Another Look at the Consumer Price Index - A wavelet Approach

Abstract

A wavelet approach was applied to a consumer price index (CPI) series to address the draw backs of some periodic models. The method requires no assumption of the data generating process but involves the spliting of a given signal into several components with each component reflecting the evolution trough of the signal at a particular time. The multi-level stationary Haar  wavelet decomposition was applied to the series which gave rise to a dyadic sequence of  , and the series was decomposed accordingly using a computer program written for the purpose. Multi-resolution wavelet method was then used to reconstruct the series and the significant details ) that captured the season were added to the  trend  component for the estimatation of the series  The resulting wavelet model was subjected to diagnostic checks and were found to be adequate. Comparative study was carried out with some hilighted CPI models built by some researchers. It was discovered that the wavelet models performs better.

Keywords: Mother wavelets, Haar  wavelet decomposition, Multi-resolution, Auto-correlation and Partial auto-correlation function.