- Email: [email protected]
Exploring built environment correlates of walking distance of transit egress in the Twin Cities
Journal of Transport Geography 64 (2017) 132–138
Contents lists available at ScienceDirect
Journal of Transport Geography journal homepage: www.elsevier.com/locate/jtrangeo
Exploring built environment correlates of walking distance of transit egress in the Twin Cities
MARK
Jueyu Wanga,⁎, Xinyu Caob a b
Humphrey School of Public Affairs, University of Minnesota, Twin Cities., 301 19th Ave S, Minneapolis, MN 55455, United States Humphrey School of Public Affairs, University of Minnesota, Twin Cities., 301 19th Ave S, Minneapolis, MN 55455, United States
A R T I C L E I N F O
A B S T R A C T
Keywords: Station area planning Land use Pedestrian behavior Transit planning Transit-oriented development
Most studies on walking distance to transit stops either emphasize transit access or do not distinguish transit access and egress. Furthermore, environmental correlates of walking distance may differ by stop location. Using the 2010 Transit Onboard Survey in the Minneapolis and St. Paul Metropolitan Area, this study develops four models to compare the effects of the built environment around transit stops on walking distance of transit egress. Job density is negatively correlated with walking distance, consistent in all four models. Other built environment variables exhibit different impacts by stop location. Particularly, land use mix has positive impacts on walking distance for stops outside of downtown and suburban employment centers whereas job density is more important for suburban centers. Job accessibility and the number of intersections have significant effects on stops within downtown areas but have no significant impacts on stops outside of downtown areas. The number of transit stops has opposite impacts on walking distance for stops within and outside of downtown. Moreover, the built environment tends to have a larger impact on walking distance in downtown areas than non-downtown areas. We then discuss the implications for stop area land use planning and transit stop location choice.
1. Introduction Public transportation plays an important role in providing access to diverse opportunities, activities, and services. It can also help reduce the growth of traffic congestion and improve air quality. To enjoy these benefits, it is important to enhance transit accessibility and encourage transit use. As a key access/egress mode, walking distance to/from transit stops/stations (called stops for simplicity) greatly influences individuals' use of transit services (Loutzenheiser, 1997; Zhao et al., 2003). The more close people live and/or work to transit stops, the more likely transit services are used (Murray et al., 1998). Furthermore, walking distance to transit stops is often used to define stop catchment areas, which are fundamental for evaluating land use impacts of transit infrastructure and designing policies for transit-oriented development (TOD). This study aims to offer researchers, transit planners, and policymakers a better understanding of built environment characteristics affecting walking distance of transit users at their destination-ends, and provide implications for stop area land use planning and the siting of transit stops. A full transit trip consists of at least three segments: an access segment from origins to transit stops, an in-vehicle segment, and an egress segment from transit stops to final destinations. This study explores
⁎
built environment correlates of walking distance of the egress segments between transit stops and non-home destinations. Using the 2010 Transit Onboard Survey in the Minneapolis-St. Paul Metropolitan area (Twin Cities), it aims to answer the following two questions: 1) how does the built environment around transit stops affect walking distance of transit egress? 2) Do these impacts differ between stops within and outside of downtown areas? The paper extends the research on walking distance of transit users in two ways. First, previous studies on walking distance to/from transit stops either focus on the access from home to transit stops and overlook the egress from stops to destinations, or do not distinguish them in data analysis. Planning implications can be different between origin stops and destination stops. This study specifically examines transit egress at the non-home ends. Second, this study differentiates the impacts of built environment characteristics on walking distance to stops within and outside of downtown areas, which carries different policy implications for traditional downtown-oriented transit systems and multidestination transit systems. Taken together, we aim to offer planning implications from the following two aspects. First, from the perspective of stop area planning, we want to identify built environment characteristics that are positively associated with the observed walking distance of transit users. This will inform land use planners how to
Corresponding author. E-mail addresses: [email protected] (J. Wang), [email protected] (X. Cao).
http://dx.doi.org/10.1016/j.jtrangeo.2017.08.013 Received 30 November 2016; Received in revised form 7 May 2017; Accepted 23 August 2017 0966-6923/ © 2017 Elsevier Ltd. All rights reserved.
Journal of Transport Geography 64 (2017) 132–138
J. Wang, X. Cao
However, few studies have focused on transit egress from transit stops to final destinations and built environment correlates of transit egress. Egress travel of transit users at the destination-ends is very important for a transit trip. Given the hypothesis of travel time budget, egress travel distance/time to destinations plays an important role in determining the choice of transit (Loutzenheiser, 1997). Moreover, transit users often have multiple choices to access transit stops at the home-ends; they can park & ride, kiss & ride, bike, or walk to transit stops. However, at the destination-ends, walking is the only choice for most transit users. Therefore, any walking barriers for transit egress may deter transit users from taking transit. Understanding built environment correlates of transit egress is critical for the siting of transit stops and stop area planning at the destinations. Among few studies on transit egress, Townsend and Zacharias (2010) showed that the destination type, a proxy for land use and activity, is the only variable significantly correlated with walking distance of transit egress. After studying subway commuters' egress in Downtown Boston, Guo (2009) concluded that improved path environment increases the utility of walking and possibly increases transit riders' willingness to walk longer. Walking distance to transit stops differs between downtown and non-downtown areas. O'Sullivan and Morrall (1996) found that the walking distance to CBD LRT stations is much shorter than that to suburban LRT stations. Presumably, built environment attributes in downtown and non-downtown areas contribute to the difference in walking distance. For example, destinations tend to be closer in downtown areas and hence walking distance is generally shorter. Alshalalfah and Shalaby (2007) also found that the dense transit network in Downtown Toronto, Canada, makes walking distance of access shorter, compared to other areas of the city. Furthermore, many metropolitan areas have experienced job suburbanization and multiple employment centers have emerged in suburban areas. Accordingly, transit planners are interested in the following two questions: How far do transit users walk from transit stops to non-downtown destinations, particularly destinations located within suburban employment centers? What factors influence the walking distance? These questions call for an investigation of the correlates of walking distance by differentiating stops within and outside of downtown areas. The answers also have implications for transit planning of grid transit systems. Traditional radial-line transit systems are oriented to serve the CBD, which is often characterized as pedestrian-friendly areas with high density, mixed land use, good sidewalks, and so on. However, some metropolitan areas (such as Phoenix and Las Vegas) without a strong CBD deploy a grid transit system to serve a dispersed array of travel destinations (Brown and Thompson, 2008), which vary greatly in pedestrian environments. The study fills the two gaps in the literature and extends the stream of these studies by examining how built environment characteristics around destination-end transit stops influence walking distance of transit egress and comparing the influences between stops in downtown areas, and non-downtown areas.
