This paper compares two methods of analyzing aggregate data that is classified by period and age. Because there is a linear relationship among age, period, and cohort, it is not possible to distinguish the separate effects without employing an identifying assumption. The first method, which is applied in the economics literature, assumes that period effects are orthogonal to a linear time trend. The second method, which is applied in the statistics literature, assumes that the effect parameters change gradually. Simulation results suggest that the performances of both methods are comparable. The results of applying the second method to household saving rates suggest that period effects had a negligible influence in the United States but considerable influence in Japan.
