Forecasting Corporate Defaults in the German Stock Market

Abstract

In this paper, we estimate and test several default risk models using new and unique data on corporate defaults in the German stock market. While defaults were extremely rare events in the 1990s, they have been a characteristic feature of the German stock market since the early 2000s. Here, we apply the structural Merton distance to default (DD) as well as several reduced-form models. A variety of performance evaluation tools, including receiver-operating-characteristics analysis, calibration tests and a loan market simulation, are used, which suggest that the Campbell, Hilscher and Szilagyi failure score outperforms the other models. Although the performance evaluation metrics show that the failure score performs slightly worse in our case than for US data, we recommend it as a benchmark default risk model for research as well as the industry. We show that lenders who apply this model are more profitable than those who apply others. Moreover, we warn of several pitfalls associated with the Altman z-score and the DD. The former has very weak discriminatory power and the latter is severely miscalibrated.

Keywords: default risk, credit risk, risk management, forecasting, internal models.