Forecasting Key Macroeconomic Indicators Using DMA and DMS Methods

Abstract

This paper studies the application of dynamic model averaging (DMA) and dynamic model selection (DMS) methods for forecasting the main macroeconomic indicators of Russia: gross domestic product, consumption, gross capital formation, and gross fixed capital formation at both constant and current prices for one to six quarters horizon. The key point of these methods is their ability to change, at each time step, both the coefficients on the independent variables and the composition of the predictors involved in the forecasting, which in theory can provide more accurate and reliable forecasts. The study reveals higher quality in forecasts of the major macroeconomic indicators using DMA and DMS methods compared to the use of alternative methods such as na.ve forecasts, first- and second-order autoregressions, first-order autoregression with exogenous variables, a dynamic factor model, Bayesian vector autoregression, the Bayesian model averaging method, and Bayesian model selection. The results obtained substantiate the practical significance of DMA and DMS methods and show the potential of their use in decision-making in a constantly changing economic environment.