Forecasting Romanian GDP Using a BVAR Model

In this study I use the Bayesian VAR framework to forecast the dynamics of output for the Romanian economy. I estimate several versions of Bayesian VARs and compare them in terms of forecasting statistics with two standard models, the OLS and the unrestricted VAR, as well as with a naïve forecast. The findings confirm that the BVAR approach outperforms the standard models. The best BVAR model is used for forecasting quarterly GDP in the short run. The results show that the recovery will be slow and that the output gap will continue to be negative for a few quarters even after the economy starts to grow.

Keywords: forecasting methods, VAR models, Bayesian methods, simulation methods
JEL Classification: C11, C15, C32