Abstract

This paper addresses the satisfaction effect of real-time information applications (apps) and their interaction with the actual quality of service of public transport (PT) in 13 Chilean cities. Our study has two methodological innovations; first, we combine a discrete choice model and a sentiment analysis conducted with a Large Language Model (ChatGPT3.5-turbo), which classifies an open-ended satisfaction question, allowing us to embed qualitative data into a quantitative model. Second, we used an objective quality of service metric (a headway reliability index, based on bus GPS data) as input in the passenger satisfaction model, which is an improvement over previous passenger satisfaction models that rely on perceived (rather than measured) service attributes only. Therefore, the model accounts for the relationship between user satisfaction, service attributes, app-induced behavior, and perception changes. The results highlight a symbiotic relationship between having access to real-time information and the PT level of service. That is, the use of real-time information makes passengers more satisfied, but the effect is greater under higher PT service regularity conditions. Sentiment analysis revealed that satisfied users (70% of the sample) value the waiting time information provided by the app and the resulting ability to manage their time. Unsatisfied users (12% of the sample) mainly criticize the frequency of service and the accuracy of the data displayed in the app. These findings promote user-centered policy making, especially in developing countries where high-quality real-time information should be more widely available. Finally, service regularity is statistically significant in explaining user satisfaction, even when controlled for the use of realtime information. This indicates that having regular headways matters to users, regardless of whether they receive real-time information about waiting times.