Pick a Number between 0 and 100: An Examination of Percentage-Based Scales

Abstract

This study provides evidence on the relationship between scale characteristics and participant responses for percentage-based scales (i.e., 101 points) in accounting research. A 4×1 between-subjects experiment examines how common labeling designs affect various statistical properties, including means, variance, normality of the distribution, and frequency of responses. The results indicate that labels on percentage-based scales have a significant impact on the distribution of participants’ responses. Labeling only the endpoints is the lone condition that results in normally distributed data. Additional analyses suggest that labels on percentage-based scales influence participant responses in multiple ways. First, as the number of labels increases, participants may not adequately consider, and thus ultimately select, unlabeled points. Second, while participants seem to inherently interpret percentage-based scales in quartiles and deciles, labeling as such exacerbates this tendency. Finally, when more labels are present, participants seem to engage an anchoring heuristic when selecting their response. Taken as a whole, the results suggest that accounting researchers may benefit from labeling only the endpoints of percentage-based scales.