Modeling the joint tails of multiple nancial time series has important implications for risk management. Classical models for dependence often encounter a lack of t in the joint tails, calling for additional exibility. In this paper we introduce a new nonparametric time-varying mixture copula...