Proportional structural effects of formative indicators

Formative constructs must influence two or more distinct outcome variables for meaningful tests of the formative conceptualization. Because the construct mediates the effects of its indicators, the indicators must have effects on the outcomes that are proportional to their effects on the formative construct itself. This constraint has important implications for developing and testing formative models. This study demonstrates the existence of the constraint, shows that researchers must consider proportionality as a criterion for evaluating the formative conceptualization, provides examples of indicators having different effects and interpretations depending on the outcome variables used, discusses the selection of outcomes to provide rigorous rather than trivial tests of the formative conceptualization, and contends that the formative nature of constructs cannot be justified in isolation from the consideration of outcome variables. In addition, the study demonstrates the importance of considering how the scaling of the formative construct influences the significance of the effects in the model. © 2008 Elsevier Inc. All rights reserved.