CMStatistics 2016: Start Registration
View Submission - CFE
A1637
Title: Nonlinear conditional Value-at-Risk Granger causality Authors:  Cees Diks - University of Amsterdam (Netherlands) [presenting]
Marcin Wolski - European Investment Bank (Luxembourg)
Abstract: A new methodology is proposed to assess the effects of individual companies' risk on other institutions and on the system as a whole. We build upon the Conditional Value-at-Risk (CoVaR) approach, introducing explicit Granger causal linkages while accounting for possible nonlinearities in the financial time series considered. The resulting causality measure is referred to as Nonlinear CoVaR (NCoVaR) Granger causality. The natural U-statistics estimator of NCoVaR Granger causality is shown to be asymptotically normally distributed with a long-term variance that can be consistently estimated using a HAC estimator, allowing us to construct a test for the absence of NCoVaR Granger causality. We investigate our testing approach numerically, and find it to have good size and power properties. In an empirical application we assess the feedback risk transmissions between sovereigns and banking sectors in the Euro area.