| Summary: | 碩士 === 義守大學 === 財務金融學系碩士班 === 99 === The Capital Asset Pricing Model (CAPM) says that the expected returns of securities and its systematic risk (β value) is a linear relationship. β values as a measure of systematic risk concept has been accepted and receive academic and financial practices broader application, but for the calculation of the value of β was also constantly challenged its accuracy.
Generally it use market model by ordinary least squares to estimated of β value. But Scholes and Williams (1977) pointed out that when daily data for the individual companies are not prone to frequent trading conditions, resulting in non-synchronous trading phenomenon caused by β estimates underestimated, which is called the estimate of β the intervalling bias effect (The Intervalling-Effect Bias in Beta). Therefore, some scholars hope to make some amendments to correct intervalling estimation method effect. (Scholes and Williams, 1977, Dimson, 1979)
In this study, the Taiwan Stock Exchange publicly traded common stock for the study of the subject, taking a total of 585 companies, the study period from December 29, 2003 to January 6, 2009. Focuses, the phenomenon of the existence of intervalling effects on the Taiwan market? And scholars have proposed amendments to the estimation method is valid? In the application of statistical methods using repeated-measures analysis of variance model to analyze flow rates and test periods, as well as turnover and Amendment Act estimates for β have explanatory power. Empirical results using the OLS method to monthly data derived estimates of β > weekly data estimates of β > datly data estimates of β, there are intervalling effects support for Taiwan''s securities market. S-W method, Dimson method and estimated by the OLS method the β values statistically significant difference, and the Dimson method of correction effect is better than the S-W method. But did not change after correction on the original monthly β > weekly β > datly β on the order, did not improve the intervalling effect, so that the correction is no real effect.
|
|---|
