A Predictive Model for Financial Crises: An Exploratory Public Policy Tool

Abstract

This paper aims to establish an early warning system for Malaysia to project future fnancial crises. We investigate the macroeconomic and fnancial factors in predicting fnancial crises using the logit econometric model and monthly data with the sample period from January 1980 to December 2003. Four indicators namely regional crisis dummy, current account balance, M2 growth and real exchange rate show signifcance and therefore, are incorporated in crisis prediction. The early warning system’s predictive power is then examined against various crises periods including the 1997 Asian fnancial crisis. The empirical results show that the early warning system exhibits forecasting ability. Both in-sample and out-ofsample performance evaluations affrm the system’s ability to predict the 1997 Asian fnancial crisis with at least one early warning signal issued within the 12-month period prior to the actual crisis. The early warning system should be treated as a supplement to policymakers existing forecasting tools in estimating the probability of crisis and assessing a country’s vulnerability rather than as an exclusive tool for crisis prediction.