Improving Rating Scale Measures by Detecting and Correcting Bias Components in Some Response Styles

Abstract

The author examines whether the response styles of yeasaying and standard deviation in rating scale responses convey information on respondents’ attitudes or create bias that distorts attitude information and marketing research. A method is proposed to identify attitude information components and bias components in response styles, using prediction errors in attitude-behavior models. Analysis of data from a large-scale consumer survey supports the presence of both attitude information and bias components in standard deviation, and an attitude information but not a bias component in yeasaying. This finding suggests that correcting rating scale data by removing the bias but not the attitude information in standard deviation can increase the accuracy of survey research. Examples are given of how bias in standard deviation, and the scoring correction, affect segmentation research.