On the Overuse and Misuse of Mediation Analysis: It May Be a Matter of Timing

Kenny (2008) credited Hyman (1955) with originally discussing mediation analysis under the name elaboration. Of importance, Hyman's elaboration required a time-ordered relationship among variables, such that the mediator must always intervene in time between the predictor and outcome. However, in the modern discussions of mediation (e.g., Baron & Kenny, 1986; Preacher & Hayes, 2004), this crucial point about time-ordered relationships appears to be underemphasized. This article shows that by employing a conceptual timing criterion for all mediation analyses, the overuse of this technique can be curbed, and, simultaneously, researchers will understand when mediation analyses are appropriate across the behavioral and medical science literatures.