Characterising the financial cycle: a multivariate and time-varying approach
We introduce a methodology to characterise financial cycles combining a novel multivariate spectral approach to identifying common cycle frequencies across a set of indicators, and a time varying aggregation emphasising systemic developments. The methodology is applied to 13 European Union countries as well a synthetic euro area aggregate, based on a quarterly dataset spanning 1970-2013. Results suggest that credit and asset prices share cyclical similarities, which, captured by a synthetic financial cycle, outperform the credit-to-GDP gap in predicting systemic banking crises on a horizon of up to three years. Financial cycles tend to be long, particularly in upswing phases and with important dispersion across country cases. Concordance of financial and business cycles is observed only 2/3 of the time. While a similar degree of concordance for financial cycles is apparent across countries, heterogeneity is high - whereby a cluster of countries tends to exhibit a high synchronisation in their financial cycle phases.