Journal of Econometrics 89 (1999) 393 — 421 Representation of measurement error in marketing variables: Review of approaches and extension to three-facet designs Richard P. Bagozzi*, Youjae Yi, Kent D. Nassen  University of Michigan Business School, Ann Arbor, MI 48109-1234, USA  Seoul National University, USA  University of Michigan, USA Abstract This paper explores approaches for modeling measurement error in marketing re- search, including random, method and measure specific sources of error. The following approaches are considered: classic confirmatory factor analysis, second-order models, panel models, additive trait-method models, correlated uniqueness models, covariance components analysis, additive trait-method-measure specific-error models, and the direct product model, where traits and methods interact. Finally, a three-facet multiplicative model is addressed wherein latent variables underlying a phenomenon under investiga- tion are shown to interact with multiple methods and occasions of measurement. The three-facet model is illustrated on a study of consumer attitudes toward losing weight explicitly conducted for this paper.  1999 Elsevier Science S.A. All rights reserved. Keywords: Construct validity; Multitrait-multimethod matrices; Method error; Confir- matory factor analysis Questionnaire items and rating scales are used throughout marketing re- search to operationalize theoretical constructs and make policy decisions. Among other areas, items and rating scales form the basis for measuring perceptions of service quality (Parasuraman et al., 1994), customer satisfaction * Corresponding author. E-mail:bagozzi@umich.edu 0304-4076/99/$ — see front matter  1999 Elsevier Science S.A. All rights reserved. PII: S 0 3 0 4 - 4 0 7 6 ( 9 8 ) 0 0 0 6 8 - 2