How to Predict Financial Stress? An Assessment of Markov Switching Models

Abstract

This paper predicts phases of the financial cycle by combining a continuous financial stress measure in a Markov switching framework. The debt service ratio and property market variables signal a transition to a high financial stress regime, while economic sentiment indicators provide signals for a transition to a tranquil state. Whereas the in-sample analysis suggests that these indicators can provide an early warning signal up to several quarters prior to the respective regime change, the out-of-sample findings indicate that most of this performance is due to the data gathered during the global financial crisis. Comparing the prediction performance with a standard binary early warning model reveals that the MS model is outperforming in the vast majority of model specifications for a horizon up to three quarters prior to the onset of financial stress.

Keywords: time-varying transition probability Markov switching model, early warning model, continuous coincident financial stress measure

JEL Classification: C54, G01, G15