Description

In this paper an exercise is performed to determine the usefulness of utilizing disaggregated price data to forecast headline inflation more accurately. A number of methods based on univariate and multivariate autoregressive models are used for different levels of disaggregation for a period of stable inflation and a period of accelerating inflation. The results show that a certain level of disaggregation could be beneficial when inflation is not low and stable, suggesting that under certain circumstances the disaggregate approach captures the underlying dynamics of inflation more efficiently. The benefits are noticeable for the three-, six- and twelve-month horizons, as opposed to the one-month horizon, where improvements seem negligible.