Journal of Risk Model Validation

Risk.net

A model combination approach to developing robust models for credit risk stress testing: an application to a stressed economy

Georgios Papadopoulos

  • A model combination approach for macro-financial model development is implemented.
  • Historical adverse conditions are used to make scenario-conditional forecasts.
  • The use of individual models can result in misspecification of credit risk.
  • Model combinations display better and more stable forecasting performance.

An integral part of advanced stress-testing frameworks is the macrofinancial model, which maps the impact of macroeconomic shocks on bank-specific risk factors. The standard practice for the development of such models is the use of a single model. However, such an approach exposes the whole framework to model risk, since any individual model may be misspecified and/or affected by structural changes in ways that can be very difficult to predict in advance. To address this issue, we use a model combination approach to develop robust macrofinancial models for credit risk stress testing. The empirical part utilizes data from the Greek economy, which experienced a sharp change from normal to distressed conditions; this makes it a particularly challenging case to forecast. The results expose the inadequacy of any individual model to capture the evolution of credit risk and indicate that relying on a single model can result in significant misestimation of risk. This holds whether the estimation sample includes only precrisis information or data from both tranquil and stressed periods. However, model combination gives a better forecasting performance than standard benchmark models. Overall, the proposed approach can lead to a more robust assessment of risk.

To continue reading...

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an indvidual account here: