Using component securities of the Dow Jones Industry Index from 1998 to 2003 as the samples. This paper uses a grey forecasting model GM(1, 1) on improving the estimation of systematic risk of the classical capital asset pricing model. Due to a lot of estimation problems, like unstable estimation, the author tries to establish a more stable and correct forecasting model by using the grey forecasting model GM(1, 1). The author uses the Theil's U as a tool on the examination of each forecasting model, estimating systematic risk coefficient β. The results show that the Theil's U of the GM(1, 1) is only 19.0138%, and declines 39.8599% estimation error than classical Moving Average.