Calculating Betas with Daily Data: Estimation Period Effects on Prediction Error

Abstract

Minimization of sampling error is critical for studies requiring beta estimation. Previous studies utilizing monthly data found estimation periods ranging from four to nine years to be optimum. However, appropriate estimation periods, using daily data, has received limited attention. This study focused on beta estimation periods from 11 to 330 days, using daily data. Results show a significant reduction in sampling error with 60 to 70 observations. Additional gains were obtained by increasing the sample size but were not statistically significant. Although a minimal estimation period was established, a single optimal estimation period, found in some of the monthly studies, was not determined. (Gl4, D81)