【主讲】Tae-Hwy Lee (Department of Economics, University of California, Riverside)
【主题】Comparing Nested Predictive Regression Models with Persistent Predictors
【时间】2015年11月27日 (周五) 15：30-17：00
【地点】上海财经大学经济学院楼701室
【语言】英文
【摘要】Inference on stock return predictability is commonly conducted by the in-sample inference on the coefficient estimator of the predictive regression, for which several problems have been identified such as the finite sample bias (when predictors are weakly stationary) and the non-pivotal and non-standard asymptotic distribution and un-correctable bias (when predictors are persistent). Various solutions to these problems have been suggested. In this paper, we adopt the out-of-sample inference of the predictive regression model by the encompassing statistic (ENC). ENC was studied by Clark and McCracken (2001) for weakly stationary predictors. This paper considers persistent predictors. While many technical problems arise for in-sample inference on the predictive regression due to persistence of predictors, the out-of-sample inference based on ENC is actually benefited from persistence of predictors. Our goal is to find situations where the ENC statistic can have the asymptotic standard normal distribution even when predictors are persistent. Monte Carlo simulation shows that the asymptotic results hold in finite samples when predictors are persistent. An application to the predictive regression of the equity premium reveals strong predictive ability of several persistent predictors.
