| 摘要: | The size and book-to-market (BM) 「anomalies」 identified by Fama and French (1992) have been examined extensively worldwide over the past decade. Rationality supporters such as Fama and French (1993) viewed both characteristics as missing factors, and advocate alternative multifactor model specifications to 「replace」 the single-factor CAPM. Behavioral theorists, on the other hand, view such characteristics as true anomalies resulting from investors』 behavioral biases such as over- and/or under-reaction. Commonly such empirical studies are conducted using Fama and MacBeth』s (1973) twopass cross-sectional regression (CSR) method, which is known to suffer from an errors-in-variables (EIV) problem. Although 「financial econometricians」 have investigated the two-pass methodology, and proposed alternative estimators, most of the solutions are too technically complicated, and none of them (to my knowledge) has considered the possibility that size and BM effect may in fact be consistent with the CAPM. In this project, I propose two ways to deal with the problem. First, I propose an EIV-corrected estimator for the CSR model based on individual securities, which is consistent, and is easy to implement. Second, as size and BM are highly correlated to beta estimates, it is natural to use size and BM as instrumental variables. Hence, I propose using GMM to re-examine the market risk premium, and using an overidentifying restriction test to examine if both characteristics serve as good instruments. In both of the above two methods, the single-beta CAPM is the null hypothesis. Recently, Schwert (2002) points out that anomalies such as size effect, the value effect, the weekend effect and the dividend yield effect seem to have weakened or simply disappeared after the papers that highlighted them were published. Schwert (2002) asserts that practitioners that began investment vehicles that implemented the strategies implied by some of these academic papers cause the anomalies to disappear. Thus, Schwert suggests that research findings cause the market to become more efficient. Indeed, if the documented anomalies were of any economic significance, one would expect the anomalies to disappear as time goes by when the market participants perceive the profitable opportunities embedded in the anomalies or when a more appropriate pricing model is developed or proposed. On the contrary, if the premia persist, it may imply that it is either rational, or just of no economic significance. Schwert (2002) only uses dummy variable to denote different time periods, which treats potential structural change as exogenous. A priori structural breaks should be treated as endogenous, and may not necessarily occur only once or at known time. Thus, it is of interest to examine statistically if the size and BM premia exhibit structural breaks, and especially if the time of structural shifts is consistent with the publication of the related articles. A natural tool to examine this issue is Bai and Perron』s (1998) change-point model. Hence, this project shall contribute to the literature mainly from the econometric perspective. In the first year, I will examine the significance of beta based on the simple EIV-corrected estimator. The second-year project will examine the validity of beta using size and BM as instrumental variables. Finally, in the third year, I will use change-point threshold model to identify possible patterns of changes in the size and value premia. 研究期間：9508 ~ 9607 |
