摘要：Using monthly data from 1975-2001, we consider the stability of bivariate and multivariate models for short run in-sample predictability of Canadian stock returns. We test for model stability using a range of tests including the Andrews SupF statistic, Bai subsample procedure, and Bai and Perron sequential SupF procedure. We find evidence of instability in two of our nine bivariate cases considered as well as our preferred multivariate model. When estimated to account for these breaks, we find the degree and direction of predictability can change markedly
