Using Past Prices and Earnings to Derive Abnormal Returns over a Stock Index

Abstract

Our objective is to determine whether one can derive returns in excess of a chosen benchmark by using readily available information such as past prices and earnings. A key aspect of our method is that we test the estimation results in conjunction with the portfolio optimization process that incorporates those same estimates, as they are generated, into a rationally-diversified portfolio. We rely on a sampling process that randomly pairs companies from a pool of available estimates for any given date, coupled with the Black–Litterman optimization algorithm, in order to derive a distribution of average returns for the 2006–2014 period, using data available for companies listed at Bucharest stock Exchange. We find that even when using information such as earnings and past prices, one can still improve the performance of a given benchmark, both on an absolute and risk-adjusted metrics. We show how the variability of the results coming from the calibration of the Black–Litterman model itself can easily be mitigated by carefully selecting the model’s parameters.