Extremal Dependence in International Output Growth: Tales from the Tails

The statistical modelling of extreme values has recently received substantial attention in a broad spectrum of sciences. Given that in a wide variety of scenarios, one is mostly concerned with explaining tail events (say, an economic recession) than central ones, the need to rely on statistical methods well qualified for modelling extremes arises. Unfortunately, several classical tools regularly applied in the analysis of central events, are simply innapropriate for the analysis of extreme values. In particular, Pearson correlation is not a proper measure for assessing the level of agreement of two variables when one is concerned with tail events. This paper explores the comovement of the economic activity of several OECD countries during periods of large positive and negative growth (right and left tails, respectively). Extremal measures are here applied as means to assess the degree of cross-country tail dependence of output growth rates. Our main empirical findings are: (i) the comovement is much stronger in left tails than in right tails; (ii) asymptotic independence is claimed by the data; (iii) the dependence in the tails is considerably stronger than the one arising from a Gaussian dependence model. In addition, our results suggest that, among the typical determinants for explaining international output growth synchronization, only economic specialization similarity seems to play a role at extreme events.