RESEARCH QUESTIONS AND HYPOTHESIS
Research on bicycling to transit, particularly in a North American suburban context, is limited (Bachand-Marleau, Larsen, and El-Geneidy 2011; Mitra and Schofield 2019). A recent systematic literature review (Pinder et al. 2018) identified 28 existing studies that have focused on transit-cycle integration; 11 of these have a North American focus, and none of these 11 studies includes a disaggregate analysis of cycling mode choice behavior. For example, Ravensbergen et al. (2018) analyzed data collected by Metrolinx, the public operator of the regional commuter transit system- GO Transit- in Southern Ontario, Canada, and explored stated barriers to cycling to train stations and bus stops across the region. The paper did not specifically focus on the suburban context nor examine mode choice behaviour, likely due to the unavailability of geographical data.
In this paper, we explore the potential for bicycling to commuter rail stations, to inform station access planning and policy that would reduce demand for car parking. More specifically, we examine the socio-demographic, attitudinal, and environmental correlates of the likelihood of bicycling for transit access trips. Our key hypothesis is that a passenger’s travel-related attitudes and the built environment conditions of the travel route will affect their probability of cycling.
METHODS AND DATA
Passenger Survey of Bicycling Behavior
GO Transit has an extensive network of bus and train services across Sourthern Ontario. Trains operate along seven corridors centering on Union Station in downtown Toronto, the largest employment center in Canada (Fig. 1).
We conducted an online survey of GO Train passengers in fall 2017 at three suburban stations. The selected stations — Burlington, Appleby, and Bronte — are located on the Lakeshore West train corridor, which is GO Transit’s highest ridership corridor. The surveys were collected on six weekdays (two days at each station) under sunny and moderate weather conditions. A research assistant distributed postcards with a link to the survey during morning peak (7am to 8 am) to as many passengers as possible, who were asked to complete the survey during the journey or within 24 hours after travel. In the survey, we asked the respondents about their transportation mode choice motivations and preferences, bicycling comfort in different hypothetical environmental conditions, as well as preference for end-of-trip facilities (e.g., parking, change room, storage lockers). These perceptions and motivations were captured on a 5-point scale based on the respondent’s agreement to a group of statements. The full survey instrument can be found elsewhere (Mitra and Schofield 2019).
Approximately 1,550 postcards were distributed, and we received 306 responses, at a 19.7% response rate. Of these respondents, 265 provided data on their travel mode to a GO Train station, age, and gender. The sample characteristics are summarized in Table 1, along with the number of responses to each question.
Built Environment Data
We determined passengers’ home locations based on their six-digit postal codes (n= 213), which represent whole or part of a residential block. Several built environment variables were computed using GIS. We conducted this analysis at the Dissemination Area (DA) level, which is the smallest geographical unit for which census data is available. The data used to compute these variables came from DMTI Spatial’s land use dataset 2015 (LUR), Statistics Canada 2011 Census, and Environics Analytics Demographic Estimates. We estimated the shortest travel route between a passenger’s home and station using ArcGIS network analysis. The built environment features within all DAs located within a 500m buffer of the shortest route were summarized and normalized. Through visual inspection of the home and station locations on a Google map, we identified whether the passenger had to cross the QEW (a major controlled access highway).
Data Analysis
We explored a total of 17 variables relating to travel motivations, perceptions, and comfort bicycling in different weather and environmental conditions. To reduce the dimensionality of the data and identify clearly distinguishable ‘factors’, we conducted a principal component analysis (PCA). The PCA produced six components that cumulatively explained 72% of the total variance (Table 2).
We adopted a binomial logistic regression approach to explore the correlates of the likelihood of bicycling (versus other modes). However, our dataset is relatively small (n=265), with few instances of the desired outcome (i.e., bicycling to GO station), which could potentially lead to a ‘separation’ problem in model estimation. To overcome this potential problem, Firth logistic regression models with penalized likelihood method were estimated (Firth 1993; Heinze and Puhr 2010). Also, we estimated partially adjusted models that explored the correlation between one variable of interest and the log-likelihood of bicycling, controlling for variations in age and gender, neither of which were associated with the propensity of bicycling at α=0.05.
FINDINGS
Of the socio-demographic variables tested, respondents from multi-vehicle households were less likely to bicycle to a GO Train station (Table 3). We also found a strong positive correlation between past-month’s bicycling frequently for transportation and recreational purposes and the likelihood of bicycling to a station, unsurprisingly indicating that those who use a bicycle more for various other activities would also be more likely to use bicycles for station access trips. A passenger’s household income was not associated with cycling to suburban stations, contrary to findings from research that has focused broadly on all transit access trips (Chakour and Eluru 2014; Chan and Farber 2019; Flamm and Rivasplata 2014; Ravensbergen et al. 2018).
Next, and in regard to passenger attitudes and preferences, we found that ecological motivation (i.e., preference toward environmentally sustainable transportation choices and physical activity) was positively associated with the odds of bicycling to the station (OR = 2.83). In addition, comfort bicycling in adverse weather and climatic conditions was positively associated with bicycling (OR = 2.57). In contrast, preference toward facilities at a GO Station (e.g., secure parking room, change room, shower) did not explain bicycling for station access.
Several built environment characteristics along the shortest route may explain bicycling for transit access trips. Specifically, distance between home and station locations was moderately associated with the probability of cycling; those living less than 2 km from station were 4.6 times more likely to use a bicycle (α=0.1) when compared to respondents living more than 5 km away. Having to cross the QEW (a major highway) to get to a station reduced the likelihood of cycling. In other words, those residing on the other side of the QEW from their respective access station would be 0.06 times or 94% less likely to bicycle. Additionally, density of streets was negatively correlated with bicycling for transit access trips, perhaps because in a suburban context higher street density can represent a built environment with higher traffic volume.
Contrary to the popular assumption (Cervero, Caldwell, and Cuellar 2013), the presence of bicycling facilities (which were mostly painted bicycle lanes on major suburban roads) did not change the probability of cycling. In addition, the self-reported comfort level in different road conditions had no impact on bicycling. While empirical research has produced mixed evidence of the impact of bicycle infrastructure (Buehler and Dill 2016), suburban contexts are understudied and a more nuanced examination of the relationship between suburban bicycle infrastructure and mode choice remains a topic for future research.
Many of our findings contradict what is expected in urban settings and emphasize the importance of bicycling research focusing on the North American suburban context. The empirical findings can be used by municipalities and commuter rail operators alike in developing or improving policies, plans and programs that would reduce automobile dependence and car parking demand at stations. For example, 62% of current GO Train passengers drive to a station, and as Metrolinx plans for major service expansion that might produce up to an additional 150,000 passengers per day, a continuation of current station access behaviour will create significant pressure on existing parking facilities (Metrolinx 2016). As a result, promoting alternative means for station access trips, such as bicycling, has become a more important policy issue than ever, and the results from this study begin to identify customers who could be targeted for marketing campaigns as well as to emphasize the importance of safer bicycling environments.
ACKNOWLEDGMENTS
This research was supported by a Social Sciences and Humanities Research Council of Canada (SSHRC) Insight Grant. James Schofield at Ryerson University conducted the surveys.