encourage transit riders to walk longer. Second, from the perspective of stop location choice, we want to identify built environment characteristics that tend to shorten riders' walking distance. Accordingly, transit planners could locate transit stops in the places with these built environment attributes. Thus, stop area planning should encourage riders walking longer while stop location choice should minimize riders' walking distance. This paper is organized as follows. The next section reviews the literature on walking behavior associated with transit trips. Section 3 describes study area, the data and methodology. Section 4 presents modeling results. The last section summarizes key findings and discusses policy implications. 2. Literature review Previous studies have explored pedestrian access to transit stops extensively because walking is a primary access/egress mode of transit (e.g. Hsiao et al., 1997). Many studies investigated the correlates of the propensity of walking to transit stops. Demographic characteristics of transit users (such as gender, ethnicity, age, income, having a driver's license, and so on) affect walking mode choice (Loutzenheiser, 1997; Kim et al., 2007). From a planning perspective, access mode choice of transit users is influenced by stop-area built environment characteristics, including distance to transit stops (e.g. Chalermpong and Wibowo, 2007), employment and residential density (Loutzenheiser, 1997; Cervero, 2001), land use mix (Cervero, 2001), parking availability (Loutzenheiser, 1997; Cervero, 2001), sidewalk and street network (Maghelal, 2011), and pedestrian path characteristics such as the numbers of ascending steps, road crossings, and traffic conflicts (Olszewski and Wibowo, 2005). Furthermore, when transit users choose walking routes, they often prioritize walking time or distance to transit stops, as well as safety (Weinstein Agrawal et al., 2008). Transit planners generally define transit catchment areas as a quarter-mile (400 m) for bus stops and half a mile (800 m) for rail stations (Hsiao et al., 1997; Zhao et al., 2003; Gutiérrez and GarcíaPalomares, 2008). The catchment areas are often used for ridership prediction and economic impact assessment. A number of empirical studies have questioned the accuracy and appropriateness of these “rules of thumb” and incorporated various factors to explain the variation of walking distance to transit stops. They found that walking distance to transit stops is influenced by transit attributes, trip characteristics and demographics of transit users. For example, walking distance is positively associated with transit services with high frequency and short waiting time (O'Sullivan and Morrall, 1996; Alshalalfah and Shalaby, 2007). The number of transfers has a negative association with walking distance whereas total trip length is positively associated with walking distance (El-Geneidy et al., 2014). Transit users' demographic characteristics, such as gender, age, income, and the number of vehicles, are also important determinants of walking distance (Loutzenheiser, 1997; García-Palomares et al., 2013; El-Geneidy et al., 2014; Chia et al., 2016). Some studies have examined the impacts of built environment characteristics around transit stops on walking distance because they are crucial for walking distance and transit use (Weinstein Agrawal et al., 2008). O'Sullivan and Morrall (1996) found that although the average walking distance to LRT stations in suburban areas of Calgary is 444 m, users of a suburban LRT station with a pedestrian-friendly environment walk 1.1 km to the station on average. Furthermore, walking distance is found to be positively associated with population density (El-Geneidy et al., 2014; Jiang et al., 2012), intersection density (ElGeneidy et al., 2014), and sidewalk density (Maghelal, 2011). Jiang et al. (2012) concluded that transit users walk longer to BRT stations in Jinan, China, when the route environment is highly walkable. Overall, these studies have shown the impacts of built environment attributes on walking distance of transit access from home to transit stops and offered important implications for stop area planning.
3. Data and methdology 3.1. Study area The Minneapolis-St. Paul (Twin Cities) metropolitan area consists of seven counties. The area includes two central cities, Minneapolis, the economic center, and Saint Paul, the political center. When defining downtown Minneapolis and Saint Paul, we used 20 jobs per acre as the minimum threshold to select continuous blocks. The suburban employment centers are defined using the criteria of 10,000 jobs as the minimum threshold of total number of jobs and seven jobs/acre as the minimum threshold of job density (Fig. 1). 3.2. Data and variables This study used the 2010 Transit Onboard Survey administered by 133
Journal of Transport Geography 64 (2017) 132–138
J. Wang, X. Cao
Fig. 1. Study area. k
the Metropolitan Council (Metropolitan Planning Organization) in the Twin Cities. The region had an extensive transit network with 215 bus routes, a light rail transit (LRT) line, and a commuter rail line when the survey was taken. The survey captured the following information of transit trips: trip purposes, origin and destination addresses, access and egress modes, transit routes, and demographic characteristics and transit use patterns of individual transit users (Metropolitan Council, 2011). Because the commuter rail has only one station in the downtown area, it was removed from the analysis. This study aims to examine walking distance of transit egress. Again, a transit trip contains an access segment from the origin to the boarding stop, an in-vehicle segment, and an egress segment from the alighting stop to the destination. The segments before and after the invehicle segment are generally called transit access and egress, respectively. In this analysis, however, transit egress means the segment between transit stops and attraction locations of home-based trips, and the segment between transit stops and origins/destinations of non-homebased trips. Thus, we excluded transit access at the home-ends and access/egress by other modes of transport (such as automobile, bicycling, and so on). Walking distance between transit stops and non-home-ends, the dependent variable, was measured based on the shortest path in street network using ArcGIS. Some egress have very large walking distances, possibly because of the errors in the survey process. In this study, we excluded egress with a walking distance of more than 3009 m (the 90th percentile of walking distance). Independent variables include three categories (Table 1). Demographic and trip characteristics were measured in the onboard survey. Job accessibility was obtained from the Accessibility Observatory (http://access.umn.edu/). Population density, job density, the number of intersections, and the number of transit stops were derived from ArcGIS analysis and total trip length were derived from network analysis. Land use mix is measured using the entropy index, based on the following equation (Frank et al., 2005).
