RDP 2008-04: A Small BVAR-DSGE Model for Forecasting the Australian Economy 6. Conclusions

With the principal objective of macroeconomic forecasting, we have used a simple, small open economy DSGE model to provide prior information for a structural Bayesian VAR model. The performance of the BVAR-DSGE model in forecasting the key variables of output growth and inflation is competitive with the three benchmark models we have considered; for example, its inflation forecasts outperform those from the DSGE alone at most horizons. However, the DSGE model we have used as the source of prior information is particularly simple. Future work could extend the BVAR-DSGE model in at least two ways: introducing common features used to improve the fit of DSGE models (such as habit persistence in consumption); and improving its open economy aspects. The main result we take from our analysis is that the BVAR-DSGE methodology is a useful way of balancing theoretical and data coherence, particularly when the aim is to build a model for forecasting.