Urban rail transit network topology evolutionary stage has influence on rail ridership: Insights from linear mixed-effects models with heterogeneity in variances

Abstract

Determining the impact of the evolutionary stage (ES) of urban rail transit (URT) network topology on daily ridership (DR) can be beneficial for policymakers when analyzing recent stagnated ridership in some cities. This paper used city-level data from 26 Chinese cities from 2013 to 2019 to investigate this. Eleven variables comprising socioeconomic attributes (population density and the average wage of employees), the availability of other transit modes (the number of taxis, the number of private cars, the bus network length, and the street network length), service-related attributes (the number of trains in operation, the average headway during peak period, the opened days, and rail fare), and a network topological characteristic (the average path length) were introduced as covariates. In addition, city class was introduced as a grouping variable to capture the unobserved characteristics of the investigated cities. Two linear mixed-effects models (M1 and M2) were developed, with an interaction term between city class and ES being introduced in M2. The variances were allowed to vary to account for the heterogeneity of the predictor, which improved the overall fit of the models. The estimation results reveal a significant positive within-subject effect of ES on DR in both models, which varied according to city class. This implies that for each city, improving ES increased DR and the scale of the increase in DR is associated with the city’s class. However, there is no evidence of a relationship between ES and DR across different cities. In addition, the heterogeneity in the variance indicates that underestimating the variances of the effects could lead to inaccurate conclusions. The results of this study can help transit agencies in mastering ridership and assessing the designs of URT network topology.