An extended paradigm for measurement analysis of marketing constructs applicable to panel data

In an effort to advance measurement analysis in marketing research, the authors propose three extensions of the current scale development paradigm. First, the authors focus attention on the trait–state distinction and present a model that separates stable sources of substantive variation in a construct from transient sources. Second, the authors develop a classification of measurement error that distinguishes six types of error on the basis of (1) whether the error is stable or transient and (2) whether the error affects individual items, subsets of items, or all items in a scale. They also show how these errors can be modeled using a factor-analytic specification. Third, the authors argue that marketing researchers should make the means of the scale items an explicit component of measurement analysis and should test for the invariance of item loadings and intercepts as a prerequisite for meaningful comparisons of scale means across samples and over time. To illustrate the benefits of the proposed procedure, the authors present an extended application of the model to the constructs of brand loyalty and deal proneness.