A didactic example of multilevel structural equation modeling applicable to the study of organizations

Abstract

This article provides a didactic example and application of new developments in structural equation modeling (SEM) that allow for the modeling of multilevel data. Such data often arise naturally from organizational structures in which within‐group units (employees, students, etc.) are observed in larger between‐group units (firms, schools, etc.) The article begins with an overview of the basic ideas of SEM and multilevel linear regression. The synthesis of both methods developed by Muthéin (1994) is then presented in the simple case of a multilevel path analysis model, in which the variations in within‐group level intercepts are modeled as a function of between‐group variables following their own path model. An application motivated by a real problem in the field of education that focuses on validating indicators of the quality of science education in the United States follows. The results show that it is possible to statistically capture the salient complexities of organizations through the application of multilevel SEM. The article concludes with a discussion of the utility of multilevel SEM for organizational studies.