Title: Long-run predictability revisited
Authors: Paulo Rodrigues - Universidade Nova de Lisboa (Portugal) [presenting]
Matei Demetrescu - University of Kiel (Germany)
Robert Taylor - University of Essex (United Kingdom)
Abstract: Long-run predictability is revisited and a simple yet powerful new approach is proposed. The procedure is based on the bias reduced IVX framework recently proposed which extends previous contributions. The new approach has several advantages over existing procedures designed to test for long-run predictability. The first is its simplicity of application, which makes it very appealing for empirical applications, when compared, for instance, with Bonferroni based methods, as these require the computations of confidence intervals for the near-integrated parameter $c$ which characterises the persistence of the predictor; the second advantage is that left, right and two sided hypotheses tests can be immediately computed from our approach, whereas Bonferroni based methods require the computations of different confidence intervals for $c$ depending on whether left or right tailed intervals are to be computed; and third our approach is easily generalized to a multi-predictor context.