CMStatistics 2016: Start Registration
View Submission - CFE
Title: Conditional quantile-dependent autoregressive value-at-risk Authors:  Giovanni Bonaccolto - University of Padova (Italy) [presenting]
Massimiliano Caporin - University of Padova (Italy)
Sandra Paterlini - European Business School Germany (Germany)
Abstract: The Conditional Value-at-Risk (CoVaR) and the Conditional Autoregressive Value-at-Risk (CAViaR) have shown to be useful tools in evaluating the connections among financial institutions in the context of extreme events. We show that it is possible to obtain relevant improvements by conditioning the estimation of CoVaR and CAViaR with respect to the support of the depedent or the explanatory variables. This corresponds to the introduction of Quantile-on-Quantile dependencies, or, equivalently, to moving toward non-parametric quantile regression estimation. Our analysis is applied on a large dataset consisting of the returns generated by the financial institutions operating in the U.S. market in the years 2000-2015 and the empirical findings support our methodological contributions.