Improving the quality of the forecast through methods of combining forecasts

Abstract

Subject of research: The article is devoted to the comparison of forecasting methods and methods of combining forecasts when forecasting average monthly wages for some regions of the Russian Federation for the period from January 2013 to July 2022. Purpose of the study: To test the hypothesis of improving forecast quality through forecast pooling methods. Methods and objects of research: Time series of average monthly wages in the Khanty-Mansi Autonomous Okrug Yugra, Yamal-Nenets Autonomous Okrug and Sverdlovsk region are used to compare forecasting methods and methods of combining forecasts. The paper considers forecasting methods: TBATS, ARIMA, exponential smoothing, ETS, Theta, STL, polynomial regression and the approach in combining forecasts by the Granger-Ramanathan method. Main results of research: The paper presents the results of comparison of forecasting methods and approach in combining forecasts by Granger-Ramanathan method. Time series of average monthly wages were taken from statistical collections of Rosstat. The forecast horizon was set at 12, 18 and 24 points. The Granger-Ramanathan method showed that in most cases it is possible to improve the quality of the forecast by combining private forecasts.