Does Using Disaggregate Components Help in Producing Better Forecasts for Aggregate Inflation?

Abstract

This paper analyzes how the information contained in the disaggregate components of aggregate inflation helps improve the forecasts of the aggregate series. Direct univariate forecasting of the aggregate inflation data by an autoregressive (AR) model is used as the benchmark with which all autoregressive (AR), moving average (MA) and vector autoregressive (VAR) models of the disaggregates are compared. The results show that directly forecasting the aggregate series from the benchmark model is generally superior to aggregating forecasts from the disaggregate components. Additionally, including information from the disaggregates in the aggregate model rather than aggregating forecasts from the disaggregates performs best in all forecast horizons when appropriate disaggregates are used. The implication of these results is that better inflation forecasts for Ghana are produce by using information from relevant disaggregates in the aggregate model rather than direct forecasts of the aggregate or aggregating forecasts from the disaggregates.