Longitudinal macro-analysis of car-use changes resulting from a TOD-type project: The case of Metro do Porto (Portugal)
Introduction
Rapid urbanization has always been a major concern for urban planners challenged by the need to accommodate population growth and address increasing travel demand while preserving the environment and quality of life. From a planning perspective, this challenge can be addressed through the concept of transit-oriented development (TOD), which aims to tackle traffic congestion and urban growth simultaneously by providing dense and mixed-used settlements around public transport nodes. Transit-oriented development was defined by Calthorpe (1993, 7) as a “mixed-use community within average 2000-foot [600 meters] walking distance of a transit stop and core commercial area. TODs mix residential, retail, office, open space, and public use in a walkable environment, making it convenient for residents and employees to travel by transit, bicycle, foot, or car.” Since the 1990s, TOD has been gaining prominence, with TOD projects implemented worldwide (like the Grand Paris project and the Corridors of Freedom Initiative in Johannesburg launched in 2010 and 2014, respectively). This growth is also reflected in a substantial increase of TOD-related publications in scientific journals (Ibraeva et al., 2020).
Since TOD is supposed to foster a reduction in car trips and the transition to sustainable transport modes, its influence on travel behavior has been the focus of numerous studies. The findings vary due to different national contexts and methods used for assessment, yet, in general, TOD is associated with fewer car trips and greater public transport patronage than in comparable automobile-oriented neighborhoods (see Section 2). Despite recent notable progress in the analysis of TOD effects on travel behavior (Ibraeva et al., 2020), studies addressing this issue using a longitudinal research approach are rare. Nevertheless, longitudinal analysis can bring several advantages compared to the typically adopted cross-sectional research design. First of all, a longitudinal analysis of panel data allows exploring more informative data and more variability, while still accounting for heterogeneity (Baltagi, 2005). Second, incorporating the temporal dimension in the analysis permits detection of a transport evolution occurring over years, which is essential in the analysis of TOD influence on travel behavior: as a new public transport service is introduced, it takes some time for people to adjust their habits and mode choice to the new transport option. The same applies to slowly occurring changes in the built environment of station areas.
In this paper, we develop and apply a longitudinal research approach to analyze the impact of the Metro do Porto – a metro system launched in Portugal in the early 2000s – on the use of private cars for commute trips (work or study). Introduced in just nine years on a territory that had no rapid transit service until then, Metro do Porto can be considered as a natural experiment in the sense that we analyze actual post-intervention changes in mode choice as opposed to preliminary project feasibility studies or studies based on stated preferences. To evaluate metro effects on mode choice, we have used a difference-in-differences (DID) model. This is a type of model that, to the best of our knowledge, has never been used before in the context of TOD travel behavior, but is highly appropriate for before/after analyses. A DID model estimates the effect of a treatment (in this case, the effect of the introduction of the metro) by comparing the average differences in an outcome variable (car trips) between a treated group and a control group (metro-served and non-metro-served civil parishes, respectively). To address potential bias from spillover effects and spatial autocorrelation, we also used a spatial DID (SDID) model.
The Metro do Porto network has a total length of 67 km and comprises 82 stations, of which 14 are underground (https://www.metrodoporto.pt). In addition to serving dense urban areas (notably the central area of Porto), it also serves residential suburbs and rural outskirts. In several cases, the introduction of the metro was accompanied by the rehabilitation and/or renovation of adjacent areas to make them more attractive, safe, and vibrant (Pinho and Vilares, 2009). These interventions resonate with TOD principles, and this is why we classify Metro do Porto as a TOD-type project. Note, however, that Metro do Porto was not formally launched as a TOD project, and that, depending on the surrounding environment, station types vary. Some of them ideally comply with TOD characteristics, while others can be better classified as transit-adjacent (TAD); i.e., stations located in proximity to urban settlements but not properly articulated with them – or park & ride (P&R). In our analyses, we account for the difference between station types and compare their effects on mode choice.
In contrast to previous studies that have mostly concentrated on immediate station areas, this is a macro-analysis conducted at the level of the (civil) parish (“freguesia”), as one of our goals is to know whether the effect of a large TOD investment is visible not only in the proximity of stations but also on a wider scale. The main research question is: to what extent did the introduction of the metro affect the number of (private) car trips since, in the absence of the metro, the car was the most convenient and fast mode of transport in the Porto region? We believe that a ten-year interval is appropriate for the purpose, as this period may encompass not only changes in residents' preferences but also emerging transformations in the built environment (Crowley et al., 2009; Dong, 2016). Thus, we also evaluate the influence of the additional covariates typically present in TOD research such as land-use and socio-economic variables.
