Mend It or End It

Abstract

Moderator variables are widely hypothesized and studied in the organizational sciences, but the empirical track record of moderator variable studies is very discouraging. These studies often lack sufficient statistical power and the type of designs and measures common in organizational research virtually guarantee that the moderator effects that are found are usually extremely small. We recommend that future attempts to identify and estimate moderator effects should be limited to situations where better measures, stronger research designs and a realistic cost-benefit assessment are available. Researchers should avoid moderator hypotheses in contexts where the measures and research designs employed do not allow them to be tested in a meaningful way, and should be cautious about interpreting the very small effects they are likely to find.