CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: The beta-adjusted covariance estimator Authors:  Kris Boudt - Vrije Universiteit Brussel (Belgium)
Steven Vanduffel - Vrije Universiteit Brussel (Belgium)
Kirill Dragun - VUB (Belgium) [presenting]
Abstract: Stock return covariance estimation is proposed to be improved by imposing the equality between the covariance matrix-implied stock-ETF covariance, and the estimated stock-ETF pairwise covariance. The proposed beta adjusted covariance estimation iteratively projects the realized covariance on an improved covariance respecting the constraints. The simulation study confirms that the proposed estimator efficiently deals with biased approximations by traditional estimators caused by asynchronous trading data and significantly improves accuracy of the estimated covariances.