Predicting Default Probabilities for Stress Tests: A Comparison of Models

TL;DR

This paper compares various regression models, including machine learning approaches, to identify the most effective for translating macroeconomic variables into default probabilities in stress testing, highlighting the superiority of certain models over the current standard.

Contribution

It systematically evaluates a wide range of models within a unified framework, revealing the advantages of machine learning and forecast combination methods for credit risk prediction.

Findings

Machine learning models outperform traditional regression models.

Forecast combinations improve prediction accuracy.

Some models surpass the current state-of-the-art in default probability estimation.

Abstract

Since the Great Financial Crisis (GFC), the use of stress tests as a tool for assessing the resilience of financial institutions to adverse financial and economic developments has increased significantly. One key part in such exercises is the translation of macroeconomic variables into default probabilities for credit risk by using macrofinancial linkage models. A key requirement for such models is that they should be able to properly detect signals from a wide array of macroeconomic variables in combination with a mostly short data sample. The aim of this paper is to compare a great number of different regression models to find the best performing credit risk model. We set up an estimation framework that allows us to systematically estimate and evaluate a large set of models within the same environment. Our results indicate that there are indeed better performing models than the current state-of-the-art model. Moreover, our comparison sheds light on other potential credit risk models, specifically highlighting the advantages of machine learning models and forecast combinations.