Forecasting Models for Components of the Consumer Price Index in Azerbaijan

The paper examines the main determinants of prices of food, non-food, and services components of the consumer basket of Azerbaijani households and provides forecasting models based on those determinants. For this purpose, I build vector autoregression models using quarterly data from 2003Q1 to 2020Q3. I find that the prices of items in the consumer basket are determined by both domestic and foreign factors. World food prices, consumer prices in trading partners, non-oil import weighted nominal effective exchange rate, real GDP, agricultural producer prices and own lags of food prices are main determinants of food prices, while consumer prices in trading partners, non-oil import weighted nominal effective exchange rate, output gap, manufacturing producer prices and own lags of non-food prices are determinants of non-food prices. However, services prices is mainly affected by non-oil import weighted nominal effective exchange rate, M2 (money supply), food consumer prices, non-oil tax, and own lags of services prices. In this paper, I also produce one-, two-, four-, six-, and eight-quarter forecasts for each component, combine the forecasts using corresponding weights and compare their forecasting errors to that of the direct forecasting model. The results show that the combination method has better forecasting performance than that of the direct method.