Not Merely a Matter of Drawing Arrows: The Empirical Consequences of Measurement Model Specification

Abstract

As the use of multiple-item measures become more prevalent in management and social science research, researchers begin to explore the efficiency and accuracy of the different types of measurement models (i.e., formative, reflective, and MIMIC). Perhaps one of the most important implications from existing research is that measurement specification is not merely a matter of how arrows should be drawn between a construct and its measures – it has empirical consequences (Bagozzi, 2007; Howell, Breivik, & Wilcox, 2007; Jarvis, MacKenzie, & Podsakoff, 2003; Law & Wong, 1999; MacKenzie, Podsakoff, & Javis, 2005). What these empirical consequences are and how substantial they could get are issues which need to be examined more thoroughly. The current study addresses these issues with simulation. Results show that a correctly specified model has significant impact on the estimation of substantive relationships, model fit indices, and factor loadings to various extent. More importantly, we identify items intercorr...