Measurement And Patterns Of International Synchronization-- A Spectral Approach

This paper examines the issue of international synchronization of cycles. Using spectral methods we analyze the pattern of co-movement (coherences) of growth rate cycles between countries across frequency bands and overtime. We also examine the lead-lag structure (phase shifts) of country cycles obtained from spectral methods and evaluate these against the reference chronology given by the Economic Cycle Research Institute (ECRI) based on the NBER methodology. These parameters are studied across three frequency bands, growth rate cycle frequency, low and high frequency. We also report partial coherences and confidence intervals based on Gaussian approximations to the distribution of the sample coherence which follows a complex Wishart distribution. To characterize growth rate cycles, the paper uses the growth rate of the coincident index of economic activity2 given by ECRI, a composite of variables that represent current economic activity for various countries and country groups over the period 1974 to 2010. We also evaluate how the character of co-movements has changed overtime by analyzing the sample over two periods, 1974-1990 and 1991-2010. We find high feedback effects between country cycles, and that average partial coherences are higher during the period 1991-2010 over that in 1974-1990 in at least one frequency band. During the period 1991-2010, for almost all paired comparisons, these have risen in the growth rate cycle frequency. Additionally, for some pairs, coherences rise for longer cycles (low frequency) while for others, they increase in the higher frequency band. Average phase shifts over growth rate cycle frequency indicate that the synchronization is faster in the latter period. Finally, in assessing spectral results against the reference chronology of growth rate cycles given by ECRI, we find that both methodologies yield comparable results.