Reversed-Polarity Items and Scale Unidimensionality

Abstract

A standard procedure in the development of multi-item measurement scales is to incorporate reversed-polarity items to control for and/or identify acquiescence response bias. In spite of the broad acceptance of this approach, very little work has been done to evaluate the impact of reversed-polarity items on the dimensionality of scales. This study empirically evaluates the impact of reversed-polarity items on the unidimensionality of several well-known marketing measures. The authors suggest that use of reversed-polarity items may present a substantive problem for marketing scholars because of the resulting degradation of scale unidimensionality. The existence of this phenomenon is confirmed in a multisurvey, multiscale, binational research design. Implications for marketing scale developers and measurement theoreticians are discussed, theoretical bases that might explain the phenomenon are explored, and further research suggestions are presented.