Land Use Entropy Index = − ∑ P iLn (P i )
Ln (k )
i=1
Pi: the percentage of each land use type i in the measurement area.k: the number of land use types. A total of 7077 transit egress are included in the study (Table 2): 4288 within downtown area and 2849 outside of downtown areas. Among the non-downtown egress segments, 828 occur in suburban areas and 520 are located within the suburban employment centers. Overall, the mean and 85th percentile walking distances are 494 m and 845 m. The mean and 85th percentile walking distances within downtown areas are 420 m and 683 m; the mean and 85th percentile walking distances within suburban employment centers are 571 m and 730 m; the mean and 85th percentile walking distance in other areas are 611 m and 1171 m. There are totally 19 LRT stations with three stations locating within downtown Minneapolis (US Bank stadium station and Target Field station are not within our defined downtown Minneapolis.). The data include 1100 transit egress segments from light rail stations, with 519 from light rail stations locating within downtown Minneapolis. In terms of transit service types, the mean and 85th percentile walking distances to bus stops are 464 m and 815 m, while the mean and 85th percentile walking distances to light rail stations are 657 m and 1092 m. 3.3. Methodology In the study, we adopted Poisson models with robust errors. Because the histogram of walking distance has a long right-hand tale, a loglinear regression or Poisson-family model should be employed. If the error is heteroskedastic, the estimates of log-linear models are biased and inconsistent (King, 1988; Silva and Tenreyro, 2006). By contrast, a Poisson regression with the robust error option is robust to heteroskedasticity. It also relaxes the Poisson assumption that treats the mean and variance of the dependent variable equal (Gould, 2011). Using 134
Journal of Transport Geography 64 (2017) 132–138
J. Wang, X. Cao
Table 1 Variable definition. Variables
Description
Walking distance
The distance (meter) measured using street network distance from non-home ends to transit stops
Trip characteristics Bus Total trip length (miles) # of Transfers Work-destination
A dummy variable indicating that the transit type is bus The distance measured using street network distance from trip origins to destinations Total numbers of transfers during trips A dummy variable which equals to one if the non-home end is workplace
Built environment attributes Downtown Population density Job density Job accessibility (1000) Land use entropy # of intersections # of transit stops
A dummy variable indicating that non-home-end transit stop is located in the Downtown Minneapolis or Downtown Saint Paul The number of persons per square meter in the block where a transit stop is located The number of jobs per square meter in the block where a transit stop is located The number of jobs within a 30-min walk from the block where a transit stop is located Job entropy index within half a mile of a transit stop The number of intersections within half a mile of a transit stop, a density measure The number of transit stops within half a mile of a destination, a density measure
Demographic characteristics Male White Black Young people Senior people Household income
A A A A A
dummy dummy dummy dummy dummy
variable variable variable variable variable
indicating indicating indicating indicating indicating
that that that that that
a a a a a
respondent respondent respondent respondent respondent
is is is is is
male a Caucasian an African American “under 18 years old” “over 65 years old”
0 if less than $15,000 1 if $15,000–$24,999 2 if $25,000–$34,999 3 if $35,000–$59,999 4 if $60,000–$94,999 5 if $95,000 or More A dummy variable indicating that a respondent has an access to a vehicle A dummy variable indicating that a respondent has a driver's license
Vehicle License
4. Results
Stata 14.0, we developed four models to examine the impacts of built environment attributes on walking distance to stops within downtown areas, stops within suburban employment centers, stops outside of downtown and suburban centers, and all stops. We controlled for transit attributes, demographics, and trip characteristics.
4.1. Estimation results The estimation results are presented in Table 3. Deviance R2 of the model for all stops (M1) is 0.106. The model for downtown stops (M2) has a better goodness of fit (Deviance R2 of 0.163). For non-downtown stops, the model for stops within suburban employment centers (M3)
Table 2 Descriptive statistics of variables.
Walk Distance Bus Total trip length # transfer Work-destination Population density Job density Job accessibility (1000) Land use entropy # of intersections # of transit stops Male White Black Young people Senior people Household income Vehicle License
All (N = 7077)
Downtown (N = 4228)
Suburban employment center (N = 520)
Non-downtown non suburban employment center (N = 2329)
Mean
Mean
Std. dev.
Mean
Std. Dev.
Mean
Std. dev.
437
571
459
611
618
0.33 6.80 0.26 0.28
0.375 8.96 0.58 0.66
0.48 4.91 0.64 0.47
0.89 7.20 0.33 0.52
0.31 6.06 0.56 0.50
0.0042 0.23 24.89 0.07 8 144
0.0002 0.004 14.97 0.55 57 52
0.001 0.006 9.39 0.25 35 63
0.002 0.018 52.10 0.688 84 86
0.003 0.042 50.21 0.139 22 83
0.49 0.39 0.25 0.08 0.16 1.28 0.40 0.26
0.49 0.72 0.13 0.01 0.04 2.52 0.34 0.68
0.50 0.45 0.34 0.11 0.19 1.65 0.47 0.47
0.42 0.66 0.17 0.04 0.03 2.48 0.45 0.71
0.49 0.47 0.37 0.20 0.17 1.70 0.50 0.45
Std. dev.
494 513 420 Trip characteristics 0.84 0.36 0.88 9.65 6.69 11.08 0.19 0.45 0.06 0.77 0.42 0.92 Built environment attributes 0.0012 0.0037 0.0011 0.12 0.20 0.18 105.27 60.91 145.66 0.68 0.12 0.69 106 29 124 144 133 188 Demographic characteristics 0.41 0.49 0.39 0.75 0.43 0.81 0.11 0.31 0.07 0.02 0.13 0.01 0.03 0.16 0.03 3.24 1.59 3.76 0.65 0.48 0.80 0.84 0.37 0.93
135
Journal of Transport Geography 64 (2017) 132–138
J. Wang, X. Cao
Table 3 Poisson models for walking distance of egress trips.
License Bus Total trip length Work-destination Job density Job accessibility Land use mix # of intersections # of transit stops Constant Deviance R2
All stops (M1) (N = 7077)
Downtowns (M2) (N = 4228)
Suburban employment centers (M3) (N = 520)
Non-downtown and non-suburban centers (M4) (N = 2329)
Coefficient
P > z
Coefficient
P > z
Coefficient
P > z
Coefficient
P > z
− 0.149** − 0.392** − 0.003* − 0.085** − 0.366** 0.003** 1.146** − 0.006** − 0.001** 6.481**
0.000 0.000 0.067 0.007 0.000 0.000 0.000 0.000 0.000 0.000 0.106
−0.184** −0.349** −0.005** −0.358** −0.124* 0.004**
0.002 0.000 0.029 0.000 0.063 0.000
−0.001** −0.020** 9.009**
0.000 0.000 0.000 0.163
− 0.164 0.114 0.0063 0.076 − 33.40** − 0.002 − 0.897 0.0004 0.0009* 6.378**
0.844 0.912 0.327 0.259 0.002 0.785 0.751 0.820 0.080 0.000 0.06
− 0.095** − 0.293** 0.012** 0.108** − 1.995** − 0.001 1.239** 0.0009 0.0005* 5.725**
