Earnings per Share Forecasting: Accuracy of Random Walk and Random Walk with Drift Models

The research aims to compare the forecasting power for earning per share (EPS) using two models, namely the random walk model and the random walk with drift model. The study uses historical EPS data of three companies listed in the Standard and Poor’s 500. The three companies are selected based on their ascending beta values from de fensive, neutral, and aggressive stock categories. The ascending beta values for these three companies, namely NextEra, Oracle, and FedEx, are 0.16, 1.0 and 1.3. The three companies have EPS data for 30 years between 1990 and 2019. For the respective defensive and neutral stocks of NextEra (Beta = 0.16) and Oracle (Beta = 1.0) compa nies, the results show that the random walk model has better forecasting power meas ured by both the mean absolute error (MAE) and the mean square error (MSE) fore casting accuracy dispersion measures, but with higher positive bias measure of the average error (AVE). For the aggressive stock of FedEx company (Beta = 1.3), con flicting results are found where the random walk with a drift model performs better in terms of MAE dispersion measure and the random walk model performs better in terms of MSE dispersion measure. In terms of AVE bias measure, the random walk model has a slight positive bias whereas the random walk with a drift model has a high neg ative bias. Additionally, the forecasting power for EPS gets worse as the beta of stock increases for the case of these three companies. Keywords: forecasting, earnings per share, beta, accuracy