The importance of modeling structural breaks in forecasting Russian GDP

The paper considers two types of models for forecasting seasonally adjusted Russian GDP under the structural breaks. Models that allow breaks in a deterministic trend, in which the dates of structural breaks are set exogenously, and more flexible class of models – with a stochastic trend are considered. It is shown that modeling a structural break in a deterministic trend or adding a stochastic trend significantly improves the quality of 3–4 steps ahead forecasts, and sometimes even on shorter horizons, compared to models with a constant trend growth rate.