Optimal, rational forecasts are often biased and thus the empirical finding that actual forecasts are biased is not necessarily evidence of irrational behavior. Using an asymmetric loss function and a regression forecasting problem, these points are explicitly demonstrated.
