BIAS DARI PENGGUNAAN MODEL DI MBAR

Abstract

This paper discusses biases that occurred on level, return and event study
models. In a prices lead earnings condition, the coefficient in the level model is
unbiased, while that in the return model is biased. In a prices do not lead earnings
condition, both level and earnings models yield unbiased coefficients economically.
In general, the level model suffers more serious bias econometrically than does the
return model. In an event study, daily returns with equally-weighted index are able to detect
abnormal returns better than are monthly returns with value-weighted index. When
announcements are clustered in calender events, the market model or the marketadjustement
model is less biased than the mean-adjusted model. But, when event
dates are not clustered, a simple model such as the mean-adjusted model is not worse
than other models.