A0349
Title: Assessing causality in the tails: Measurement and testing
Authors: Bikramjit Das - Singapore University of Technology and Design (Singapore) [presenting]
Xiangyu Liu - Singapore University of Technology and Design (Singapore)
Abstract: A measure of extreme tail association (ETA) is defined between two variables, which is asymmetric, easily computable in sample data, and consistent for the population measure. The asymptotic normality of the sample measure is exhibited under mild conditions on the underlying distribution. A test for bivariate tail asymmetry is also proposed, and asymptotic distributions are computed under both null and alternative hypotheses. Confidence regions for real data are computed using a multiplier bootstrap method. The measures and tests established allow inferring causality relationship between two variables. Data from movement in cryptocurrency prices is used to exhibit the the tools developed.
