Stress Testing

The introduction of stress testing

A major trigger behind the new focus for stress testing in the risk management of banks was the problem that the established risk models and VaR approaches had previously only adequately modelled the usual market fluctuations. Extreme scenarios were - and still are - insufficiently taken into account. These days, the primary goal of stress testing has to be to examine the effect(s) on the value of a single portfolio, or a bank’s overall portfolio, of crisis events, such as a stock market crash, interest-rate and exchange-rate shocks, or an economic recession. It should be possible to reliably determine potential losses.

In this regard, besides the uniform baseline scenarios specified by the regulator (e.g. "abstractly defined severe economic downturn", or operationalising this as part of the "ECB stress tests" with support provided by SKS Advisory), it is also necessary to identify where the major risks, which are specific to an institution, are to be found in the business model.

Stress tests - also for analysing and monitoring specific business risks

Stress tests are used for analysing and monitoring specific business risks. They have already become an established part of operational practice for the banks with respect to market and liquidity risks. In the course of implementing regulatory requirements, such as, e.g. the "SolvV", or the new German regulation governing large exposures and loans (Großkredit- und Millionenkreditverordnung, "GroMiKV"), if an approach based on internal ratings (IRB) is used, stress testing will become mandatory for credit risk and operational risk in banks, too. To this end, when designing stress tests for credit risks, particular crisis events will be converted into risk factors and subsequently implemented in the stress test scenario within the scope of the business risk that needs to be considered.

Risk factors can be either market parameters or model parameters  

  • Market parameters are variables, such as, e.g. yield curves, or inflation rates, which can be directly observed in the market and can be measured.
  • Model parameters are variables, such as, e.g. correlations, which are determined from market parameters, or other model parameters within a stress test model.

An overview of the relevant business risks in the banking sector and the respective risk factors is presented below.



Methods of stress testing

A distinction is made between two types of stress tests

  • Sensitivity analyses examine the isolated effect of an extreme change (on the risk factor side) on the dependent variable. This method is also called the "univariate" stress testing procedure.
  • Scenario analyses take as the object of their investigation the effect of extreme changes in several risk factors, occurring simultaneously, and the correlation between them and the dependent variable. This method is also referred to as the "multivariate" stress testing procedure.

The univariate stress test

  • As regards credit risks, the increase in the probability of default for a particular portfolio can be an example for this easy-to-implement procedure. A major drawback of this method is, however, that one single risk factor is viewed separately. Thus, other risk factors and their correlations are not taken into account.

The (multivariate) stress test

  • With the multivariate method several risk factors are used to map an extreme scenario and, moreover, correlations are included in the calculations, therefore, this method generates results that are much more realistic.

Portfolio-specific vs. non-portfolio-specific scenarios

Non-portfolio-specific stress tests are especially designed not to be used for a definite portfolio. They can be constructed as macro stress tests (e.g. deep recession, oil price shock) as well as historical tests (e.g. stock market crash 10/87) or as a standardised scenario. Their advantage consists in the good comparability of the results over a time line. However, the drawbacks include not taking into account portfolio-specific features (such as, e.g. of special financing) and being restricted to historical events.

In the implementation of portfolio-specific scenarios, as a rule, recourse is made to qualitative or quantitative scenario construction methods.

  • The qualitative scenario construction method employs a panel of experts to define stress scenarios. The advantage consists precisely in the involvement of this panel of experts. The disadvantage is the arbitrariness of the scenarios, as this does not permit a comparison between the different financial institutions over a time line.
  • The quantitative scenario construction method uses the empirical distribution of the portfolio, which involves ascertaining for each specific risk factor whether an increase or a decrease of k standard deviations would trigger a loss. Subsequently, a scenario is constructed out of the adverse changes. One advantage of this procedure, which is also referred to as the "factor push" method, is that it can be easily calculated. One drawback is, however, with non-linear positions, for example, an extreme movement in the factor does not necessarily generate an extreme portfolio movement.

The diagram below provides and overview of the stress testing methodologies.