A Study of the Bias effect between Firm size and BE/ME
Eunkwang Lee
Namwon Hyung
Banz and Breen(1986) shows that Firm size and Earning yield use a SURE about 2,500 security in U.S stock market that includes an ex post selection bias and look ahead bias. As to SURE is mainly limited to error in variables problem only ¥â and portfolio must be solve it. we are recognized to capture to reason for existence of look ahead bias that compares in empirical test by using FF 3 factor model when dating problem is induce or not. First, look ahead bias doesn't impact to estimate direction and statistical significance but estimate are consistently getting more under-estimate when dating gap be greatly in multivariate test and we found to aspect in time series model that occurs a small error in risk premium within any group. supposed that it cause from skewness bias by the long-term abnormal return about some reasons including a positive(e.g merger or dividend) or negative(e.g bankruptcy, liquidity) that have a affect bigger to returns than before when dating gap be risen that is not tightly but consistently influenced on empirical tests. Secondly, variables of ¥â including ME and B/M are also well enough to capture on risk or variation on common equity through time in our korea stock market. and mostly it checked consistently in empirical test which given that estimates of its ¥â for the preceding in 5 years and other estimates of the beta for a given period of time, 2, 3, and 4 years, that did not differences on multivariate test. In the Fama and French(1992) show that FM regression took consistently B/M proxy variable and Fama and French(1993) by using a model is good at explanations of time-variant and other factors captures the risks when ¥â does not capture. In this study we check it perfectly pattern in our market, and also Merton(1973) study that it fitted closely to 0 but when B/M became larger, risk premium also revealed the increasing pattern shouldn't satisfied. and we don't found of case that it cannot explain the average rate of premium as SMB and HML when risk premium was big. The FF 3 factor model was sensitively changed how to consider about data properties. and look ahead bias is manifested by skewness bias and researcher misconception, and we can see how bias effect has affect on cross sectional and time series in this study.