Real-time conditional forecasts with Bayesian VARs: An application to New Zealand

Abstract We develop a Bayesian VAR (BVAR) to produce conditional forecasts for the New Zealand economy. In a real-time out-of-sample forecasting exercise, we find that the BVAR outperforms a selection of other time series models, and it yields forecasts of similar accuracy to the forecasts produced internally at the Reserve Bank of New Zealand. Examining shock decompositions, we also highlight the importance of foreign shocks for the New Zealand economy. Our results suggest that the BVAR is a useful tool for policy making in real time.