Forecasting Russian GDP, Inflation, Interest Rate, and Exchange Rate Using DSGE-VAR Model

Abstract

This paper compares, on Russian quarterly data for 2003–2021, the forecast performance of a small-scale dynamic stochastic general equilibrium model (DSGE model) and the DSGE-VAR model as a Bayesian vector autoregression (VAR), which uses priors from this DSGE model. The forecast performance of the DSGE-VAR model turns out to be higher than that of the DSGE model for output growth and inflation over the one-year horizon and for the interest rate and exchange rate over a two-year horizon. Meanwhile, the DSGE-VAR model, on average, predicts GDP, inflation, and the exchange rate better and the interest rate worse than the first-order autoregressive model that serves as a benchmark.