An alternative to steady-state comparisons in perfect foresight models