Predicting Defaults with Regime Switching Intensity: Model and Empirical Evidence

Abstract

In this paper, we propose a regime-switching intensity model to predict the occurrences of default events. Individual firm's default intensity function, i.e., the instantaneous default probability function, is determined by both observable risks factors and an unobservable regime indicator. We provide estimation algorithm when the regime indicator follows the discrete-time Markovian process. The empirical study on U.S. listed corporations suggests three key criteria when specifying the intensity function. They are systematically and objectively chosen observable risk factors, a common frailty variable, and the regime-dependent risk exposure to firm's leverage proxy or observable risk factors. Comparing to classical default intensity models, the regime-switching intensity model with these components has better performance in the in-sample fit and in the out-of sample prediction ability.