A recent research paper authored by Ekaterini Panopoulou of Kent Business School (joint with L. Meligkotsidou, I. Vrontos and S. Vrontos) entitled ‘Out-of-sample equity premium prediction: A complete subset quantile regression approach’ has ranked among the TOP 10 most downloaded papers in its category within days of it being uploaded onto the Social Science Research Network (SSRN).
The paper focuses on the issue of forecasting equity returns, which is one of the most widely discussed topics in the finance literature mainly due to its central role in asset pricing, portfolio allocation and evaluation of investment managers. The authors propose a new forecasting approach based on complete subset quantile regressions. Their quantile regression approach to equity premium prediction allows them to cope with the non-linearity and non-normality patterns that are evident in the relationship between stock returns and potential predictors. Moreover, by employing quantile forecast combinations, they reduce model uncertainty and parameter instability. Finally, employing complete subset quantile regressions induces shrinkage to the respective estimates and further helps reduce the effect of parameter estimation error. A further contribution of this paper is the development of a recursive algorithm for selecting the best subset in real time, based on the past history of excess returns and predictive variables. The proposed algorithm is a likelihood-based method that chooses the best complete subset for a given quantile and is flexible enough to allow for variability of the selected value across quantiles. Their approach delivers statistically and economically significant out-of-sample US equity premium forecasts relative to both the historical average benchmark and the complete subset mean regression approach. More importantly, the economic evaluation results suggest that an investor that adopts this framework can gain sizable benefits which range from 3.91% to an impressive 6.27% per year relative to a naïve strategy based on the historical benchmark performance.
For more details, the paper is available from the following link: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2335084