Department of Economics, UCSD

An emerging literature in time series econometrics concerns the modeling of potentially nonlinear temporal dependence in stationary Markov chains using copula functions. We obtain conditions that imply a geometric rate of mixing in models of this kind. A geometric rate of beta-mixing is shown to obtain under a rather strong condition that rules out asymmetry and tail dependence in the copula function. Rho-mixing, which implies a geometric rate of alpha-mixing, is obtained under a much weaker condition. We verify one or both of these conditions for a range of parametric copula functions that are opular in applied work.