XXV Edition

1-2 December 2016"

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

Thibaut Duprey, Bank of Canada
Benjamin Klaus, European Central Bank

This paper bridges the gap between the business cycle literature and the literature on currency, banking and financial crises by comparing the continuous Markov switching (MS) approach used to identify recessions with a binary logit model widely used to predict currency and banking crises. The two models are assessed in the context of the financial cycle for the EU countries, captured by the financial stress index of Duprey et al. (2015). (i) Overall, both models have a relatively similar ability to predict high financial stress episodes, with the MS model outperforming the logit model between six to one quarters prior to the onset of financial stress episodes. (ii) Across country, the MS model is able to distinguish the drivers pushing the economy in and out of high financial stress. Debt service ratios and housing variables are found to be predictors of a transition to a high financial stress regime, while equity price growth and economic sentiment indicators provide signals for a transition to a tranquil state. (iii) Last, country-specific estimation are possible within the MS framework, so that the identification of country-specific predictors might be valuable for national authorities.

Area: Financial Stability

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

Paper file

University Network