Dynamics of cross-correlations in the stock market

Co-movements of stock price fluctuations are described by the cross-correlation matrix C. The application of random matrix theory (RMT) allows to distinguish between spurious correlations in C due to measurement noise and true correlations containing economically meaningful information. By calculating cross-correlations for different time windows, we study the time dependence of eigenvectors of C, which are related to economic sectors, and the time evolution of the largest eigenvalue, which describes the average correlation strength. We use these results to forecast cross-correlations, and test the quality of our forecast by constructing investments in the stock market which expose the invested capital to a minimum level of risk only.