Time-Varying Mixture Copula Models with Copula Selection

Modeling the joint tails of multiple financial time series has many important implications for risk management. Classical models dependence often encounter a lack fit in tails, calling additional flexibility. This paper introduces new semiparametric time-varying mixture copula model, which both weights and parameters are deterministic unspecified functions time. We propose penalized with group smoothly clipped absolute deviation penalty to do estimation selection simultaneously. Monte Carlo simulation results suggest that shrinkage procedure performs well selecting estimating constant models. Using proposed model method, we analyze evolution among four international stock markets, find substantial changes levels patterns dependence, particular around crisis periods.