Stress Testing at the Deutsche Bundesbank Dr. Philipp Koziol* Deutsche Bundesbank Edinburgh, 25th April 2014 * Disclaimer: The presentation represents the author s personal opinion and do not necessarily reflect the views of the Deutsche Bundesbank or its staff.
Agenda 0. Introduction 1. Stress test approach for large IRB banks 2. Stress test framework for small and medium-sized banks 3. Impact of copulas 4. Link to EBA/SSM stress test 2
0. Introduction 3
1. Stress test approach for large IRB banks 4
1. Stress test approach for large IRB banks 1.1 Framework Analysis of how a strongly recession in Germany impacts the corporate credit portfolios of German banks which use the Internal Rating Based (IRB) approach Impact of the stress scenario on the Tier 1 capital ratio of the IRB banks Portfolio data provided from the German central credit register at the reporting date of June 30, 2013 5
1. Stress test approach for large IRB banks 1.3 Macroeconomic stress scenario GDP Sector GDP Index 1 Multifactor Credit Risk Model Systematic Factor 1 Systematic Factor 2 Financial Crisis Scenario GDP Index 2 Correlations Stressed Expected Losses (Impairments) Total Capital Ratio GDP Index 18 Systematic Factor 17 Systematic Factor 18 6
1. Stress test approach for large IRB banks 1.2 Macroeconomic stress scenario Macroeconomic shock: modeled by the historical data of the financial crisis 2008/2009 (3,8% decrease of the GDP on an annual basis) Historical evolution of the sub-sectors of the German GDP provides the initial allocation Stress can be transmitted more precisely considering sub-sectors Consideration of the sub-sectors allows a more precise stress transmission, especially regarding the different impact of the financial crisis on the subsectors Stress scenario is based on the realized sub-sector GDP change rates (annualized values of 2008Q3 2009Q4) and applied on the initial allocation of the subsector 7
1. Stress test approach for large IRB banks 1.4 Macroeconomic stress scenario GDP-growth rates on sectoral level (a) Entire period (b) Crisis period 8
1. Stress test approach for large IRB banks 1.5 Stress test approach for the Tier 1 capital ratio Impact of a baseline and stress scenario on the solvency of the respective institutes: Tier1 capital ratio Baseline = Tier 1 capital 12.5(K Credit + K Market + K OpRisk ) Tier1 capital ratio Stress = Tier 1 capital ΔELStress 12.5(K Stress Credit + K Market + K OpRisk ) ΔEL Stress K Stress Credit K Credit Change of the expected losses related to the stress scenario Change of the minimum regulatory equity requirements ΔEL Stress is calculated using simulations on the basis of multi-factor portfolio model ΔEL Stress = λ E L Stress E L Baseline with λ = Total exposure of coporate credits subject to the SolvV Total exposure of large coporate credits 9
1. Stress test approach for large IRB banks 1.6 Results Influence of the stress scenario on the IRB bank s Tier 1 capital ratios in 2013 and the comparison of the stress impact to 2012 10
2. Stress test framework for small and medium-sized banks 11
2. Stress test framework for small and medium-sized banks 2.1 Goals Analysis of the resilience of smaller and medium-sized banks is important since they play a significant role in providing a functioning credit flow to the real economy, mainly for small and medium-sized enterprises (SMEs) Macroeconomic Portfolio Stress Test Adaptation of macroeconomic stress scenario of large IRB banks Multifactor Credit Risk Model: stress test of the banks' credit portfolios Income stress test model: stress test of the banks main income components Total stress effect on each bank: impact on the total capital ratios over a one-year horizon Basis stress scenario: macroeconomic stress scenario reflecting the experience of the recent financial crisis Basis stress scenario is transferred into both the multifactor credit risk model as well as the income stress test model 12
2. Stress test framework for small and medium-sized banks 2.2 Macroeconomic Portfolio Stress Test Design GDP Sector GDP Index 1 Multifactor Credit Risk Model Systematic Factor 1 Systematic Factor 2 Financial Crisis Scenario GDP Index 2 GDP Index 18 GDP Growth Short-term Interest Rate Systematic Factor 17 Correlations Income Stress Test Model Net Interest Income Net Commission & Fee Income Systematic Factor 18 Stressed Expected Losses (Impairments) Stressed Net Income (excl. Impairments) Total Capital Ratio Operating Expenses Long-term Interest Rate 13
2. Stress test framework for small and medium-sized banks 2.3 Macroeconomic Portfolio Stress Test Design Total capital ratio under baseline and stress conditions for t + 1: j TotCR t+1 Satellite Model Stress effect = T1C j j t + T2C t + T3C t + NIeI t+1 I t+1 12.5 K CR,t + K MkR,t + K OpR,t Multi Factor Portfolio Model j stress, baseline j TotCR t+1 Total capital ratio in t+1 under stress or baseline conditions T1C t, T2C t, T3C t Values for Tier 1, Tier 2 and Tier 3 Capital at t K CR,t, K MkR,t, K OpR,t Current reported regulatory capital charges for credit, market and operational risk at t j I t+1 Forecasted impairment charge for t + 1 in the stress or baseline case j NIeI t+1 Forecasted net income (excluding impairments) for t + 1 in the stress or baseline case 14
