Forecasting Inflation in Mongolia: A Dynamic Model Averaging Approach

Abstract

Abstract This paper investigates the use of the dynamic model averaging (DMA) approach for identifying good inflation predictors and forecasting inflation in Mongolia, one of the most commodity-dependent economies. The DMA approach allows for both a set of predictors for inflation and marginal effects of predictors to change over time. Our empirical work resulted in several novel findings. First, external variables (i.e., China’s growth, China’s inflation, and change in oil price) play an important role in forecasting inflation and change considerably over time and over forecast horizons. Second, among domestic variables, wage inflation and M2 growth are currently the best predictors for short and longer forecast horizons. Third, the use of DMA leads to substantial improvements in forecast performance, and DMA (2,15) with the chosen forgetting factors is the best performer in predicting inflation for Mongolia.