No abstract

Research question This paper deals with stress tests for credit risk and shows how exploiting the discretion when setting up and implementing the underlying model can drive the results of a quantitative credit risk stress test for default probabilities. Contribution We contribute to the scarce literature on model and estimation risk in stress tests. We employ several variations of a CreditPortfolioView-style model using US data ranging from 2004 to 2016 and compare the forecasted default probabilities of these models. Our clear focus on stress tests is the aspect that differentiates our paper from existing studies most. This is particularly relevant against the background of regulatory stress tests which have become more important in recent years. Results and policy implications This paper shows that stress forecasts of default probabilities highly depend on the modelling assumptions and that seemingly only minor variations can affect the results of stress tests considerably. That said, our findings reveal that the conversion of a shock (i.e., stress event) increases the (non-stress) default probability by 20% to 80%-this high range can be explained by the sensitivity of stress test models to model and estimation risk. Interestingly, forecasts for non-stress default probabilities are less exposed to model and estimation risk. In addition, the risk horizon over which the stress default probabilities are forecasted and whether we consider mean stress default probabilities or high quantiles seem to play only a minor role for the dispersion between the results of the different model specifications. These findings emphasize the importance of extensive robustness checks for model-based credit risk stress tests, particularly in regulatory stress tests.

Research questionThis paper deals with stress tests for credit risk and shows how exploiting the discretion when setting up and implementing the underlying model can drive the results of a quantitative credit risk stress test for default probabilities. ContributionWe contribute to the scarce literature on model and estimation risk in stress tests. We employ several variations of a CreditPortfolioView-style model using US data ranging from 2004 to 2016 and compare the forecasted default probabilities of these models. Our clear focus on stress tests is the aspect that differentiates our paper from existing studies most. This is particularly relevant against the background of regulatory stress tests which have become more important in recent years. Results and policy implicationsThis paper shows that stress forecasts of default probabilities highly depend on the modelling assumptions and that seemingly only minor variations can affect the results of stress tests considerably. That said, our findings reveal that the conversion of a shock (i.e., stress event) increases the (non-stress) default probability by 20% to 80% -this high range can be explained by the sensitivity of stress test models to model and estimation risk. Interestingly, forecasts for non-stress default probabilities are less exposed to model and estimation risk. In addition, the risk horizon over which the stress default probabilities are forecasted and whether we consider mean stress default probabilities or high quantiles seem to play only a minor role for the dispersion between the results of the different model specifications. These findings emphasize the importance of extensive robustness checks for model-based credit risk stress tests, particularly in regulatory stress tests.