Coexceedances in Financial Markets - a Quantile Regression Analysis of Contagion

This article introduces a new model to analyze financial contagion based on a modified coexceedance measure. We use the quantile regression framework to examine the occurrences and the degrees of coexceedances. Contagion is defined as the crisis specific coexceedance not explained by the covariates for different quantiles. Our approach can identify the extent of contagion and also reveal linear and non-linear linkages between contagion and its determinants. Estimation results for daily stock index returns show that bsomeQ contagion exists and is predictable within and across regions. Furthermore, contagion depends on a regional (world) market return and its volatility and is stronger for extreme negative returns than for extreme positive returns. An analysis of the evolution of coexceedances additionally reveals clusters of extremes. Finally, the computation of conditional densities shows the impact of different influence factors on the entire conditional distribution of coexceedances.

BAUR Dirk;  SCHULZE Niels; 

2006-01-31

Elsevier

JRC32470