MEASURING EXTREME RESPONSE STYLE

Research on extreme response style (ERS) in rating scale responseshas been characterized by conflicting findings and little agreementover how to assemble and validate ERS measures. This articleproposes that, when ERS is defined as a proportion of extremeresponses, an ERS measure will be more accurate if the itemsare uncorrelated and have equal extreme response proportions.Furthermore, appropriate stochastic models should be used toassess the internal reliability and convergent validity of thesemeasures. An ERS measure is created and validated with thismethod, using items from a survey administered in 1975 and 1987to large samples of U.S. adults serving on a consumer panel.We find that ERS is stable over a lengthy survey compared toa benchmark stability for a "perfect" measure. Furthermore,the distribution of ERS over this population is stable overtime. Respondents' ERS is related to their age, education level,and household income but not to their gender.