Messung in der Marketingforschung — Status Quo und Perspektiven

Abstract

Latent constructs are important elements of modern theories in empirical marketing research. The underlying latent variables are, generally, modelled as quantitative variables. The measurement of latent variables entails two distinct problems. Firstly, evidence has to be provided that the latent variable really exists as a quantity and that it is, consequently, measurable. Secondly, a concrete measurement instrument has to be devised that enables us to infer measures of latent variables from observations of manifest variables. The predominant measurement paradigm of classical test theory fails to provide this sort of evidence. It rather presumes measurement. Recently, the discussion of models with formative indicators has gained momentum again. Typically, formative indicators are seen as an alternative to reflective indicators entailed by classical test theory. The analysis of criteria for choosing the appropriate model (reflective or formative) that are offered in the literature shows that these criteria do not allow for a decision based on empirical evidence. Finally, formative indicators turn out to be incompatible with the notion of a real, truly existing latent variable. Thus, formative models are no alternative to reflective measurement models. By their nature, they are models of structural relationships, i.e. they pertain to structural theories. A scientifically satisfactory solution to the problem of measuring latent variables has to be based on a definition of measurement that is compatible with the scientific concept of quantity. With the Rasch model, item response theory offers a class of models which comply with the requirements of the classical concept of quantity.