Predictive Validity of Multiple-Item Versus Single-Item Measures
Why do marketing practitioners favor single-item measures but most academic marketing researchers believe that multiple-item measures are necessary to validly represent the major constructs or variables studied in marketing? Why do academic journals in marketing tend not to accept articles with single-item measures of major constructs?
In a recent issue of Journal of Service Research, Aimee Drolet and Donald Morrison argued in favor of single-item measures on the practical grounds that, in at least some cases, a single item provides most of the “information” that can be derived from a multiple-item scale. Further, the only theoretical argument for single-item measures was made by John Rossiter in his 2002 article in International Journal of Research in Marketing, which introduces his alternative procedure for scale development, titled C-OAR-SE. According to Rossiter, a single-item measure is appropriate if the object being rated is simple and unambiguous, such as an advertisement or a brand, and if the attribute on which the object is being rated is also simple and unambiguous to the raters, such as liking the advertisement and overall evaluation of the brand. Rossiter calls these “doubly concrete” constructs.
However, to date, there has been no test of whether a single-item measure of a doubly concrete construct is as valid as a multiple-item measure of the same construct, and there has been no test of whether the researcher “loses” anything by using a single-item measure.
In their May 2007 Journal of Marketing Research article (“The Predictive Validity of Multiple-Item Versus Single-Item Measures of the Same Constructs), Lars Bergvist and John R. Rossiter offer a test of the predictive validity of single-item versus multiple-item measures of two widely studied, doubly concrete constructs: attitude toward the ad and brand attitude. They find that it is unnecessary and wasteful to use multiple-item scales to measure basic constructs in marketing. Although the tests are based on attitude toward the ad and brand attitude, the single-item recommendation is also intended for other constructs in marketing that involve basic objects, such as companies, retailers, salespeople, or sales promotions, and basic attributes, such as perceptions, intentions, or satisfaction. Marketing research practitioners have always used single-item measures for constructs such as these, and academic researchers can now follow without concern about losing validity (r) or explanatory power (R-square).
JMR readers are invited to participate in further discussion on this controversial topic.