The remainder of the paper is structured as follows. The next section provides an overview of the existing literature on the effects of TOD on travel behavior, aiming to present existing research findings and some uncertainties (frequently associated with the lack of longitudinal research) that remain in this field. After that, we describe the case study, focusing on the socio-economic, urbanization, and travel mode trends observed in the Porto region. Special attention is given to the evolution of car use in metro-served and non-metro-served parishes. Our methodological approach is explained in the following section, where we provide a description of our DID model and its spatial extension. Subsequently, we present and discuss the modelling results we have obtained, and provide a performance analysis of TOD- and TAD-served parishes. Finally, in the last section, we summarize our study and identify some directions for future research.
Section snippets
Literature overview
This section is intended to provide an overview of the numerous studies addressing the influence of TOD on travel behavior. The resulting estimations of the TOD effects vary depending on the methodology applied, variables used, and national or urban context considered. Nevertheless, it is possible to generalize existing findings to some extent.
Considering transit-related variables, proximity to a transit station largely determines the attractiveness of transit use for residents (Cervero, 2007;
Porto region evolution
Focusing essentially on temporal changes, we provide in this section an overview of the dominant urbanization, transport system, and travel mode trends in the Porto region in the years before the launch of Metro do Porto and after the first nine years of its operation. For this purpose and, more broadly, for the analyses conducted later in this paper, we designate the Porto region as a group of seven municipalities served by the metro system: Gondomar, Maia, Matosinhos, Porto, Póvoa de Varzim,
Methodological approach
In this section, we focus on the methodological approach adopted in our study (Fig. 8). Once we decided to study the impact of Metro do Porto on travel mode choice, we looked for the data available. As stated before, the unit of analysis was the civil parish. After collecting the relevant data (population, land use, transport system, and mode choice), and performing a preliminary analysis to observe the mode choice trends for metro-served and non-metro-served parishes in the period 1991–2011,
Study results and discussion
In this section, we present, analyze and discuss the results obtained through the estimation of the models using the splm package of the R software (Millo and Piras, 2012). It is divided into two subsections, dedicated, respectively, to the DID model and the SDID model. In the last part of the second subsection, we focus on the performance of parishes depending on the predominant type of metro service (TOD, TAD, and P&R) they offer.
Conclusion
In this paper, we presented a study aimed to analyze the impact of Metro do Porto on the use of private cars for commute trips (work or study). The analysis extends over a ten-year period (2001−2011) and is essentially based on census data: 120 civil parishes (“freguesias”) were selected as units of analysis to explore whether metro, as a large infrastructure project, produced effects noticeable on a macro scale. While the majority of studies about the effect of TOD on car use comes from the
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
The research described in this paper has been carried out in CITTA (Research Center for Territory Transports and Environment) and at the Department of Engineering Systems and Services of the Delft University of Technology (3-month visit) in the framework of the doctoral thesis of Anna Ibraeva (PD/BD/135417/2017). The research was partially carried out in the Delft University of Technology. Both the research center, the doctoral thesis are funded by FCT – Fundação para a Ciência e Tecnologia.
References (58)
- et al.
Testing panel data regression models with spatial error correlation
J. Econ.
(2003) - et al.
Causal identification of transit-induced gentrification and spatial spillover effects: the case of the Denver light rail
J. Transp. Geogr.
(2018) - et al.
A spatial multiple treatment/multiple outcome difference-in-differences model with an application to urban rail infrastructure and gentrification
Transp. Res. A
(2019) - et al.
Vehicle trip reduction impacts of transit-oriented housing
J. Public Transp.
(2008) - et al.
Suburbanization and transit-oriented development in China
Transp. Policy
(2008) - et al.
Difference-in-differences techniques for spatial data: local autocorrelation and spatial interaction
Econ. Lett.
(2015) - et al.
Transit-oriented development: a review of research achievements and challenges
Transp. Res. A Policy Pract.
(2020) - et al.
Advance transit oriented development typology: case study in Brisbane, Australia
J. Transp. Geogr.
(2014) - et al.
Panel data models with spatially correlated error components
J. Econ.
(2007) - et al.
Millennials and car ownership: less money, fewer cars
Transp. Policy
(2017)