Comparing the Performance of Different Methods of Forecasting Producer Price Index in Iran

Macroeconomic indices are the essential tools to determine the effects of economic policies, in the short-, medium- and long-run planning and policy makings for an economy. One of the important indices in this area is the producer price index. Therefore, this paper investigates the forecasting of producer price index (PPI) in Iran using seasonal time series data over the period 1990-2017 (1369-1396) through Dynamic Model Selection (DMS) and Dynamic Model Averaging (DMA) in three forecast horizons (one, four, and eight season). In such methods, not only the coefficients but also the forecasting models change over time. The models used in this study (DMA, DMS, BMA, BVAR, TVP and AR) were divided into three categories, large-scale (including 101 variables in nine factor blocks), middle-scale (including 6 variables), and univariate models. The results of the study indicated that forecasting DMS and DMA compared to other approaches has an efficient forecasting performance for the producer price index in Iran.