CMStatistics 2021: Start Registration
View Submission - CFE
A0194
Title: Forecasting sovereign defaults Authors:  Ana Galvao - University of Warwick (United Kingdom) [presenting]
Michael McCracken - Federal Reserve Bank of St. Louis (United States)
Michael Owyang - Federal Reserve Bank of St Louis (United States)
Abstract: Six countries defaulted partially in their debt obligations in 2020; the highest number of defaults recorded by Moodys-rated sovereign bonds since 1983. We propose a novel Panel Qualitative Vector Autoregressive (Qual-VAR) model to measure and forecast the probability of sovereign defaults. We use the sovereign default events recorded previously, a set of global factors (activity and financial factors) and country-specific macroeconomic variables (such as government debt, external debt and FDI inflows to GDP ratios and GDP growth) sampled quarterly from 1980 to 2020 to estimate our proposed quantitative model for a panel of countries. Our panel of 50 countries includes emerging European, Asian, African, and Latin American economies. Using the panel Qual-VAR model, we can compute multi-step ahead forecasts for the probability of sovereign default for each country in our sample while allowing for serial correlation in the probability of default. The model is also useful to compute probability forecasts conditional on paths for the key global factors considered. Finally, the panel Qual-VAR allows us to assess the contribution of global versus country-specific factors in explaining sovereign default risks.