Are Likert scales unidimensional?

Abstract

Likert's method of scoring his categories can be justified by the Rasch Rating-Scale model if the categories produce unidimensional responses. Tests involving eight items from the 1956–1958–1960 American Panel Study suggest that unidimensionality often is not achieved with the Likert categories strongly agree, agree, disagree, and strongly disagree. Combining the first two and the last two of these categories can be justified under a two-dimensional, Content-by-Intensity, Rasch model, provided that the two components are not confounded in the response process. Tests using the same data suggest that such confounding may occur fairly often. Thus the methods commonly used for analyzing Likert data may not be as generally applicable as research workers have assumed. Validation of measurement involving rating scales should be an integral part of any research employing the Likert design.