Modeling the Evolution of Age and Cohort Effects in Social Research

Sam Schulhofer-Wohl, Princeton University
Yang Yang, University of Chicago

It is well known that the conventional linear age-period-cohort model suffers from an identification problem: If an outcome depends on the sum of an age effect, a period effect and a cohort effect, we cannot distinguish the effects of age, period and cohort because cohort = current year - age. Less well-appreciated is that the model suffers from a conceptual problem: It assumes that the influence of age is the same in all years, the influence of present conditions is the same at all ages and cohorts do not change over time. We argue that these assumptions fail in most applications, and we propose a model that relaxes them. Our model operationalizes Ryder's (1965) concept of cohort effects as an accumulation of age-by-period interactions. The additive model is a special case; except in special cases, our model is identified. We apply our model to analyze mortality rates in Sweden.