Short-term Inflation Forecast Combination Analysis for Uzbekistan

Abstract

In this paper, we produce the short-term inflation forecast for Uzbekistan, using univariate and multivariate econometric models. In particular, we use Auto Regressive Integrated Moving Average (ARIMA) model, Bayesian Vector Auto regression Model (BVAR) and Vector Error Correction model (VECM) to project CPI inflation and its decomposed subcomponents. The results of the forecast combination analysis are in line with the outcomes of the other research done in this field. The relative performance of combined forecasts based on the RMSE weighting scheme are on average 33% better for 6-month ahead. Despite some individual models demonstrate better performance in certain time horizons, the overall results reveal that forecast combination method permits to reduce the forecast error in comparison with the aforementioned models taken separately.