AbstractIt is not readily apparent that the single‐period mean and variance are sufficient to specify the investor choice problem when the investor horizon is greater than one period. It is pointed out that the standard justifications for use of single‐period mean‐variance analysis are inconsistent with the assumptions usually made in the multiperiod problem. It is argued, however, that if the investor horizon is long enough, the single‐period mean and variance will be approximately sufficient. The argument is much less restrictive than the standard justifications for the use of single‐period mean‐variance analysis. The predictions of this alternative justification are empirically tested and found to be quite accurate for the time frames found in the