Reflections on SEM

Abstract

For almost 40 years structural equation modeling (SEM) has been the statistical tool of choice for the assessing measurement and structural relationships in the social sciences. During the initial 30 years almost all applications of SEM utilized what has become known as covariance-based SEM. But in the past ten years an alternative structural equation modeling method, composite-based SEM, has increasingly been applied. In fact, a substantial number of social sciences scholars consider composite-based SEM the method of choice for structural equation modeling applications. In this paper, I provide an overview of the evolution of SEM, from the early years when factor-based SEM was the dominant method to the more recent years as composite-based methods have become much more prevalent. I also summarize several relevant composite-based topics including the emergence of composite-based SEM, confirmatory composite analysis (CCA), and a new method of generalized structured component analysis (GSCA). In the final section I propose some observations about current developments and future opportunities for composite-based SEM methods.