PLS-SEM and reflective constructs: A response to recent criticism and a constructive path forward

Abstract

This article addresses criticisms asserting that reflective construct measurement and its associated evaluation criteria are unsuitable for partial least squares structural equation modeling (PLS-SEM). More specifically, critics contend that reflective measurement models correspond exclusively to common factor models, a premise that is both inaccurate and misleading. Reflective measurement models represent theoretically grounded and conceptualized constructs. Statistical methods such as common factor model estimation, composite model estimation, and sum score regression enable researchers to estimate method-specific proxies that serve as approximations for theoretically established conceptual constructs in empirical research. These proxies vary depending on the statistical models and assumptions inherent to each method. In this context, it is important to highlight that the use of reflective evaluation criteria is not restricted to common factor models. When applied to composite model estimation, it does not compromise the validity of the results. Moreover, this article advocates for embracing the complementary strengths of diverse SEM methods within a multimethod approach, rather than positioning one method in opposition to another. We believe that this contribution provides critical insights and guidance, fostering advancements in SEM methodology, and its practical applications.