Reference-Day Risk and the Use of Monthly Returns Data

Investment practitioners and the empirical finance literature make extensive use of monthly stock returns, where a monthly return is based on the change in stock price between one particular day of the calendar month—the reference day—and the corresponding day of the following month. We show that the choice of reference day seriously affects estimates of the properties of monthly returns, including their means, medians, variances, correlations, and betas. We find these effects both in individual stocks and in market indexes. Our evidence indicates the effects are generally unsystematic and are caused by sampling variation but are sufficiently pervasive and serious to suggest that studies that use estimates of the properties of monthly returns as inputs, and portfolio decisions based on such estimates, should be tested for robustness against different reference days.