The Use and Misuse of Structural Equation Modeling (SEM) in Management Research: A Review and Critique

Abstract

The research practice in management research is dominantly based on structural equation modeling, but almost exclusively, and often misguidedly, on covariance-based SEM. We adumbrate theoretical foundations and guidance for the two SEM streams: covariance-based, also known as LISREL, covariance structure analysis, latent variable analysis, etc.; and variance-based SEM, also known as a component-based SEM, PLS, etc. Our conceptual framework discusses the two streams by analysis of theory, measurement model specification, sample and goodness-of-fit. We question the usefulness of Cronbach’s alpha research paradigm and discuss alternatives that are well-established in social science, but not well-known in the management research community. We conclude with discussion of some open questions in management research practice that remain under-investigated and unutilized.

Keywords: structural equation modeling, covariance- and variance-based SEM, formative and reflective indicators, LISREL, PLS

JEL Classification: C18, C3, M0