A BVAR Model for Forecasting Ukrainian Inflation and GDP

Abstract

In this paper, I examine the forecasting performance of a Bayesian Vector Autoregression (BVAR) model with a steady-state prior and compare the accuracy of the forecasts against the QPM and official NBU forecasts during the Q1 2016–Q1 2020 period. My findings suggest that inflation forecasts produced by the BVAR model are more accurate than those of the QPM model for two quarters ahead and are competitive for a longer time horizon. The BVAR forecasts for GDP growth also outperform those of the QPM but for the whole forecast horizon. Moreover, it is revealed that the BVAR model demonstrates a better performance compared to the NBU’s official inflation forecasts over the monetary policy horizon, whereas the opposite is true for GDP growth forecasts. Future research may deal with estimation issues brought about by COVID-19.