0.057 0.000 0.000 0.010 0.000 0.163 0.000 0.442 0.051 0.000 0.104
** Significant at the 95% confidence level. * Significant at the 90% confidence level.
has a worse goodness of fit (Deviance R2 of 0.06) and the model for stops outside of suburban employment centers has a deviance R2 of 0.104 (M4). Overall, these models have mediocre goodness of fit. That is, a large proportion of variation in walking distance is explained by unobserved factors. Driver's license is the only significant demographic variable in three models (M1, M2 and M4). It is negatively associated with walking distance. Transit users without a driver's license are transit-dependent. Thus, they tend to walk a longer distance than choice riders. In general, demographics have very weak correlations with walking distance as all bivariate correlations are smaller than 0.10. Household income has a lot of missing values. Adding household income in the model will reduce sample size from 7077 to 6530. Thus, we did not include it in our model. White and Black are insignificant in all models but M4. However, including them in the M4 makes the license variable insignificant. Thus, we decided not to include these two variables. In the models M1, M2 and M4, the dummy variable bus is negatively correlated with walking distance. Its negative coefficient indicates that transit users walk farther to light rail stations than to bus stops, consistent with previous studies (e.g. Daniels and Mulley, 2013; El-Geneidy et al., 2014). This is not surprising because light rail transit tends to have a larger stop spacing than bus. Note that we also developed a model for bus stops and a model for light rail stations, respectively, to examine whether the effects of land use variables differ between bus and rail transit. Although the effect sizes of these land use variables vary slightly, their sign and significance are consistent in the two models. Accordingly, we decide not to report the detailed model results in this paper. The effects of two trip characteristics on walking distance differ between downtown stops and stops outside of downtown and suburban centers. First, total trip length is positively correlated with walking distance in the M4. The larger the total trip length is, the smaller the share of walking distance of transit egress in the total trip is. Therefore, transit users are more likely to walk a longer distance when the total trip length is larger. This is consistent with previous studies (El-Geneidy et al., 2014). However, in the M2, total trip length has a negative correlation with walking distance of transit egress. In the sample, more than 90% of downtown transit users are commuters. Because suburban express services tend to have a larger total trip length and the siting of express stops is deliberated to serve very dense blocks, walking distance of transit egress tends to be shorter. Second, work-destination, a dummy variable for trip purpose, has a positive association with walking distance in the M4, but is negatively associated with walking distance in the M2. Commuters who use transit stops outside of downtown and suburban centers are more likely to be captive riders with limited choices. On the other hand, non-work trips are less
mandatory than commute trips, so individuals may choose different destinations or different means of transport if walking to destinations is too onerous. Therefore, it is plausible that commuters tend to walk longer to reach their destinations than non-commuters. By contrast, because the locations of stops within downtown areas are chosen mainly to serve commuters, it is reasonable that commuters tend to walk a shorter distance than non-commuters. Furthermore, these two variables are insignificant in the M3. After controlling for transit attributes, demographics and trip characteristics, job density is the only variable that is significant and has the same sign in all four models. It has a negative association with walking distance of transit egress. If a block has high job density, transit stops are likely to be located around the block. Accordingly, many commuters, as well as non-commuters, can walk short distances to reach their destinations. The number of jobs within a 30-minute walking distance (job accessibility) is positively associated with walking distance in the downtown model (M2). If jobs are a little far away from transit stops but are still within a walking distance, transit riders tend to walk longer to access those jobs. However, job accessibility is insignificant in the models for stops outside of downtown areas. It is worth noting that population density is insignificant for walking distance of transit egress, different from previous studies on walking distance of transit access (El-Geneidy et al., 2014; Jiang et al., 2012). In the M4, land use entropy is positively correlated with walking distance, suggesting that transit users are more likely to walk farther in a mixed-use environment outside of downtown and suburban centers. Combining with the literature (e.g. Ewing and Cervero, 2010), it seems that mixed-use around transit stops facilitates walking for both transit access and egress. However, land use entropy is insignificant for stops within suburban employment centers (M3). Because land use entropy within downtown areas has strong correlations with job accessibility and the number of intersections, we manually removed it from the downtown model (M2) because of multi-collinearity. The number of intersections within half a mile of transit stops has a negative association with walking distance in the downtown model (M2). The result differs from some studies on transit access (e.g. ElGeneidy et al., 2014). The number of intersections is an indicator of street connectivity. Streets with higher connectivity would provide more route options for pedestrians and reduce actual walk distance (e.g. Townsend and Zacharias, 2010). However, this variable is insignificant for stops outside of downtown areas. The number of transit stops is negatively correlated with walking distance in the downtown model (M2) but has a positive correlation in the two non-downtown models (M3 and M4). This variable measures the quality of transit service around destinations. In downtown areas, the higher the number of transit stops is, the more robust transit service 136
Journal of Transport Geography 64 (2017) 132–138
J. Wang, X. Cao
5. Conclusion
Table 4 Average elasticities of independent variables.
Licensea Busb Total trip length Work-destinationc Job density Job accessibility # of intersections # of transit stops
All stops (M1)
Downtowns (M2)
Suburban employment centers (M3)
Nondowntown and nonsuburban centers (M4)
− 0.17** − 0.29** − 0.06* − 0.07** − 0.07** 0.31** − 0.48** − 0.16**
− 0.17** − 0.30** − 0.06** − 0.30** − 0.02* 0.61** − 2.50** − 0.28**
−0.01 −0.01 0.05 0.08 −0.15** −0.04 −0.01 0.05*
− 0.11** − 0.26** 0.06** 0.14** − 0.05** 0.15 − 0.16 0.02*
This study examines the effects of the built environment around transit stops on walking distance of transit egress and develops four models to compare their effects between stops within and outside of downtown areas in the Twin Cities. We found that for transit egress, it is jobs rather than population that affect transit riders' walking distance, because population density is not significantly associated with walking distance, and that job density is negatively associated with walking distance in all four models. However, other built environment variables show different effects in the four models. In particular, the number of intersections has a significant and negative association with walking distance for stops within downtown areas while it is not significantly correlated with walking distance for stops outside of downtown areas. The number of transit stops has a positive association for stops within downtown areas but is negatively correlated for stops outside of downtown areas. Land use mix around transit stops has a significant and positive association with walking distance for stops outside of downtown and suburban centers but does not have any significant association with walking distance for stops within suburban employment centers. Moreover, these variables show different magnitudes of the impacts for stops within and outside of downtown areas. Specifically, the elasticity of intersection intensity is 2.5 for stops within downtown areas but is very small for stops outside of downtown areas. Job density and accessibility are inelastic. The effect of job accessibility on walking distance within downtown areas is much larger than that outside of downtown areas. Job density seems to have a larger effect within suburban employment centers than within downtown areas. This may reflect diminishing returns when job density is very high in downtown areas. Overall, the models yield interesting insights on stop area planning and the siting of transit stops. Downtown areas usually have a pedestrian-friendly environment while non-downtown areas have a great variation in pedestrian environments. Thus, land use policies should focus on transit stops outside of downtown areas, especially the stops located at secondary activity or job centers, to facilitate transit users' walking to their destinations. However, the different impacts of built environment elements outside of downtown areas on walking distance within suburban employment centers and outside of suburban employment centers provided different land use policy implications. Land use mix around transit stations outside of suburban employment centers has a positive association with walking distance. Therefore, land use policies should promote mixeduse development to encourage transit users to walk longer. On the other hand, job density improves transit users' access to transit stops and reduces their walking distance and the effect is larger within suburban employment centers. Thus, planners should reinforce job concentration in suburban centers. Moreover, offering more transit service also encourages transit users to walk farther, particularly in suburban employment centers. When choosing stop locations for new transit services within downtown areas, the priority should be given to blocks with good street connectivity, as the number of intersections has the largest effect on minimizing walking distance. This choice provides transit users a direct and convenient path to reach their destinations. Besides, job accessibility around stops within downtown areas has a positive association with walking distance. Therefore, transit stops should be located at the job-rich areas of the CBD to increase users' walking distance and serve more people by enlarging catchment areas. Planners should also locate transit stops around the blocks with high job density although its effect on walking distance reduction is not large.
