Wavelet Multidimensional Scaling Analysis of European Economic Sentiment Indicators

We propose the use of wavelet coefficients, which are generated from nondecimated discreet wavelet transforms, to form a correlation-based dissimilarity measure in metric multidimensional scaling. This measure enables the construction of configurations depicting the associations between objects across different timescales. The proposed method is used to examine the similarities between the economic sentiment indicators of the EU member states that are published monthly by the European Commission. The results suggest that economic sentiment differs considerably among the member states in the short term. In contrast, several similarities emerge when considering the associations over longer time horizons. These similarities tend to be related to the countries that are geographically close or that exhibited similar economic behaviour prior to the introduction of the euro. Furthermore, the results of a detailed simulation study suggest that the proposed dissimilarity measure is particularly well suited for identifying long-term associations between nonstationary time series.