학회소식
[2005년 제 1차] A New Test of Stock Return Predictability
작성자 : 관리자
조회수 : 909
게시일 :
2005-03-04
The predictive regressions are subject to the well-known finite samples bias in the slope estimator. To adjust for the bias, a two-step estimation is proposed under the typical assumption that the predictor variable follows a first-order autoregressive process. The first step is to obtain bias-adjusted estimate of the autocorrelation for the predictor variable. The second step is to estimate the predictive regression by making use of the residuals from the first step as an additional regressor which plays a role in correcting for the bias in the slope estimator. In light of Pagan (1984), this paper also considers a correct standard error for the slope estimator of the second step regression in the presence of a generated regressor. Our Monte Carlo experiments suggest that, unlike existing tests, the proposed test has little size distortions.
