Economic significance of predictability in Australian equities

摘要

This paper examines evidence of predictability in Australian equities using both statistical and economic metrics of significance. A probit-based predictive model is used to forecast the probability that the 1 month ahead excess market return will be positive. Funds under management are then switched between equities and fixed income on the basis of this forecast. Although the statistical evidence of the model's predictive ability is mixed, the results suggest convincing evidence of an economically significant degree of return predictability. A $A1 investment in the switching strategy (market) in January 1980 grows to over $A55 ($A39) by June 2007. Although the economic significance of the switching strategy remains even in the presence of high transaction costs, robustness checks suggest that the seemingly impressive full-sample results might be sample specific. The apparent superiority of the portfolio-switching strategy can be traced to a handful of observations early in the study during which the predictive model provides a timely signal to exit equities. There is little evidence that the predictive model has forecasting ability across the entire sample. As such, this paper serves both to illustrate how alternate metrics of return predictability can lead to divergent conclusions, and to emphasize the importance of subjecting apparent findings of predictability to robustness checks.