Research Impacts and Correction of the Upward Biased CRSP Daily Equal-Weighted Index

Abstract

Loughran and Ritter (1995), and Canina et al. (1998) show that compounding the CRSP daily Equal-Weighted Return with Dividend (EWRETD) to obtain a monthly index will generate a substantial bias. To overcome the influences of bid-ask bounce, Blume and Stambaugh (1983), non-synchronous trading, Roll (1983), we design a better method to generate a monthly EWRETD index from the CRSP daily data. Using our method, the average annual bias between the generated monthly time series and the corresponding values drawn from the CRSP monthly tape is 0.02%, considerably smaller than 6% reported in Canina et al. (1998). In the paper, we explain the reasons why the conventional way to construct a daily equal-weighted portfolio lacks the ability to filter out the undesired impacts of the microstructure and is inconsistent with corresponding monthly data. More importantly, we offer a new method to generate an unbiased CRSP daily EWRETD, which is free of the problems associated with the microstructure and consistent with the CRSP monthly index. Benchmarked on this unbiased daily time series, we study the impacts of the upward biased CRSP daily EWRETD on the estimations of beta, Jensen's alpha and event study.