Analyzing tail dependencies using simple regression models

The Global Financial Crisis shows us the importance of managing the downside risk in the financial market. In this paper we investigate the bivariate and multivariate tail dependence structure across seven international stock markets from North America, Europe and Asia. Van Oordt and Zhou (2012) introduce the linear indicator regression model to investigate the tail dependence structure. However, the amount of parameters to be estimated is very large. Therefore, we introduce two variables selection methods and the logit indicator regression. We apply the variable selection to the data set and then perform the linear and logit regression. We see that the differences between the estimations of the regressions are insignificant. Furthermore, we observe distinction between the stock markets in the multivariate analysis that does not appear from the bivariate analysis. Therefore, we recommend using the logit indicator regression in a multivariate setting. Finally, we find the dependence structure within European markets is stronger than across other markets.