** Significant at the 95% confidence level in Table 3. * Significant at the 90% confidence level in Table 3. a Proportion reduction in walking distance relative to people without license. b Proportion reduction in walking distance relative to light rail stations. c Proportion reduction in walking distance relative to non-work destination.
around destinations is. The presence of multiple transit alternatives allows transit users to choose the stop closer to their destinations. This finding is consistent with Alshalalfah and Shalaby (2007). However, in the non-downtown areas, people will use public transit to reach destinations served by fewer transit alternatives only if the actual walking distance is acceptable. Thus, the positive association makes sense in the non-downtown models. Taken together, the effects of land use variables on walking distance vary by stop location. Job density and the number of transit stops are the only significant variables in the model for suburban employment centers. Land use mix is significant in the model for stops outside of downtown and suburban centers. In addition, job accessibility and the number of intersections are significant in the downtown model. Furthermore, the number of transit stops has opposite effects in the downtown and non-downtown models. If we do not differentiate stop location, we will end up with a general model M1, which is mostly consistent with the downtown model M2. Accordingly, we are likely to misinform planning practice for stops outside of downtown areas.
4.2. Average elasticity To understand practical significance of built environment variables and make models comparable, we used a consistent model specification for the four models (by excluding land use entropy) and calculated their elasticities (Table 4). Generally, built environment characteristics tend to have effects equivalent to other variables, and their effects within downtown areas seem to be larger than those outside of downtown areas. Job density has a small effect on walking distance for stops within downtown areas and for stops outside of downtown and suburban centers. However, its effect on walking distance for stops within suburban employment centers is much larger. Particularly, a 1% increase in job density around transit stops shortens walking distance by 0.15%. Job accessibility has a larger effect on walking distance of transit egress than job density for stops within downtown areas. In particular, a 1% increase in job accessibility around transit stops increases walking distance by 0.61%. The number of intersections has the largest effect size for downtown stops (M2): increasing street connectivity by 1% around downtown stops reduces transit users' walking distance to reach their destinations by 2.5%. Furthermore, a 1% increase in the number of transit stops decreases walking distance by 0.28% for stops within downtown areas, whereas its effect size reduces to 0.05% for stops within suburban employment centers and 0.02% outside of downtown and suburban centers.
References Alshalalfah, B.W., Shalaby, A.S., 2007. Case study: relationship of walk access distance to transit with service, travel, and personal characteristics. J. Urban Plan. Dev. 133 (2), 114–118.
137
Journal of Transport Geography 64 (2017) 132–138
J. Wang, X. Cao
Hsiao, S., Lu, J., Sterling, J., Weatherford, M., 1997. Use of geographic information system for analysis of transit pedestrian access. Transp. Res. Rec. 1604, 50–59. Jiang, Y., Zegras, P.C., Mehndiratta, S., 2012. Walk the line: station context, corridor type and bus rapid transit walk access in Jinan, China. J. Transp. Geogr. 20 (1), 1–14. Kim, S., Ulfarsson, G.F., Hennessy, J.T., 2007. Analysis of light rail rider travel behavior: impacts of individual, built environment, and crime characteristics on transit access. Transp. Res. A Policy Pract. 41 (6), 511–522. King, G., 1988. Statistical models for political science event counts: bias in conventional procedures and evidence for the exponential Poisson regression model. Am. J. Polit. Sci. 838–863. Loutzenheiser, D., 1997. Pedestrian access to transit: model of walk trips and their design and urban form determinants around Bay area rapid transit stations. Transp. Res. Rec. 1604, 40–49. Maghelal, P.K., 2011. Walking to transit: influence of built environment at varying distances. Institute of transportation engineers. ITE J. 81 (2), 38. Metropolitan Council, 2011. Metropolitan Council Travel Behavior Inventory Transit Onboard Survey., Cambridge Systematics, Inc. http://www.metrocouncil.org/ Transportation/Publications-And-Resources/TBI-OnBoard-Survey-Final-Report.aspx. Murray, A.T., Davis, R., Stimson, R.J., Ferreira, L., 1998. Public transportation access. Transp. Res. Part D: Transp. Environ. 3 (5), 319–328. Olszewski, P., Wibowo, S., 2005. Using equivalent walking distance to assess pedestrian accessibility to transit stations in Singapore. Transp. Res. Rec. 1927, 38–45. O'Sullivan, S., Morrall, J., 1996. Walking distances to and from light-rail transit stations. Transp. Res. Rec. 1538, 19–26. Silva, S.J.M.C., Tenreyro, S., 2006. The log of gravity. Rev. Econ. Stat. 88, 641–658. Townsend, C., Zacharias, J., 2010. Built environment and pedestrian behavior at rail rapid transit stations in Bangkok. Transportation 37 (2), 317–330. Weinstein Agrawal, A., Schlossberg, M., Irvin, K., 2008. How far, by which route and why? A spatial analysis of pedestrian preference. J. Urban Des. 13 (1), 81–98. Zhao, F., Chow, L.F., Li, M.T., Ubaka, I., Gan, A., 2003. Forecasting transit walk accessibility: regression model alternative to buffer method. Transp. Res. Rec. 1835, 34–41.
