Forecasting monthly earnings per share-Time series models
Date of Original Version
The purpose of this paper is to compare the accuracy of five forecasting models for monthly earnings per share data, a seasonal time series. The five models are Holt-Winters exponential smoothing model (HW), Box-Jenkins ARIMA Model (BJ), linear regression of data deseasonalized by the Census II-X11 method (X11), linear regression of data deseasonalized by the X11-ARIMA method (X11ARIMA), linear regression of data deseasonalized by the ratio-to-moving-average method. Study of earnings per share data is important because (1) these data exhibit all the systematic components of a time series, (2) earnings per share forecasts are important for purposes of financial decision making and strategic planning, and (3) previous studies of earnings per share data did not compare these five forecasting models. © 1989.
Jarrett, J.. "Forecasting monthly earnings per share-Time series models." Omega 17, 1 (1989): 37-44. doi:10.1016/0305-0483(89)90018-2.