The Comparison of Forecasting Performance of a DSGE Model With Stochastic Volatility and a Bvar Model

This paper compares point based macroeconomic forecasting performance of two models on various time horizons. Point – based macroeconomic forecasts in the following work are obtained using structural and non-structural models. The structural model is represented by New-Keynesian small-scale DSGE model in two specifications: the first specification includes stochastic volatility (DSGE- SV), and second does not (DSGE). Non-structural models are represented by VAR and Bayesian VAR (BVAR). Accuracy is estimated using standard metric - root of mean squared forecast error (RMSFE). The results suggest that, DSGE – SV performs quite poorly in comparison with non-structural models while forecasting interest rate and inflation on all time horizons. At the same time, by utilizing DSGE –SV it is possible to obtain a forecast for output growth rates more accurate than BVAR model forecasts.