Title: Intraday variance-covariance matrix estimation: A point process approach
Authors: Ingmar Nolte - Lancaster University (United Kingdom) [presenting]
Sandra Nolte - Lancaster University (United Kingdom)
Yifan Li - Lancaster University (United Kingdom)
Abstract: The point process based volatility estimator provides an important alternative to the popular Realized Variance (RV)-type estimators in estimating the high-frequency volatility. It has been shown that the volatility estimates from the duration or intensity-based model performs at least equally well with the estimates from the RV-type models. Moreover, this type of volatility estimator has several advantages over the RV-type ones, in particular the fully parametric volatility estimation which utilizes data beyond the daily aggregation window. The parametric design also allows for the inclusion of other market microstructure covariates. We propose an estimator of the high-frequency variance-covariance matrix based on a point process approach by extending the univariate intensity-based framework. We show that our estimator provides reliable variance-covariance estimates while inheriting the advantages of the univariate version over the realized approach. In principle, our estimator can be applied to any sets of price series that provide enough observations in each series for intraday covariance estimation. The parametric structure allows intraday covariance inference that takes the autoregressive structure and diurnal pattern into consideration.