Article ID Journal Published Year Pages File Type
6869087 Computational Statistics & Data Analysis 2016 22 Pages PDF
Abstract
A dynamic copula model is introduced, in which the copula structure is inferred from the realized covariance matrix estimated from within-day high-frequency data. The estimation is carried out in a method-of-moments fashion using Hoeffding's lemma. Applying this procedure day by day gives rise to a time series of daily copula parameters which can be approximated by an autoregressive time series model. This allows one to capture time-varying dependence. In an application to portfolio risk-management, it is found that this time-varying realized copula model exhibits very good forecasting properties for the one-day ahead value at risk.
Related Topics
Physical Sciences and Engineering Computer Science Computational Theory and Mathematics
Authors
, ,