Stevens’ measurement scales in marketing research – A continuation of discussion on whether researchers can ignore the Likert scale’s limitations as an ordinal scale

This article discusses the use of Stevens’ measurement scales in marketing research, contributing to a broader discussion, underway for over 70 years, as to whether researchers can ignore the Likert scale’s limitations as an ordinal scale. The central question explored is whether the use of various statistical methods and techniques in marketing research has gone too far, limiting researchers’ horizon of thought, leading erroneous conclusions to be drawn, and diverting attention from trying to explain the non-quantitative attitudes of consumers (who are people, not machines or AIs). Stevens’ measurement scales are still widely used in data analysis across social sciences, including marketing research. Although they were revolutionary, they had certain flaws which have fueled an ongoing debate about the acceptability or permissibility of using different tests and statistical techniques at different scales and levels of measurement. The Likert scale, one of the scales most frequently used to measure customer attitudes, was intended to overcome the limitations of simple scales, having the advantage of being multi-item. However, historically, two competing views have evolved independently of each other, in the related literature and in the practice of empirical research: one emphasizing the ordinal nature of Likert scales, the other interpreting them as having interval-scale properties. This debate has significant consequences for the permissible scope of statistical analysis of empirical data. The problem discussed here is likely to become even more complex with the development of artificial intelligence (AI), machine learning, data science and big data, as data scientists perform computational analysis but are not often involved in data collection or deciding about how data is represented.