Thinking Clearly About Age, Period, and Cohort Effects with Bounding Analyses

Psychological researchers are interested in how things change over time and routinely make claims about age effects (e.g., personality maturation) or cohort effects (e.g., generational differences in narcissism) and sometimes period effects (e.g., the effects of the pandemic years 2020 to 2022 on well-being). The age-period-cohort identification problem means that these claims are not possible based on the data alone: Any possible temporal pattern can be explained by an infinite number of combinations of age, period, and cohort effects. This concern holds regardless of the study design—it also applies to longitudinal designs covering multiple cohorts—and regardless of the number of observations available—it also applies if we observe the whole population. Researchers rely on statistical models that impose assumptions to pick one specific combination of effects. But these assumptions are often opaque and researchers may be unaware of them, resulting in a lack of scrutiny. Here, I provide a non-technical introduction to a framework which allows researchers to reason about age, period, and cohort effects in a more systematic manner. I summarize advances in our understanding of the precise nature of the identification problem—it only affects the underlying linear effects but not the nonlinearities, making some statements about effects possible based on the data alone—and show how transparent assumptions can be used to put bounds on the effects of interest, following a framework developed by the sociologists Ethan Fosse and Christopher Winship. Lastly, using data from the German General Social Survey, I illustrate this approach by analyzing age, period and cohort effect on people’s agreement with the statement “Given what the future looks like, it is barely responsible to bring children into the world.”