Brown, J.R., Thompson, G.L., 2008. Examining the influence of multi-destination service orientation on transit service productivity: a multivariate analysis. Transportation 35 (2), 237–252. Cervero, R., 2001. Walk-and-ride: factors influencing pedestrian access to transit. J. Publ. Transp. 7 (3). Chalermpong, S., Wibowo, S.S., 2007. Transit Station Access Trips and Factors Affecting Propensity to Walk to Transit Stations in Bangkok, Thailand. In Proceedings of the Eastern Asia Society for Transportation Studies. The 7th International Conference of Eastern Asia Society for Transportation Studies (Vol. 2007, No. 0, pp. 232-232). Eastern Asia Society for Transportation Studies. Chia, J., Lee, J., Kamruzzaman, M., 2016. Walking to public transit: exploring variations by socioeconomic status. Int. J. Sustain. Transp. 10 (9), 805–814. Daniels, R., Mulley, C., 2013. Explaining walking distance to public transport: the dominance of public transport supply. J. Transp. Land Use 6 (2), 5–20. El-Geneidy, A., Grimsrud, M., Wasfi, R., Tétreault, P., Surprenant-Legault, J., 2014. New evidence on walking distances to transit stops: identifying redundancies and gaps using variable service areas. Transportation 41 (1), 193–210. Ewing, R., Cervero, R., 2010. Travel and the built environment: a meta-analysis. J. Am. Plan. Assoc. 76 (3), 265–294. Frank, L.D., Schmid, T.L., Sallis, J.F., Chapman, J., Saelens, B.E., 2005. Linking objectively measured physical activity with objectively measured urban form: findings from SMARTRAQ. Am. J. Prev. Med. 28 (2), 117–125. García-Palomares, J.C., Gutiérrez, J., Cardozo, O.D., 2013. Walking accessibility to public transport: an analysis based on microdata and GIS. Environ. Plan. Plan. Des. 40 (6), 1087–1102. Gould, W., 2011. Use Poisson Rather than Regress; Tell a Friend. Stata Blog. http://blog. stata.com/2011/08/22/use-poisson-rather-than-regress-tell-a-friend/. Guo, Z., 2009. Does the pedestrian environment affect the utility of walking? A case of path choice in downtown Boston. Transp. Res. Part D: Transp. Environ. 14 (5), 343–352. Gutiérrez, J., García-Palomares, J.C., 2008. Distance-measure impacts on the calculation of transport service areas using GIS. Environ. Plan. Plan. Des. 35 (3), 480–503.
138
Contents lists available at ScienceDirect
Journal of Transport Geography journal homepage: www.elsevier.com/locate/jtrangeo
Exploring built environment correlates of walking distance of transit egress in the Twin Cities
MARK
Jueyu Wanga,⁎, Xinyu Caob a b
Humphrey School of Public Affairs, University of Minnesota, Twin Cities., 301 19th Ave S, Minneapolis, MN 55455, United States Humphrey School of Public Affairs, University of Minnesota, Twin Cities., 301 19th Ave S, Minneapolis, MN 55455, United States
A R T I C L E I N F O
A B S T R A C T
Keywords: Station area planning Land use Pedestrian behavior Transit planning Transit-oriented development
Most studies on walking distance to transit stops either emphasize transit access or do not distinguish transit access and egress. Furthermore, environmental correlates of walking distance may differ by stop location. Using the 2010 Transit Onboard Survey in the Minneapolis and St. Paul Metropolitan Area, this study develops four models to compare the effects of the built environment around transit stops on walking distance of transit egress. Job density is negatively correlated with walking distance, consistent in all four models. Other built environment variables exhibit different impacts by stop location. Particularly, land use mix has positive impacts on walking distance for stops outside of downtown and suburban employment centers whereas job density is more important for suburban centers. Job accessibility and the number of intersections have significant effects on stops within downtown areas but have no significant impacts on stops outside of downtown areas. The number of transit stops has opposite impacts on walking distance for stops within and outside of downtown. Moreover, the built environment tends to have a larger impact on walking distance in downtown areas than non-downtown areas. We then discuss the implications for stop area land use planning and transit stop location choice.
1. Introduction Public transportation plays an important role in providing access to diverse opportunities, activities, and services. It can also help reduce the growth of traffic congestion and improve air quality. To enjoy these benefits, it is important to enhance transit accessibility and encourage transit use. As a key access/egress mode, walking distance to/from transit stops/stations (called stops for simplicity) greatly influences individuals' use of transit services (Loutzenheiser, 1997; Zhao et al., 2003). The more close people live and/or work to transit stops, the more likely transit services are used (Murray et al., 1998). Furthermore, walking distance to transit stops is often used to define stop catchment areas, which are fundamental for evaluating land use impacts of transit infrastructure and designing policies for transit-oriented development (TOD). This study aims to offer researchers, transit planners, and policymakers a better understanding of built environment characteristics affecting walking distance of transit users at their destination-ends, and provide implications for stop area land use planning and the siting of transit stops. A full transit trip consists of at least three segments: an access segment from origins to transit stops, an in-vehicle segment, and an egress segment from transit stops to final destinations. This study explores
⁎
built environment correlates of walking distance of the egress segments between transit stops and non-home destinations. Using the 2010 Transit Onboard Survey in the Minneapolis-St. Paul Metropolitan area (Twin Cities), it aims to answer the following two questions: 1) how does the built environment around transit stops affect walking distance of transit egress? 2) Do these impacts differ between stops within and outside of downtown areas? The paper extends the research on walking distance of transit users in two ways. First, previous studies on walking distance to/from transit stops either focus on the access from home to transit stops and overlook the egress from stops to destinations, or do not distinguish them in data analysis. Planning implications can be different between origin stops and destination stops. This study specifically examines transit egress at the non-home ends. Second, this study differentiates the impacts of built environment characteristics on walking distance to stops within and outside of downtown areas, which carries different policy implications for traditional downtown-oriented transit systems and multidestination transit systems. Taken together, we aim to offer planning implications from the following two aspects. First, from the perspective of stop area planning, we want to identify built environment characteristics that are positively associated with the observed walking distance of transit users. This will inform land use planners how to
Corresponding author. E-mail addresses: [email protected] (J. Wang), [email protected] (X. Cao).
http://dx.doi.org/10.1016/j.jtrangeo.2017.08.013 Received 30 November 2016; Received in revised form 7 May 2017; Accepted 23 August 2017 0966-6923/ © 2017 Elsevier Ltd. All rights reserved.
Journal of Transport Geography 64 (2017) 132–138
J. Wang, X. Cao
However, few studies have focused on transit egress from transit stops to final destinations and built environment correlates of transit egress. Egress travel of transit users at the destination-ends is very important for a transit trip. Given the hypothesis of travel time budget, egress travel distance/time to destinations plays an important role in determining the choice of transit (Loutzenheiser, 1997). Moreover, transit users often have multiple choices to access transit stops at the home-ends; they can park & ride, kiss & ride, bike, or walk to transit stops. However, at the destination-ends, walking is the only choice for most transit users. Therefore, any walking barriers for transit egress may deter transit users from taking transit. Understanding built environment correlates of transit egress is critical for the siting of transit stops and stop area planning at the destinations. Among few studies on transit egress, Townsend and Zacharias (2010) showed that the destination type, a proxy for land use and activity, is the only variable significantly correlated with walking distance of transit egress. After studying subway commuters' egress in Downtown Boston, Guo (2009) concluded that improved path environment increases the utility of walking and possibly increases transit riders' willingness to walk longer. Walking distance to transit stops differs between downtown and non-downtown areas. O'Sullivan and Morrall (1996) found that the walking distance to CBD LRT stations is much shorter than that to suburban LRT stations. Presumably, built environment attributes in downtown and non-downtown areas contribute to the difference in walking distance. For example, destinations tend to be closer in downtown areas and hence walking distance is generally shorter. Alshalalfah and Shalaby (2007) also found that the dense transit network in Downtown Toronto, Canada, makes walking distance of access shorter, compared to other areas of the city. Furthermore, many metropolitan areas have experienced job suburbanization and multiple employment centers have emerged in suburban areas. Accordingly, transit planners are interested in the following two questions: How far do transit users walk from transit stops to non-downtown destinations, particularly destinations located within suburban employment centers? What factors influence the walking distance? These questions call for an investigation of the correlates of walking distance by differentiating stops within and outside of downtown areas. The answers also have implications for transit planning of grid transit systems. Traditional radial-line transit systems are oriented to serve the CBD, which is often characterized as pedestrian-friendly areas with high density, mixed land use, good sidewalks, and so on. However, some metropolitan areas (such as Phoenix and Las Vegas) without a strong CBD deploy a grid transit system to serve a dispersed array of travel destinations (Brown and Thompson, 2008), which vary greatly in pedestrian environments. The study fills the two gaps in the literature and extends the stream of these studies by examining how built environment characteristics around destination-end transit stops influence walking distance of transit egress and comparing the influences between stops in downtown areas, and non-downtown areas.
