Predicting Tail-Risks for the Italian Economy

Abstract This paper investigates the empirical performance of various econometric methods to predict tail risks for the Italian economy. It provides an overview of recent econometric methods for assessing tail risks, including Bayesian VARs with stochastic volatility (BVAR-SV), Bayesian additive regression trees (BART) and Gaussian processes (GP). In an out-of-sample forecasting exercise for the Italian economy, the paper assesses the point, density, and tail predictive performance for GDP growth, inflation, debt-to-GDP, and deficit-to-GDP ratios. It turns out that BVAR-SV performs particularly well for Italy, in particular for the tails. It is then used to also predict expected shortfalls and longrises for the variables of interest, and the probability of specific interesting events, such as negative growth, inflation above the 2% target, an increase in the debt-to-GDP ratio, or a deficit-to-GDP ratio above 3%.