3. Impact of copulas 15
3. Impact of copulas 3.1 Copulas A Copula C(.) is a multivariate distribution function on the n-dimensional unit cube with uniformly distributed marginals on [0,1] and represents the multivariate dependence structure Multivariate Distribution: F(x 1,, x n ) Dependence structure of the random variables: C(u 1,, u n ) Marginal distributions of each random variables: F 1 x 1,, F n (x n ) F x 1,, x n = C F 1 x 1,, F n (x n ) x R n 16
3. Impact of copulas 3.2 Distribution of Tier1 capital ratios Distribution of Tier1 capital ratios under normal and stressed conditions using different copulas Unanticipated result: The Gaussian copula gives the harshest forecast 17
3. Impact of copulas 3.3.1 Conditional expectation of the risk factors - Impact of the correlation impact of correlation on E copula X 2 X 1, X 2 c for fix c = 2 Clayton copula most severe for moderate correlation, Gaussian copula most severe for high correlation, t copula relatively robust to correlation of the risk factors 18
3. Impact of copulas 3.3 Conditional expectation of risk factors Compute precise results for E copula X 2 X 1, X 2 c c, ρ Graph shows level curves: If level curve for one copula lies above the others, this copula will give the more severe forecast For extreme stress and high correlation of the risk factors, the Gaussian copula is the most severe one! 19
3. Impact of copulas 3.3.2 Conditional expectation of the risk factors - Impact of number of risk factors E copula X i X 1,, X n c for fix c = 2 and n 1,, 10 As the number of truncated risk factors/sectors increases, the relative severity of the Gaussian copula increases 20
4. Link to EBA/SSM stress test 21
4. Link to EBA/SSM stress test EBA conducts an EU wide stress test on a sample of banks covering systemic banks (at least 50% of the national banking sector in each EU Member State) Bottom-up fashion, using consistent methodologies, scenarios and key assumptions developed in cooperation with the ESRB, the European Central Bank (ECB) and the European Commission (EC). ECB in preparation of the Single Supervisory Mechanism (SSM) is conducting a comprehensive assessment comprising of a supervisory risk assessment, asset quality review and a stress test Comprehensive Assessment: Risk Assessment, Asset Quality Review, Stress Test Exercise Single Supervisory Mechanism (SSM) appoints how the collaborative banking supervision by the ECB and the national supervisory bodies is organized ECB and EBA conduct a stress test for significant banks in the Eurozone, which are defined by: Value of assets exceeds 30 billion Value of assets exceeds both 5 billion and 20% of the GDP of the member state in which it is located One of the three most significant banks in the respective country Assistance from Eurozone bailout fund Large cross-border activities 22
4. Link to EBA/SSM stress test 4.1 Treatment of different types of risk Credit risk in the banking book: Stress effect on capital through: Rise in Expected Loss (EL) New defaults and resulting impairments Higher impairments on loans that were already in default Stress effect on risk weighted assets (RWA) through: Higher equity requirements, e.g. through an increase in PD and LGD RWA level of 2013 as floor Data provided by the national central banks contributes to benchmark parameter estimation of PDs and LGDs (problem in EBA ST 2011) 23
4. Link to EBA/SSM stress test 4.1 Treatment of different types of risk Sovereign risk: Stress effect depends on accounting category: HtM and LaR: analogous to credit risk methodology Market risk: HfT, FVO and AfS: valuation haircuts on government bonds For AfS positions, prudential filters as in CRR/CRD IV are applied 60 % phase-out in 2016 Banks with significant trading activities or model banks : Capital: scenario dependent losses at fair value in categories HfT, AfS and FVO RWA: Increase in VaR, SVaR, IRC, CRM and CVA 24
4. Link to EBA/SSM stress test 4.1 Treatment of different types of risk Securitization risk: Positions at amortized cost: Capital: impairments RWA: deterioration of credit quality through rating migration Positions at fair value: Mark-to-market valuation analogous to market risk approach Funding risk: increase in RWA through rating migration Banks determine increase in funding cost through higher contango money for: Wholesale funding (adverse macroeconomic development and increase in risk aversion) Retail funding (enhanced competition) Expired LTROs can be substituted by cheap refinancing operations with the ECB Part of the rise in funding cost can be mitigated by a pass-through for expiring loans 25