encourage transit riders to walk longer. Second, from the perspective of stop location choice, we want to identify built environment characteristics that tend to shorten riders' walking distance. Accordingly, transit planners could locate transit stops in the places with these built environment attributes. Thus, stop area planning should encourage riders walking longer while stop location choice should minimize riders' walking distance. This paper is organized as follows. The next section reviews the literature on walking behavior associated with transit trips. Section 3 describes study area, the data and methodology. Section 4 presents modeling results. The last section summarizes key findings and discusses policy implications. 2. Literature review Previous studies have explored pedestrian access to transit stops extensively because walking is a primary access/egress mode of transit (e.g. Hsiao et al., 1997). Many studies investigated the correlates of the propensity of walking to transit stops. Demographic characteristics of transit users (such as gender, ethnicity, age, income, having a driver's license, and so on) affect walking mode choice (Loutzenheiser, 1997; Kim et al., 2007). From a planning perspective, access mode choice of transit users is influenced by stop-area built environment characteristics, including distance to transit stops (e.g. Chalermpong and Wibowo, 2007), employment and residential density (Loutzenheiser, 1997; Cervero, 2001), land use mix (Cervero, 2001), parking availability (Loutzenheiser, 1997; Cervero, 2001), sidewalk and street network (Maghelal, 2011), and pedestrian path characteristics such as the numbers of ascending steps, road crossings, and traffic conflicts (Olszewski and Wibowo, 2005). Furthermore, when transit users choose walking routes, they often prioritize walking time or distance to transit stops, as well as safety (Weinstein Agrawal et al., 2008). Transit planners generally define transit catchment areas as a quarter-mile (400 m) for bus stops and half a mile (800 m) for rail stations (Hsiao et al., 1997; Zhao et al., 2003; Gutiérrez and GarcíaPalomares, 2008). The catchment areas are often used for ridership prediction and economic impact assessment. A number of empirical studies have questioned the accuracy and appropriateness of these “rules of thumb” and incorporated various factors to explain the variation of walking distance to transit stops. They found that walking distance to transit stops is influenced by transit attributes, trip characteristics and demographics of transit users. For example, walking distance is positively associated with transit services with high frequency and short waiting time (O'Sullivan and Morrall, 1996; Alshalalfah and Shalaby, 2007). The number of transfers has a negative association with walking distance whereas total trip length is positively associated with walking distance (El-Geneidy et al., 2014). Transit users' demographic characteristics, such as gender, age, income, and the number of vehicles, are also important determinants of walking distance (Loutzenheiser, 1997; García-Palomares et al., 2013; El-Geneidy et al., 2014; Chia et al., 2016). Some studies have examined the impacts of built environment characteristics around transit stops on walking distance because they are crucial for walking distance and transit use (Weinstein Agrawal et al., 2008). O'Sullivan and Morrall (1996) found that although the average walking distance to LRT stations in suburban areas of Calgary is 444 m, users of a suburban LRT station with a pedestrian-friendly environment walk 1.1 km to the station on average. Furthermore, walking distance is found to be positively associated with population density (El-Geneidy et al., 2014; Jiang et al., 2012), intersection density (ElGeneidy et al., 2014), and sidewalk density (Maghelal, 2011). Jiang et al. (2012) concluded that transit users walk longer to BRT stations in Jinan, China, when the route environment is highly walkable. Overall, these studies have shown the impacts of built environment attributes on walking distance of transit access from home to transit stops and offered important implications for stop area planning.
3. Data and methdology 3.1. Study area The Minneapolis-St. Paul (Twin Cities) metropolitan area consists of seven counties. The area includes two central cities, Minneapolis, the economic center, and Saint Paul, the political center. When defining downtown Minneapolis and Saint Paul, we used 20 jobs per acre as the minimum threshold to select continuous blocks. The suburban employment centers are defined using the criteria of 10,000 jobs as the minimum threshold of total number of jobs and seven jobs/acre as the minimum threshold of job density (Fig. 1). 3.2. Data and variables This study used the 2010 Transit Onboard Survey administered by 133
Journal of Transport Geography 64 (2017) 132–138
J. Wang, X. Cao
Fig. 1. Study area. k
the Metropolitan Council (Metropolitan Planning Organization) in the Twin Cities. The region had an extensive transit network with 215 bus routes, a light rail transit (LRT) line, and a commuter rail line when the survey was taken. The survey captured the following information of transit trips: trip purposes, origin and destination addresses, access and egress modes, transit routes, and demographic characteristics and transit use patterns of individual transit users (Metropolitan Council, 2011). Because the commuter rail has only one station in the downtown area, it was removed from the analysis. This study aims to examine walking distance of transit egress. Again, a transit trip contains an access segment from the origin to the boarding stop, an in-vehicle segment, and an egress segment from the alighting stop to the destination. The segments before and after the invehicle segment are generally called transit access and egress, respectively. In this analysis, however, transit egress means the segment between transit stops and attraction locations of home-based trips, and the segment between transit stops and origins/destinations of non-homebased trips. Thus, we excluded transit access at the home-ends and access/egress by other modes of transport (such as automobile, bicycling, and so on). Walking distance between transit stops and non-home-ends, the dependent variable, was measured based on the shortest path in street network using ArcGIS. Some egress have very large walking distances, possibly because of the errors in the survey process. In this study, we excluded egress with a walking distance of more than 3009 m (the 90th percentile of walking distance). Independent variables include three categories (Table 1). Demographic and trip characteristics were measured in the onboard survey. Job accessibility was obtained from the Accessibility Observatory (http://access.umn.edu/). Population density, job density, the number of intersections, and the number of transit stops were derived from ArcGIS analysis and total trip length were derived from network analysis. Land use mix is measured using the entropy index, based on the following equation (Frank et al., 2005).
