Density forecasting with Bayesian Vector Autoregressive models under macroeconomic data uncertainty

Macroeconomic data are subject to revisions as later vintages released. Yet, the usual way of generating real-time density forecasts from BVAR models makes no allowance for this form uncertainty. We evaluate two methods that consider uncertainty when forecasting with with/without stochastic volatility. First, model is estimated on vintages. Second, a included, so on, and conditioned estimates revised values. show both these improve accuracy US UK output growth inflation. also investigate how characteristics underlying processes affect performance, provide guidance may benefit professional forecasters.