Do Habits Moderate the Intention-Behavior Link? Statistical Analyses of the Interaction Effect Are Inadequate for Testing the Underlying Cognitive Mechanism

The habit-intention interaction effect tested in behavior prediction studies is a prominent example of statistical moderation analysis in health psychology. Motivated by the mixed empirical findings concerning the effect and long-standing criticisms on estimating and interpreting interaction effects, I used a computational modeling approach to evaluate whether the habit-intention interaction effect estimated from statistical analysis is adequate for inferring the underlying cognitive mechanism assumed by contemporary habit theories. By manipulating the presence or absence of a habit-moderating mechanism in the data-generating model, statistical analyses were performed on the simulated data to examine the correspondence of the estimated interaction effect to the manipulation. Results suggest that while the conventional statistical approach can be indicative of the underlying mechanism, the direction and size of the estimated effect are often misleading in relation to the substantive theory. I identified the cause of the problem from a causal perspective (i.e., intention as a common cause of habit strength and behavior) and discussed potential remedies.