Analysis of the Chinese stock market correlations in high frequency data

Abstract

This paper analyzes the correlation structure for various time window intervals using high frequency data of 230 actively traded stocks in Chinese stock market. It is found that the number of nonrandom eigenvalues is more than the developed markets' through the random matrix theory (RMT) analysis. It is also found that the eigenvector components corresponding to the largest eigenvalue cannot be regarded as describing a broad ‘index’ composed of all the stocks as usual. And the analysis of inverse participation ratio (IPR) also proved the above conclusion. Its IPR values are in noise level in the high frequency region.