Can a Variable be both a Mediator and

Can a Variable be both a Mediator and a Moderator? Dominique Muller (University of Grenoble/University Institute of France) Charles M. Judd (University of Colorado) Vincent Yzerbyt (Catholic University of Louvain)

Mediation (Baron & Kenny, 1986; Judd & Kenny, 1981) § X is a treatment and units are randomly assigned to conditions § Me is a mediator measured after X (and before Y) § Y is the dependent variable 2

Mediation Me X Y 3

Mediation Me X Y Without a causal argument, these are nothing else than regression equations, not a strong case for mediation Social psychologists have been very concerned with causality in mediation models (e. g. , Spencer, Zanna, & Fong, 2005) 4

Mediation in the Mac. Arthur approach The Mac. Arthur approach (see Kraemer, Kiernan, Essex, & Kupfer, 2008) takes its root in clinical interventions and randomized clinical trials (RCT) c, a, and b conjointly significant or c, a, and d conjointly significant Mediation 5

Moderation in social psychology P S Y 6

Moderation in social psychology S P Y 7

Moderation in social psychology P S S Y P Y Social psychologists are fine with both models Only preferences dictate which one will be used Moderation is equated with simply a statistical interaction 8

Moderation in the Mac Arthur approach § X is a treatment and units are randomly assigned to conditions § Mo, the moderator, is measured PRIOR to the treatment § And therefore, the Treatment and the moderator are uncorrelated Moderation is not the same as simply a statistical interaction 9

Definition of moderation In social psychology: A moderator P is a variable that affects a S – Y relationship or a moderator S is a variable that affects a P – Y relationship In Mac Arthur Approach: A moderator is a variable that affects a causal effect of X on Y 10

Can a variable be both a mediator and a moderator? From a social psychology perspective: -> Me is a mediator (c, a, and b are conjointly significant) -> Me is a moderator (the d interaction is significant) For the Mac. Arthur approach: -> Me is only a mediator (c, a, and d are conjointly significant) 11

For social psychologists Me X Y 12

For social psychologists Me X Y This explains why it has been argued that this is a case of moderated mediation (Preacher, Rucker, & Hayes, 2007) 13

For the Mac. Arthur approach XMe Me X Y 14

Can a variable be both a mediator and a moderator? Social psychology perspective Mac. Arthur Approach Me X Me Y X Y Y 15

Conclusion Can a variable be both a mediator and a moderator? Yes, it can be both if moderation and statistical interactions are equated No, if moderation and statistical interactions are not equated How should we define moderation? 16