EMPIRICAL PRICE SURVEY IN BiH 2006 – 2010 АND INFLATION FORCASTING FOR 2011 – USE OF ARIMA MODEL

Abstract

The survey of price trends during the period from 2005 to 2010 was aimed at estimating the best quality of ARIMA models used to forecast consumer price index and/or inflation in BiH in 2011. A very process to create and select ARIMA models was based on the recommended Box-Jenkins procedure involving data collection for the selected time series, identification of its stationary level, creation of model, diagnostic checking, model stationary checking (simulation of real series) and finally inflation forecasting based on the selected models. This is an iterative process, i.e. it is repeated over and over again to obtain a set of desired and efficient models used for inflation forecasting. Out of 64 tested models following the Akaike Info Criterion (AIC), Schwarz Criterion (SIC), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE) and Theil Inequality Coefficient (TIC) criteria, three ARIMA models identified to fit best the forecasting procedure were selected. Their efficiency and predictive power will be tested during the course of 2011.