2023

Aguirre‐Urreta, Miguel I.; Rönkkö, Mikko; Marakas, George M.

The literature on formative modelling (“formative measurement”) in the information systems discipline claims that measurement model misspecification, where a reflective model is used instead of a more appropriate formative model, is widespread. In this research, we argue that this cannot be true because models misspecified in this way would fail the measurement validation procedures used with reflective models and thus would not be publishable. To support this argument, we present two extensive simulation studies. The simulation results show that in most cases where data originates from a formative model, estimating a reflective model would not produce results that satisfy the commonly used measurement validation guidelines. Based on these results, we conclude that widespread publication of models where the direction of measurement is misspecified is unlikely in IS and other disciplines that use similar measurement validation guidelines. Moreover, building on recent discussions on modelling endogenous formatively specified latent variables, we demonstrate that the effects of misspecification are minor in models that do pass the model quality check. Our results address important issues in the literature on the consequences of measurement model misspecification and provide a starting point for new advances in this area.