Forecasting performance of grey prediction for education expenditure and school enrollment

► The study carried out an out-of-sample forecasting competition among GM(1,1), GM(1,1) rolling, and exponential smoothing. ► Annual time-series data of education expenditure and school enrollment from 1991 to 2004 documented by NCES were used to generate individual forecasts. ► Accuracy measurements were estimated based on the criteria of MAD, MAPE, MSE and the F-statistics for the test of equal forecast accuracy. ► The GM(1,1) rolling model over-performed the other two models in predicting education expenditure and student enrollment.