Three approaches to estimate latent interaction effects: intention and perceived behavioral control in the theory of planned behavior

Abstract

Interaction effects between explanatory constructs are an important part of many social theories. Analyses of interaction effects between variables using regression techniques have low power because they do not control for measurement errors. Therefore, latent interaction modeling using structural equation modeling (SEM) has been proposed as a better alternative to test for interaction effects. In contrast to traditional and complicated ‘constrained’ SEM approaches, two recent developments, the unconstrained approach and the residual centering approach, are especially attractive for applied researchers as they are much easier to implement. However, applied researchers still seem to be unsure about how to apply these approaches. In this study, we illustrate the use of the unconstrained and the residual centering approach and compare these approaches with the constrained approach of Algina and Moulder (2001) using data from a field study of 1,442 students. Theoretical background is the theory of planned behavior (Ajzen, 1991) in which we test the proposed interaction between an individual’s intention to perform a behavior and perceived behavioral control (PBC)
on behavior. The illustration should assist researchers interested in testing interaction effects using structural equation modeling.