Making leading indicators more leading: A wavelet-based method for the construction of composite leading indexes

This paper proposes a novel wavelet-based approach for constructing composite indicators. The wavelet-based methodology exploits the ability of wavelet analysis to analyse the relationships between variables on a scale-by-scale, rather than aggregate, basis. A wavelet-based index which combines several scale-based subindexes is constructed by using a scale-by-scale selection of the components included in the OECD composite leading indicator (CLI) for the US. The comparison with the CLI and its derived measures indicate that the wavelet-based composite index tends to provide early signals of business cycle turning points well in advance of the OECD CLI. Moreover we find that the reliability of the signals tends to increase considerably when the sub-index obtained from the time scale components corresponding to minor cycles, that is, 2-4 years, is removed from the overall wavelet-based index. Keywords: wavelets; composite leading indicators; early warning signals JEL classification: C1; C3; C5; E3