Everything you always wanted to know about SEM (structural equations modeling) but were afraid to ask

This article is intended to serve as a primer for structural equations models for the behavioral researcher. The technique is not mysterious—it is a natural extension of factor analysis and regression. The measurement part of a structural equations model is essentially a confirmatory factor analysis, and the structural part of the model is like a regression but vastly more flexible in the types of theoretical models that may be tested. The models and notation are introduced and the syntax is provided to replicate the analyses in the paper. Part II of this article will appear in the next issue of the Journal of Consumer Psychology, and it covers advanced issues, including fit indices, sample size, moderators, longitudinal data, mediation, and so forth. © 2009 Society for Consumer Psychology. Published by Elsevier Inc. All rights reserved. Structural equations modeling (SEM) is an important tool for marketing researchers. Our top journals contain numerous articles with SEM analyses of data as varied as B2B relationship marketing studies (e.g., Anderson and Narus, 1990; Selnes and Sallis, 2003; Steenkamp and Baumgartner, 2000) or B2C customer satisfaction surveys (e.g., Luo and Bhattacharya, 2006). Survey data from studies such as these are a natural fit for SEM, but when SEM is considered as a natural extension of regression (to be demonstrated shortly), it should be clear that SEM is a statistical tool, orthogonal to the substantive domain of data on which it is implemented. Thus, SEM may also be applied to secondary data bases (e.g., behavioral CRM or loyalty program data) or data collected in experimental settings (e.g., measures of memory, attitudes, and intentions), etc. For example, it is not unusual for an experimental consumer psychologist to inquire about mediation in an attempt to clarify the theoretical processes by which some phenomenon occurs (e.g., Iacobucci, 2008). If mediation clarifies the conceptual picture somewhat, with the insertion of just one new construct— the mediator—imagine howmuch richer the theorizing might be if researchers tried to formulate and test even more complex nomological networks. The Journal of Consumer Psychology presents this article to encourage more frequent and knowledgeable use of SEMs. We begin with an overview of the SEM approach: (1) we discuss the E-mail address: [email protected]. 1057-7408/$ see front matter © 2009 Society for Consumer Psychology. Publish doi:10.1016/j.jcps.2009.09.002 motivation underlying SEMs and a perennial cautionary concern regarding the interpretation of causality, (2) we review the source models of factor analysis and path analysis, and (3) we combine these to develop the full SEM model, and describe the entirety of parameters that may be estimated. Software syntax is introduced to enhance the likelihood of readers' trial and adoption.