Search results

This paper aims to extend Rossiter’s C-OAR-SE method of measure design (IJRM, 2002, p. 19, p. 4, pp. 305-335; EJM, 2011, p. 45, p. 11, p. 12, pp. 1561-1588) by proposing five distinct construct models for designing optimally content-valid multiple-item and single-item measures.

The paper begins by dismissing convergent validation, the core procedure in Nunnally’s (1978) and Churchill’s (1979) psychometric method of measure design which allows alternative measures of the same construct. The method of dismissal is the mathematical demonstration that an alternative measure, no matter how highly its scores converge with those from the original measure, will inevitably produce different findings. The only solution to this knowledge-threatening problem is to agree on an optimal measure of each of our major constructs and to use only that measure in all future research, as is standard practice in the physical sciences. The paper concludes by proposing an extension of Rossiter’s C-OAR-SE method to design optimal standard measures of judgment constructs, the most prevalent type of construct in marketing.

The findings are, first, the mathematical dismissal of the accepted practice of convergent validation of alternative measures of the same construct, which paves the way for, second, the proposal of five new C-OAR-SE-based construct models for designing optimal standard measures of judgment constructs, three of which require a multiple-item measure and two of which a single-item measure.

The common practice of accepting alternative measures of the same construct causes major problems for the social sciences: when different measures are used, it becomes impossible, except by remote chance, to replicate findings; meta-analyses become meaningless because the findings are averaged over different measures; and empirical generalizations cannot be trusted when measures are changed. These problems mean that we cannot continue to accept alternative measures of the constructs and that, for each construct, an optimal standard measure must be found.

The ideas in this paper, which have untold value for the future of marketing as a legitimate science, are unique to Rossiter’s C-OAR-SE method of measure design.

New measures in marketing are invariably created by using a psychometric approach based on Churchill's “scale development” procedure. This paper aims to compare and contrast Churchill's procedure with Rossiter's content‐validity approach to measurement, called C‐OAR‐SE.

The comparison of the two procedures is by rational argument and forms the theoretical first half of the paper. In the applied second half of the paper, three recent articles from the Journal of Marketing (JM) that introduce new constructs and measures are criticized and corrected from the C‐OAR‐SE perspective.

The C‐OAR‐SE method differs from Churchill's method by arguing for: total emphasis on achieving high content validity of the item(s) and answer scale – without which nothing else matters; use of single‐item measures for “basic” constructs and for the first‐order components of “abstract” constructs; abandonment of the “reflective” measurement model, along with its associated statistical techniques of factor analysis and coefficient alpha, arguing that all abstract constructs must be measured as “formative”; and abandonment of external validation methods, notably multitrait‐multimethod analysis (MTMM) and structural equation modeling (SEM), to be replaced by internal content‐validation of the measure itself. The C‐OAR‐SE method can be applied – as demonstrated in the last part of the article – by any verbally intelligent researcher. However, less confident researchers may need to seek the assistance of one or two colleagues who fully understand the new method.

If a measure is not highly content‐valid to begin with – and none of the new measures in the JM articles criticized is highly content‐valid – then no subsequent psychometric properties can save it. Highly content‐valid measures are absolutely necessary for proper tests of theories and hypotheses, and for obtaining trustworthy findings in marketing.

C‐OAR‐SE is completely original and Rossiter's updated version should be followed. C‐OAR‐SE is leading the necessary marketing measurement revolution.