The limited usefulness of macroeconomic Bayesian VARs when estimating the probability of a US recession

► We estimate the probability of a US recession using Bayesian VAR models. ► In contrast to judgemental probabilities, the BVARs assign very low probabilities. ► Also when survey data are included in the models, the estimated probabilities are low. ► Bayesian VAR models appear to be of limited usefulness for predicting recessions.