Portfolio performance under reference-day risk

Abstract

Reference-day risk has been previously identified as a type of sampling variation phenomenon, and its effect on the estimation of stock returns and their volatility and market betas has been documented. Using a dataset of daily equity mutual fund returns, we extend previous studies to analyse the effect of reference-day risk on regression alphas, a metric that is widely used to measure fund performance. The results follow the trend observed for other financial asset statistics: for a large proportion of funds that show significant alphas, significance depends only on the reference day used as a basis for the estimation of the returns. This finding casts doubt on all inferences previously obtained using this measurement. Further tests show that the variability of the alpha and its significance does not appear to have a systematic pattern and is not related to fund characteristics. Methodologies tested that may diminish the problem, which include the use of factor models and robust regression using Student-t distributed errors, do not yield encouraging results.