This paper written by Phillip Gray provides very insightful analysis into the predictability of Australian equities. In the theoretical literature that exists out there, there have been attempts to predict the returns on equity employing diverse methodologies with varying degrees of success. Some of these employed the use of R-squared and t-statistics used with predictive regressions. This was problematic as statistical indicators can not conclusively point to the economic importance of some phenomena. More in tune with the economic perspective was the attempt to use a strategy to see if it could beat the classic buy and hold returns. When this predictive analysis points to a positive market, investments are switched to fixed income. There is little literature present regarding the predictability of equity returns in Australia.
Attempts have been made to predict using economic and financial variables and encompassing dividend yields and term spreads but to little avail. The dividend yield is seen as a good indicator of stock returns along with the book to market and price-earnings ratios. Anyhow, the literature review here is a good step towards aiding data snooping efforts, particularly in exploring some variables such as the Westpac indices which have not been employed before.
Having some basis for the theory that is present and the one employed here, the data used and methodology assume considerable importance to assess the impact of any study. The study here uses sources such as the value-weighted Australian All Ordinaries Index, the Reserve Bank of Australia for interest rates, and valuation ratios from MSCI from 1974 to June 2007. This indicates easy availability and hence verification through the use of these variables subsequently. The investigation of the economic significance of return predictability is done using a dynamic portfolio-switching strategy. Although previous research has employed the OLS method, this research favors the probit model, deeming it favorable for a strategy that switches funds between stocks and fixed income.
The predictability is done based on seven variables including dividend yield, price-earnings and book ratios, coincident index, leading index, short rate, and the term spread. It should be taken into note that two series of market returns are employed which helps the scope of this study. Secondly, arbitrary choice of the probability-based threshold embedded in the switching strategy is avoided which is good as it shields the study from data snooping criticism. An examination of the buy and hold strategy and the portfolio switching strategy analyzed in the research shows that the switching strategy produces an excess return of 0.6229 percent per month, nearly 9 basis points higher than the market portfolio. Annualized, this advantage goes even higher in terms of basis points. The probit predictive model hence supports the switching strategy but in this regard, the t-test and median rank-sum tests should not be ignored as they help avoid distortion. A statistical result such as this could be attributed to chance; however, the study does incorporate the means and standard deviations of returns to analyze whether it arises by chance and supplements this by appraising the economic significance of the results. This is followed by the incorporation of transaction costs and sub-period results to check for the robustness of the results.
The conclusion through the statistical and economic metrics points to the predictive ability of the methodology of this paper. A $A1 investment in the portfolio-switching strategy in January 1980 grows to over $A55 by June 2007. Over the same period, a $A1 passive investment in the market portfolio grows to just $A39. Bootstrapping simulations strongly suggest that this difference cannot be attributed to chance. A sampling bias could be considered as one possible drawback to the conclusions. As the sub-period analysis shows, most of the high returns are attributable to a few broken-down periods and an all-pervasive conclusion about superior predictive power is not possible based on this paper. The robustness checks would have to be better passed for that to happen. Otherwise, this paper provides an intriguing insight into how varying methods can provide various signals as to the predictability of equity returns.
Works Cited
Gray, Philip. “Economic significance of predictability in Australian equities.” Accounting and Finance 482008 783-805. 2009