Please use this identifier to cite or link to this item: http://hdl.handle.net/10419/39326
Authors: 
Hautsch, Nikolaus
Kyj, Lada M.
Oomen, Roel C.A.
Year of Publication: 
2009
Series/Report no.: 
SFB 649 discussion paper 2009,049
Abstract: 
We introduce a regularization and blocking estimator for well-conditioned high-dimensional daily covariances using high-frequency data. Using the Barndorff-Nielsen, Hansen, Lunde, and Shephard (2008a) kernel estimator, we estimate the covariance matrix block-wise and regularize it. A data-driven grouping of assets of similar trading frequency ensures the reduction of data loss due to refresh time sampling. In an extensive simulation study mimicking the empirical features of the S&P 1500 universe we show that the 'RnB' estimator yields efficiency gains and outperforms competing kernel estimators for varying liquidity settings, noise-to-signal ratios, and dimensions. An empirical application of forecasting daily covariances of the S&P 500 index confirms the simulation results.
Subjects: 
covariance estimation
blocking
realized kernel
regularization
microstructure
asynchronous trading
JEL: 
C14
C22
Document Type: 
Working Paper

Files in This Item:
File
Size
486.86 kB





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.