Forecasting the Inflation Rate in Poland and U.S. Using Dynamic Model Averaging (DMA) and Google Queries

The purpose of this paper is to apply the recently proposed Dynamic Model Averaging (DMA) to modelling the inflation rate in U.S. and Poland with the additional analysis of the usefulness of Google Trends data. One of the analysed economies is quite uniform, but the time-series available for it are quite short. The second is the developed leading economy. It is found that in the case of U.S. the DMA methodology is quite useful and produces more accurate forecasts than the alternative ones. In particular, all features of DMA (i.e., model averaging, time-varying parameters, dynamically updated weights in model averaging) improve the forecast quality. Similar analysis for Poland does not lead to such conclusions. As two types of models are considered for the U.S. (with long and short time-series) it can be suspected that the problem with applying DMA to the Polish economy comes from the length of the available time-series. Anyway, in the case of U.S. inflation the DMA produced interesting outcomes, i.e., time-varying inflation drivers could have been identified. The practical implications for Poland are that unemployment rate is the major driver of inflation. For the U.S., the drivers are change in number of new houses, money supply, stock prices, energy prices, industrial production and level of short-term interest rate, government longterm bond yield and term spread.