Stochastic Properties of Time‐Averaged Financial Data: Explanation and Empirical Demonstration Using Monthly Stock Prices

This article considers the potential statistical problems resulting from the use of averaged rather than end-of-period data in financial research. Averaged data are widely employed throughout the literature without explicit recognition that the use of such data results in biased estimates of the variance, covariance and autocorrelation of the first as well as higher order changes. We illustrate the magnitude of the biases, using the S&P 500 end-of-month series over the period March 1957 to February 2001. Results confirm the predictions of Working and of Schwert. In addition, an analysis of the properties of higher-order lags indicates that the bias persists, a result not previously suggested in the literature. We also find that these statistical biases are time varying--which has significant implications for empirical financial research. Copyright 2001 by MIT Press.