Forecasted Sharpe Ratio Ranking Strategy

Benjamin Callesen & Christian Buhl Møller

Student thesis: Master thesis


The purpose of this thesis is to understand if a portfolio constructed on the basis of ranking stocks by their predicted Sharpe ratio can produce a mean excess returns, a mean volatility, or a mean Sharpe ratio that is superior and different from an equally weighted portfolio of the same stock universe with statistical significance. The relevant stock universe is the current constituents of the Dow Jones Industrial Average, excluding 3 of the stocks, resulting in sample data consisting of 27 stocks over the period 28-03-1991 to 26- 03-2018. The study uses four different implementations of linear regressions to predict the Sharpe ratio as the dependent variable and the lagged 12-month forward, the lagged 12-month trailing earnings per share, the lagged 3-month excess return, the lagged 6-month excess return, the lagged 9-month excess return, and the lagged 12-month excess returns as independent variables for all 27 stocks. The portfolio findings are based on holding the 20 stocks with the highest predicted Sharpe ratio with a holding period and rebalancing frequency of 3-months. The portfolios are tested against each other and the equal weight portfolio to determine both the internal hierarchy of the model portfolios and the performance relative to the simpler portfolio. The models are found to have some differentiating traits regarding their ability to minimize squared errors and maximize the number of significant Sharpe ratio estimates in relation to the actual observed Sharpe ratios of the different stocks. However, the linear relations between the dependent variables and the independent variables are found to be weak, which means the predicted Sharpe ratio is most often not statistically significant, and as a result, the wrong stocks are excluded or included based on the Sharpe ratio criteria. Consequently, the result is that the mean excess return, mean volatility, and mean Sharpe ratio produced by the four model portfolios, are not found to be different from the equal weight portfolio with statistical significance. Although it is shown that if the ranking procedure is based on ex post Sharpe ratios, a portfolio with statistically significant and superior excess return and Sharpe ratio can be constructed.

EducationsMSc in Finance and Accounting, (Graduate Programme) Final Thesis
Publication date2018
Number of pages124
SupervisorsSøren Plesner