Drivers of firm-level tail dependence: A machine learning approach

Abstract

The paper studies the determinants of firm-level tail dependence of companies with respect to foreign markets using machine learning. We measure dependence for a comprehensive international set of firms using copulas and we find that left tail dependence is consistently stronger than right tail dependence with their gap widening in recessionary periods. We then apply random forest regressions to identify and characterize the factors that account for the total panel variation of tail risk. The World Uncertainty Index, the R2 integration measure and coskewness with respect to foreign markets are the most important determinants. For US firms individual ownership variables such as the number of total or foreign investors dominate the remaining firm-level characteristics in explaining tail dependence. Our results contribute to the understanding of crash risk in the modern global financial landscape with implications for asset managers.