Removing noise from correlations in multivariate stock price data

This paper examines the applicability of Random Matrix Theory to portfolio management in finance. Starting from a group of normally distributed stochastic processes with given correlations we devise an algorithm for removing noise from the estimator of correlations constructed from measured time series. We then apply this algorithm to historical time series for the Standard and Poor’s 500 index. We discuss to what extent the noise can be removed and whether the resulting underlying correlations are sufficiently accurate for portfolio management purposes.