Testing a model of pull production: a paradigm for manufacturing research using structural equation modeling

Abstract

The measurement of unobservable (latent) variables has been a recent phenomenon in the manufacturing research area. Most available empirical research in manufacturing has been exploratory in nature and has borrowed its methods extensively from other fields such as psychology, sociology, and marketing. Traditional exploratory techniques have been used to provide preliminary scales and assess measurement properties. Manufacturing researchers have, however, overlooked the assessment of unidimensionality, an essential measurement property and a basic assumption of measurement theory. An explicit evaluation of unidimensionality can be accomplished with a confirmatory factor analysis (CFA) of individual measures as specified by a multiple‐indicator measurement model. A paradigm for scale evaluation incorporating CFA for the assessment of unidimensionality is outlined here along with methodology to assess other measurement properties such as convergent validity, discriminant validity, composite reliability, and average variance extracted. A measurement model is tested first followed by a structural model of interest. The hypothesized structural model relates pull production with two of its antecedents, setup improvement and preventive maintenance practices. It further relates pull production to one of its consequences, delivery dependability. Responses from 244 firms are used to test the measurement and structural model.