Complete the account of the three effects that any study of change over time must separate.

An age effect is what changes as people grow older, in any period. A period effect is what changes for everybody at once at a given moment. A cohort effect is what is distinctive about people born at a particular time. Because the third is exactly the second minus the first, they are linearly dependent and cannot be separated without an outside assumption.

Stating the dependence precisely is what turns a vague warning into a usable rule. Any published estimate that reports all three effects has imposed a constraint — equal effects for two adjacent cohorts, a zero linear trend, a functional form — and the credibility of the result rests entirely on whether that constraint is defensible for the case at hand.