Financial cycles as early warning indicators - Lessons from the Nordic region

Frameworks to handle cyclical systemic risk usually contain a wide selection of early warning indicators. Different indicators sometimes send diverging signals which can be hard to interpret. However, measures of aggregate financial cycles can serve as a way to synthesize information from many indicators. There are however many ways to construct a measure of such cycles. Many methods exist for cycle extraction, variable choice represents another dimension, and cycle aggregation the third. We tackle each step of the way by selecting the best out of six cycle extraction methods, then comparing variables from three groups: credit, house prices and bank funding, and lastly arguing for a simple method of cycle aggregation based on cycle correlation and frequency domain analysis. We then construct a trivariate financial cycle measure which outperforms the ’Basel gap’, all univariate cycles and all other multivariate combinations for the Nordic countries in terms of a noise-to-signal ratio. In addition, it peaks much closer to crisis onset and does relatively well at real-time turning point identification. The trivariate band-pass filtered measure contains the best variable from each group, and outperforms them all. This indicates that aggregate cycles can be more than the sum of their parts, as early warning indicators. Furthermore, we examine potential weaknesses of our analysis in terms of small-sample problems, spurious cycles and the timing of crisis onset. We conclude with 15 lessons from the Nordic countries.