A Random Matrix Approach to Cross-Correlations in Financial Data

TL;DR

This paper applies random matrix theory to analyze cross-correlations in stock returns, identifying significant eigenvalues and eigenvectors related to market-wide and sector-specific influences, with implications for portfolio stability.

Contribution

It introduces a systematic RMT-based method to distinguish meaningful correlations from noise in large financial datasets, revealing stable sector structures.

Findings

Most eigenvalues fall within RMT bounds, indicating randomness.

Eigenvectors outside RMT bounds correspond to market and sector influences.

Deviating eigenvectors are stable over time.

Abstract

We analyze cross-correlations between price fluctuations of different stocks using methods of random matrix theory (RMT). Using two large databases, we calculate cross-correlation matrices C of returns constructed from (i) 30-min returns of 1000 US stocks for the 2-yr period 1994--95 (ii) 30-min returns of 881 US stocks for the 2-yr period 1996--97, and (iii) 1-day returns of 422 US stocks for the 35-yr period 1962--96. We test the statistics of the eigenvalues $\lambda_i$ of C against a ``null hypothesis'' --- a random correlation matrix constructed from mutually uncorrelated time series. We find that a majority of the eigenvalues of C fall within the RMT bounds $[\lambda_-, \lambda_+]$ for the eigenvalues of random correlation matrices. We test the eigenvalues of C within the RMT bound for universal properties of random matrices and find good agreement with the results for the Gaussian orthogonal ensemble of random matrices --- implying a large degree of randomness in the measured cross-correlation coefficients. Further, we find that the distribution of eigenvector components for the eigenvectors corresponding to the eigenvalues outside the RMT bound display systematic deviations from the RMT prediction. In addition, we find that these ``deviating eigenvectors'' are stable in time. We analyze the components of the deviating eigenvectors and find that the largest eigenvalue corresponds to an influence common to all stocks. Our analysis of the remaining deviating eigenvectors shows distinct groups, whose identities correspond to conventionally-identified business sectors. Finally, we discuss applications to the construction of portfolios of stocks that have a stable ratio of risk to return.