Predicting sovereign debt crises: an early warning system approach

In light of the renewed challenge to construct effective "Early Warning Systems" for sovereign debt crises, we empirically evaluate the predictive power of econometric models developed so far across developed and emerging country regions. We propose a different specification of the crisis variable that allows for the prediction of new crisis onsets as well as duration, and develop a more powerful dynamic-recursive forecasting technique to generate more accurate out-of-sample warning signals of sovereign debt crises. Our results are shown to be more accurate compared to the ones found in the existing literature.