Latent Variable Modeling in Business Research: a Comparison of Two- Step Approach with Structural Equation Modeling

Researchers in marketing, organization behaviour, and the other fields of business research frequently study variables that are abstract and not directly observable, for example, country image, brand loyalty, service quality, job satisfaction, work stress, managerial competence, etc. To measure such variables, referred to herein as latent variables, latent traits, or constructs, a multiple-item scale of measurement is often used, in which the scale measure is usually a weighted or unweighted sum of the individual items that form the scale. These items in turn are usually measured as dichotomous or polytomous variables. The prevalent guidelines for scale construction in marketing and business research textbooks are based on classical test theory (CTT) (e.g., Aaker, Kumar, & Day, 2004; Churchill & Iacobucci, 2002; Cooper & Schindler, 1998), a measurement technology that has its origin in the field of psychometrics. The recent edition of the Handbook of Marketing Scales by Bearden and Netemeyer (1999) describes numerous scales that have been constructed using CTTbased procedures. Concepts such as reliability, item-total correlations, and Cronbach’s alpha are all based on the CTT model, which posits that the observed scale score is the sum of a true score and an error term.