Spillover effects in transit networks: A parameterized weight matrix spatial lagged approach
Journal
Journal of Urban Mobility
Date Issued
December 2024
Author(s)
DOI
10.1016/j.urbmob.2024.100081
Abstract
Public Transit (PT) systems aim to provide social, economic, and environmental benefits to modern cities offering reliable transportation service to users and ultimately reducing the problems related to traffic externalities due to car dependency. With the growing trend of urbanization and the associated phenomenon of cities’ sprawl being increasingly evident, significant attention should be given to understanding PT systems’ performance and then improving their efficiency. Considering the spatial characteristics of the PT system's performance promotes the environmentally friendly transport “character” that every city endeavors. This paper aims to incorporate the spatial spillover effects of a realistic PT system by augmenting information about the service network along with socio-economic variables, in a spatial demand-supply econometric framework. In detail, geographically separated demand and supply information on bus stops and lines was analyzed by a spatial econometric model, namely, the Spatial Lagged X (SLX) model which may be formatted so that can soundly handle, spatial and –in particular- network data like those coming from the General Transit Feed Specification (GTFS) protocol widely used in PT systems. The novelty and the importance of the proposed model rely on the ability to introduce transit network structure within the framework of spatial econometric regression, fostering the explanatory statistical analysis over networked information. The developed modeling approach was applied to the urban and rural PT system of Nicosia (Cyprus), incorporating the spatial spillover effects of the system over more than 1,500 bus stops, 40 lines, and 252 different but spatially connected postcode areas. The results of the SLX model were compared with other demand models of this form typically used, namely the Ordinary Least Square model and standard Spatial Autoregressive models, providing solid evidence of the benefits of incorporating the network structure in spatial demand modeling, giving valuable input for further planning purposes.