Financial Crises and the State of the Real Economy: An Extreme Value Approach

Financial crises typically occur both during economic recessions and expansions. The objective of this paper is to quantify the likelihood of financial crises and crisis spill-overs across the business cycle in order to assess whether and to what extent economic recession episodes are more inclined towards financial crises and crisis co-movements than expansion periods. Statistical extreme value analysis (EVT) is put at work to calculate these marginal and joint tail likelihoods for recessions and expansion subsamples. We find that tail risk increases during recessions for most financial assets. The same seems to happen with extreme linkages between financial markets. Moreover, cross-asset crisis spillovers like flight-to-quality effects between stocks, bonds or gold become more pronounced during recessions. Finally, we show that diversifying portfolio tail risk becomes more difficult during recessions. To our knowledge, applying EVT techniques to economically meaningful sample partitions is novel to the literature on financial extremes and extreme value analysis. From a statistical point of view, neglecting the existence of regimes in tail behaviour biases full sample estimates of measures of tail fatness or tail dependence because the presence of regimes induces higher order tail behaviour. EVT measures can also be made dependent on multiple regimes and regime determination can be made endogenous.

This event is organised by the Econometric Institute.
Twitter: @MetricsSeminars