Reinterpretation of the Likert Scale for Public Transportation User Satisfaction: Pattern Recognition Approach

Abstract

Myriad cities around the world conduct public transportation satisfaction investigations, and most of them use a linguistic Likert scale (i.e., it runs from strongly dissatisfied to strongly satisfied). Most such studies neglect two critical issues: users' complex cognitive systems and the fundamental problem of applying the Likert scale. Thus the present study considered those issues and analyzed public transportation user satisfaction intervals of heterogeneous groups with different cognitive systems. The study used a Gaussian mixture model and a pattern recognition approach, with satisfaction data from public transportation users in South Korea. The results indicated that users expressed considerably different satisfaction intervals for the same linguistic Likert items: five intervals instead of the 7-point Likert scale in the survey. In addition, each Likert item was clustered into two distinguishable satisfaction intervals, and those heterogeneous groups were classified not randomly but systematically by sociodemographic attributes and characteristics of public transportation use. On the basis of the results, significant implications were found about public transportation satisfaction surveys and their analysis.