The paper compares the accuracy of standard econometric models and machine learning algorithms in forecasting fiscal crises in emerging markets and low-income countries, and identifies the variables that are the best predictors of fiscal crises.
The analysis shows that machine learning algorithms allow for more accurate forecasts of fiscal crises compared with traditional approaches. The best predictors are the stock of public debt, especially the component held by foreign investors, the historical frequency of similar crises, and the quality of public institutions. Finally, we derive aggregate risk indices that point to a significant increase in the probability of fiscal crises after the pandemic.
