Model validation and model risk: reaching the end of the line?

In this paper, we argue that the current standard process of model validation can itself be seen as a model and that the quantification of model risk is very often built upon additional assumptions, which results into an infinite regress. Therefore, model validation in its current form does not constitute an endpoint of the critical examination of all models in use by a financial institution. We will show what this observation means for model uncertainty, validation and model selection. Finally, we make suggestions how this infinite regress could be resolved. In the light of our findings, we recommend a more qualitative judgement of remaining model uncertainties with regard to the regulatory and supervisory point of view rather than the introduction of more complexity into the model risk process.