Dynamic Copulas and Long Range Dependence

This paper extends the evolution equation of Patton (2006) for the time variation of the copula parameters by specifying an autoregressive fractionally integrated term. For any copula parameter there is a suitable one-to-one transformation so that the maximum likelihood estimation method may be employed. It is suggested an exploratory tool based on the copula data crossproducts for detecting the presence of long range dependence on the copula level of real data. We simulate from copula models possessing long range dependence and work out two examples using real data. Modeling long range dependence on the level of dynamic copulas has the potential for providing improved forecasts and are useful for financial and economic applications.