	IT	Portugal	PT	Egypt	EG
Netherlands	NL	Romania	RO	India	IN
Spain	ES	Slovakia	SK	Japan	JP
		Sweden	SE	Mexico	MX
		United Kingdom	GB	New Zealand	NZ
				Peru	PE
				Russia	RU
				Singapore	SG
				Switzerland	CH
				Turkey	TR
				United States	US

Topics

Countries

Series

About

Resources

Abstract

Executive Summary

Introduction1

A. A Complex Financial System

B. Banking Sector Trends

C. Supervisory Stress Tests in the Euro Area

D. Stress Testing in the FSAP

E. Sample of Banks

F. FSAP Stress Testing Strategy

G. Banking Sector Risks

Solvency Stress Test

A. Key Elements of the Stress Test

B. Scenarios

C. Credit Risk Modelling Approach

D. Market Risk Approach

E. Interest Rate Risk in the Banking Book Setting the Stage

F. Other Profit and Loss Items

G. Solvency Results

H. Sensitivity Tests

Liquidity Risk Analysis and Stress Tests

A. Overview

B. Structural Liquidity Risks

C. Cash Flow Liquidity Stress Tests

Introduction¹