Title: Estimating the rank of cojumps in high-dimensional financial data with market microstructure noise
Authors: Wenying Yao - Deakin University (Australia) [presenting]
Lars Winkelmann - Freie Universitaet Berlin (Germany)
Abstract: Jumps explain a significant percentage of the overall variation of asset prices. The aim focuses on the reduced rank structure of contemporaneous jumps across assets. The cojump matrix is estimated from high-frequency financial data which are potentially corrupted by market microstructure noise. In general, noise can distort high-frequency statistics, such as volatility and jump estimates. We use the pre-average method to establish a noise-robust estimator of the cojump matrix, which follows a mixed normal distribution. Then the estimation procedure of the cojump rank comes from Random Matrix Theory (RMT), employing the asymptotic results of spiked covariance matrix. This estimator is consistent even when the number of assets and the number of cojump events both diverge to infinity. We use this method to investigate the number of factors in the term structure of U.S. interest rates at macroeconomic news announcement times.