On the Nature and Estimation of Age. Period, and Cohort Effects in Demographic Data

This paper develops a general procedure for estimating age, period, and cohort effects in demographic data. The procedure involves structuring, mathematically, the effect of cross-cohort changes in the timing and level of a vital event on period rates of occurrence of the event. The procedureis illustrated and tested in an application to data on the first birth rates of American women. Overall, the empirical results provide support for the procedure. The results also provide evidence that period effects are highly age-specific and that the size of cohort effects may be substantially overestimated by models which fail to allow for the age specificity of period effects.