A comparison of quarterly earnings per share forecasts using James-Stein and unconditional least squares parameter estimator

This study provides evidence which suggests that the use of the James-Stein shrinkage estimator of parameters from multiple ARIMA models of quarterly earnings per share results in forecasts of earnings with lower mean square percentage forecast error (MSPFE) than can be obtained using the unconditional least square estimator. Moreover, this reduction in MSPFE is available at low marginal computational cost.