Having learned how to evaluate reflective measurement models in the previous chapter (Stage 5a of applying PLS-SEM), our attention now turns to the assessment of formative measurement models
(Stage 5b of applying PLS-SEM). The internal consistency perspective that underlies reflective measurement model evaluation cannot be applied to formative models since formative measures do not
necessarily covary. Thus, any attempt to purify formative indicators based on correlation patterns can have negative consequences for a construct measure’s content validity. This notion
especially holds for PLS-SEM, which assumes that the formative indicators (more precisely, composite indicators) fully capture the content domain of the construct under consideration. Therefore,
instead of employing measures such as composite reliability or average variance extracted (AVE), researchers should rely on other criteria to assess the quality of formative measurement
The chapter begins with an introduction to the criteria needed to evaluate formative measures. This includes a discussion of the bootstrapping routine that facilitates significance testing of PLS-SEM estimates, including formative indicator weights. These criteria are then applied to the corporate reputation model that is extended for this purpose. While the simple model contains only three reflectively measured constructs as well as one single-item construct, the extended model also includes four antecedent constructs of corporate reputation that are measured using formative indicators. This chapter concludes with the evaluation of measurement models. In Chapter 6, we move to the evaluation of the structural model (Stage 6 of applying PLS-SEM).
Copyright © 2021 A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM)