Estimation of Common Factors for Microstructure Noise and Efficient Price in a High-frequency Dual Factor Model

CINET Number: 2122

Li, Y-N., Chen, J. and Linton, O.


We develop the Double Principal Component Analysis (DPCA) based on a dual factor structure for high-frequency intraday returns data contaminated with microstructure noise. The dual factor structure allows a factor structure for the microstructure noise in addition to the factor structure for efficient log-prices. We construct estimators of factors for both efficient log-prices and microstructure noise as well as their common components, and provide uniform consistency of these estimators when the number of assets and the sampling frequency go to infinity. In a Monte Carlo exercise, we compare our DPCA method to a PCA-VECM method. Finally, an empirical analysis of intraday returns of S&P 500 Index constituents provides evidence of co-movement of the microstructure noise that distinguishes from latent systematic risk factors.

Classification JEL