On the Limit Distribution of the t-statistics in Bias-reduced Predictive Regressions

Resumo/Abstract:This paper focuses on the asymptotic and finite sample evaluation of reduced bias based predictability test procedures introduced by Amihud and Hurvich (2004). In particular, we investigate the behaviour of these tests when the predictor(s) is (are) stationary or (near) integrated. Surprisingly, we find through simulations that the test by Amihud and Hurvich (2004) presents some desirable features such as, a degree of robustness in the context of persistent predictor(s), a behaviour which is contrary to other tests available in the literature and to what could have been expected. This is an empirically relevant result in that in contrast with most available procedures the test is applicable regardless of the stochastic nature of the predictors and allows for more than one predictor in the test regression whereas most available procedures only consider a single predictor in the regression equation.

Outros seminários / Other seminars: Programa completo / Full programme.