Uribe Gil, Jorge MarioValencia, Oscar M.2024-12-112024-12-112024https://hdl.handle.net/2445/217020This study offers novel monthly estimates of the latent probability of fiscal crises for 163 countries, from January 1970 to December 2023. These indicators are constructed with minimal data requirements on prices and exchange rates and serve as a global early warning system for fiscal risk. The probabilities are estimated using a Random Forest model within a Mixed-Data Sampling (MIDAS) framework, trained on manually compiled fiscal crisis events. Using these indicators, we test nine hypotheses on the effects of country characteristics, time periods, and policy choices on the probability of fiscal crises. Countries with inflation-targeting regimes, on average, experience lower fiscal distress. Fiscal rules reduce the probability of crises while higher debt levels increase their likelihood. Our findings are particularly relevant for developing countries, where fiscal risk is higher than in advanced economies, even after controlling for policy choices and country-specific characteristics.73 p.application/pdfengcc-by-nc-nd, (c) Uribe Gil et al., 2024http://creativecommons.org/licenses/by-nc-nd/3.0/es/Crisis financeresInflacióAprenentatge automàticFinancial crisesInflationMachine learninginfo:eu-repo/semantics/workingPaperinfo:eu-repo/semantics/openAccess