Customer satisfaction in urban rail: a study on transferability of structural equation models

Studies on the quality of service thrive to quantify passenger satisfaction of public transport services which is a key factor in how these services are used. Service quality evaluation is performed by means of objective and subjective approaches. Considering that people are attracted to public transport services based on their own perception, subjective approaches are utilized to capture user judgment. However, there are various subjective approaches in the literature to model service quality in which different lists of factors and model structures are identified. This paper aims to test the spatial transferability of these approaches at various levels. All the adopted approaches are based on a Structural Equation Model (SEM) to evaluate the relationship between perceived service quality and customer satisfaction. First a synthesis of the literature on model transferability is presented to compare these approaches in terms of model characteristics (structural equivalence) and parameter values (measurement equivalence). Three classes of model characteristics, including six distinct SEM structures in total, are tested to identify the factors affecting service quality: (1) recommendations of previously published models from the literature, (2) established customer satisfaction theories, and (3) adopting an exploratory approach and developing it based on the previous two approaches and the nature of locally collected data. Tehran Metro Line 3, which is a heavy urban rail mode, is used as a case study. The dataset consists of three hundred validated responses collected through a customer satisfaction survey. The results indicate that model characteristics are not transferable to this case study (i.e., none of the SEM structures from previous studies were directly transferable). The best model is developed when the entire modeling procedure is repeated using a wide range of affecting factors and gradually narrowing them down through model development. The final model comprises four latent variables, namely major services, comfort, security, and minor services, the first two of which have the largest effect on service quality.