Early warning systems for economic crises in South Africa.

Abstract

This paper develops a series of Early Warning System models for debt crises. This paper uses a Debt Pressure index to define crisis periods and then demonstrates how one can go about trying to forecast these periods using Logit and Markov-switching Models. An alternative approach, whereby ordinary least squares (OLS) is used to create Early Warning System models, is introduced. A graphical analysis is also conducted. Three useful Early Warning System models emerge from this study.