A discussion paper, Bank of England, October 2013
- The Bank expects to use a suite of models to estimate the impact of stress scenarios.
- to guard against the risk that the framework becomes excessively exposed to the unavoidable weaknesses of any single model.
- It also leaves a greater role for judgement in combining model outputs to reach an overall view on capital adequacy the exercise will focus primarily on risks to bank solvency. At least initially, funding and liquidity vulnerabilities are likely to be incorporated only as amplification mechanisms.
- All PRA-regulated banks are already required to carry out stress tests and scenario analysis as part of their Internal Capital Adequacy Assessment Process (ICAAP).
- this framework will not replace the requirement that banks undertake stress tests for their own risk management purposes
- In most countries, the results of stress-testing exercises are used as one of a set of inputs into policy decisions around bank capital.
- Many different types of risk, including market, credit, funding and liquidity risks.
- Moreover, reliance on a single quantitative tool heavily exposes the stress-testing exercise to ‘model risk’.
- Finally, from a system-wide perspective, relying on a single model could have adverse consequences for diversity in risk measurement and risk management practices across the system. Broadly speaking, the Bank expects to employ four main types of models and associated analysis the Bank’s Risk Assessment
- Model of Systemic Institutions (RAMSI), are better able to take a system-wide view of risk. 1) Burrows, Learmonth and McKeown (2012).
- For example, when considering a bank’s mortgage book, relevant risk factors might include the loan to value (LTV) ratio, product type (for example, interest-only or repayment), type of borrower (prime, sub-prime or buy-to-let) and whether or not repayments are fully up-to-date or in arrears
- For example, ignoring the distribution of LTV ratios across borrowers could result in an over or underestimation of risk in different banks’ mortgage books when assessing the possible impact of a shock originating in the housing market.
- It may be that the correlations between risk factors or other interactions are ultimately more important indicators of systemic risks than the individual risk factors themselves
- There will remain a number of other complementary sources of information, including asset quality reviews (discussed in Section 6). In fact, asset quality reviews may inform the appropriate level of granularity of stress-testing models.
- Comparison of banks’ projections with those from regulatory models
- Peer comparison
- It is reasonable to incorporate at least some of banks’ proposed responses to macroeconomic and market shocks when evaluating the impact of scenarios on bank capital adequacy.
- Amplification mechanisms. Such effects are typically not modelled in bank-specific stress tests, which tend to take a partial-equilibrium approach to risk measurement.(1)
- Amplification mechanisms can occur between banks, between banks and the real economy, and between banks and non-bank financial companies.
- Interbank exposures, Fire sales, Liquidity hoarding, Confidence channels
- A change in the supply of credit is a key channel through which the real economy may be affected.
- Spillover and feedback analysis is an important element of stress tests in increasingly interconnected financial systems. But the development of stress-testing models in this area remains at a relatively early stage.
- The Bank of Canada, for example, incorporates a network of interbank linkages as well as asset fire sale effects in its stress-testing approach (8) Kapadia et al (2012). (9) Gauthier and Souissi (2012).
What to avoid
- insufficient engagement by banks’ Boards and senior management with the stress-testing process;
- insufficient integration of stress testing with banks’ annual business planning process, including the use of stress tests as a challenge to business plans;
- inadequacies in scenario design, including the failure to identify key vulnerabilities, overly optimistic baseline assumptions and insufficiently stressful adverse scenarios;
- difficulties in reconciling risk data with reported balance sheets and risk-weighted assets;
- stress-testing infrastructures that have not been suitable for bank-wide stress testing;
- insufficiently justified or internally challenged assumptions and judgements around the translation of macroeconomic shocks into projected losses, including overestimation of banks’ ability to control margins and generate profits in stress scenarios; and
- inadequate determination and quantification of relevant management actions under different stress scenarios.
What to expect
- Banks will be expected to supply the necessary data to allow Bank staff to run the common scenarios independently using their own models.
- The 2014 stress test will incorporate three scenarios: a common baseline; a common stress scenario; and an institution-specific stress scenario for each bank.
source: Discussion paper