Title: Clustering estimator of the HAC for high-frequency data
Authors: Anastasija Tetereva - University St Gallen (Switzerland) [presenting]
Ostap Okhrin - Dresden University of Technology (Germany)
Abstract: A computationally simple estimator is introduced for the multivariate hierarchical Archimedean copulae. It is proposed to estimate the structure and the parameters of a copula simultaneously based on the correlation matrix only. The advantage of the average correlation estimator is the significant reduction of the computational costs and that it can be used in cases when the maximum likelihood type estimation can not be performed. Extensive simulation studies show the superior performance and the low computational costs of the proposed estimator in comparison to the benchmark models.In the case of high-frequency data, the proposed algorithm enables the estimation based on the realized covariance matrix. The application of the estimator to the one-day-ahead Value at Risk prediction using high-frequency data gives rise to the hierarchical realized copula (RHAC). The RHAC exhibits good forecasting properties for a multivariate portfolio in comparison to the dynamic copula and realized covariance models and does not suffer under the curse of dimensionality.