The forward unbiasedness regression is revisited by varying the prediction horizons from 1 day to 1 year. The panel data suggests some possibility of a positive slope coefficient at a short horizon while the negative coefficient improves forecasting performance at longer horizons