Abstract

This research is aimed at testing the robustness of the positive correlation between average stock risk and stock market returns documented by Goyal and Santa-Clara (2003). This is addressed by investigating this relationship in the Canadian market by using an improved market timing methodology. We examine both the statistical and financial properties of average stock risk, as measured for equal- and value-weighted portfolios using a daily sample of all TSE-listed Canadian firms between January 1975 and December 2001. In order to further test the robustness of the relationship, we examine two sub-samples, January 1980 to December 1989 and January 1990 to December 1999. We find that the positive trend in average stock risk previously noted by Campbell et al (2001) and by Goyal and Santa-Clara (2003) is only present over our entire time period and not over each of the sub-periods. This questions the generality of the result for the entire period. Further, our findings do not support the positive correlation between average stock risk and stock market return. Our analysis shows only a weak statistical link and practically no financial timing ability when this time-series is used in an active market timing investment strategy. When macro-economic variables are introduced into our forecasting models, the in-sample statistical properties of the risk-return relationship increases significantly at the expense of the out-of-sample forecasting accuracy.