WP/13/01 Short-Term Inflation Forecasting in Belarus
The paper outlines the methodology of obtaining inflation forecasts published in the bulletin “Belarus Short-Term Economic Trends: Inflation”. We describe the properties of the statistical data, as well as the nature of the approaches applied to inflation forecasting: autoregressive model (AR), autoregressive moving average model (ARMA), forecasting inflation as a weighted sum of inflation components, vector autoregression (VAR), error correction model (ECM), and P* model. We compare the forecasting performance of the models using the pseudo-out-of-sample forecasting procedure over January 2011 – October 2012. According to the results, the best forecasting performance is demonstrated by the disaggregated models, AR models with multiple lags, ECM models based on the demand for nominal M3 aggregate. The forecasting performance of the ARMA and VAR models is relatively bad, although some VARs generate good forecasts for some periods. Forecast averaging in most cases do not substantially improve the forecasting performance of the individual models.