Land Use Entropy Index = − ∑ P iLn (P i )
Ln (k )
i=1
Pi: the percentage of each land use type i in the measurement area.k: the number of land use types. A total of 7077 transit egress are included in the study (Table 2): 4288 within downtown area and 2849 outside of downtown areas. Among the non-downtown egress segments, 828 occur in suburban areas and 520 are located within the suburban employment centers. Overall, the mean and 85th percentile walking distances are 494 m and 845 m. The mean and 85th percentile walking distances within downtown areas are 420 m and 683 m; the mean and 85th percentile walking distances within suburban employment centers are 571 m and 730 m; the mean and 85th percentile walking distance in other areas are 611 m and 1171 m. There are totally 19 LRT stations with three stations locating within downtown Minneapolis (US Bank stadium station and Target Field station are not within our defined downtown Minneapolis.). The data include 1100 transit egress segments from light rail stations, with 519 from light rail stations locating within downtown Minneapolis. In terms of transit service types, the mean and 85th percentile walking distances to bus stops are 464 m and 815 m, while the mean and 85th percentile walking distances to light rail stations are 657 m and 1092 m. 3.3. Methodology In the study, we adopted Poisson models with robust errors. Because the histogram of walking distance has a long right-hand tale, a loglinear regression or Poisson-family model should be employed. If the error is heteroskedastic, the estimates of log-linear models are biased and inconsistent (King, 1988; Silva and Tenreyro, 2006). By contrast, a Poisson regression with the robust error option is robust to heteroskedasticity. It also relaxes the Poisson assumption that treats the mean and variance of the dependent variable equal (Gould, 2011). Using 134
Journal of Transport Geography 64 (2017) 132–138
J. Wang, X. Cao
Table 1 Variable definition. Variables
Description
Walking distance
The distance (meter) measured using street network distance from non-home ends to transit stops
Trip characteristics Bus Total trip length (miles) # of Transfers Work-destination
A dummy variable indicating that the transit type is bus The distance measured using street network distance from trip origins to destinations Total numbers of transfers during trips A dummy variable which equals to one if the non-home end is workplace
Built environment attributes Downtown Population density Job density Job accessibility (1000) Land use entropy # of intersections # of transit stops
A dummy variable indicating that non-home-end transit stop is located in the Downtown Minneapolis or Downtown Saint Paul The number of persons per square meter in the block where a transit stop is located The number of jobs per square meter in the block where a transit stop is located The number of jobs within a 30-min walk from the block where a transit stop is located Job entropy index within half a mile of a transit stop The number of intersections within half a mile of a transit stop, a density measure The number of transit stops within half a mile of a destination, a density measure
Demographic characteristics Male White Black Young people Senior people Household income
A A A A A
dummy dummy dummy dummy dummy
variable variable variable variable variable
indicating indicating indicating indicating indicating
that that that that that
a a a a a
respondent respondent respondent respondent respondent
is is is is is
male a Caucasian an African American “under 18 years old” “over 65 years old”
0 if less than $15,000 1 if $15,000–$24,999 2 if $25,000–$34,999 3 if $35,000–$59,999 4 if $60,000–$94,999 5 if $95,000 or More A dummy variable indicating that a respondent has an access to a vehicle A dummy variable indicating that a respondent has a driver's license
Vehicle License
4. Results
Stata 14.0, we developed four models to examine the impacts of built environment attributes on walking distance to stops within downtown areas, stops within suburban employment centers, stops outside of downtown and suburban centers, and all stops. We controlled for transit attributes, demographics, and trip characteristics.
4.1. Estimation results The estimation results are presented in Table 3. Deviance R2 of the model for all stops (M1) is 0.106. The model for downtown stops (M2) has a better goodness of fit (Deviance R2 of 0.163). For non-downtown stops, the model for stops within suburban employment centers (M3)
Table 2 Descriptive statistics of variables.
Walk Distance Bus Total trip length # transfer Work-destination Population density Job density Job accessibility (1000) Land use entropy # of intersections # of transit stops Male White Black Young people Senior people Household income Vehicle License
All (N = 7077)
Downtown (N = 4228)
Suburban employment center (N = 520)
Non-downtown non suburban employment center (N = 2329)
Mean
Mean
Std. dev.
Mean
Std. Dev.
Mean
Std. dev.
437
571
459
611
618
0.33 6.80 0.26 0.28
0.375 8.96 0.58 0.66
0.48 4.91 0.64 0.47
0.89 7.20 0.33 0.52
0.31 6.06 0.56 0.50
0.0042 0.23 24.89 0.07 8 144
0.0002 0.004 14.97 0.55 57 52
0.001 0.006 9.39 0.25 35 63
0.002 0.018 52.10 0.688 84 86
0.003 0.042 50.21 0.139 22 83
0.49 0.39 0.25 0.08 0.16 1.28 0.40 0.26
0.49 0.72 0.13 0.01 0.04 2.52 0.34 0.68
0.50 0.45 0.34 0.11 0.19 1.65 0.47 0.47
0.42 0.66 0.17 0.04 0.03 2.48 0.45 0.71
0.49 0.47 0.37 0.20 0.17 1.70 0.50 0.45
Std. dev.
494 513 420 Trip characteristics 0.84 0.36 0.88 9.65 6.69 11.08 0.19 0.45 0.06 0.77 0.42 0.92 Built environment attributes 0.0012 0.0037 0.0011 0.12 0.20 0.18 105.27 60.91 145.66 0.68 0.12 0.69 106 29 124 144 133 188 Demographic characteristics 0.41 0.49 0.39 0.75 0.43 0.81 0.11 0.31 0.07 0.02 0.13 0.01 0.03 0.16 0.03 3.24 1.59 3.76 0.65 0.48 0.80 0.84 0.37 0.93
135
Journal of Transport Geography 64 (2017) 132–138
J. Wang, X. Cao
Table 3 Poisson models for walking distance of egress trips.
License Bus Total trip length Work-destination Job density Job accessibility Land use mix # of intersections # of transit stops Constant Deviance R2
All stops (M1) (N = 7077)
Downtowns (M2) (N = 4228)
Suburban employment centers (M3) (N = 520)
Non-downtown and non-suburban centers (M4) (N = 2329)
Coefficient
P > z
Coefficient
P > z
Coefficient
P > z
Coefficient
P > z
− 0.149** − 0.392** − 0.003* − 0.085** − 0.366** 0.003** 1.146** − 0.006** − 0.001** 6.481**
0.000 0.000 0.067 0.007 0.000 0.000 0.000 0.000 0.000 0.000 0.106
−0.184** −0.349** −0.005** −0.358** −0.124* 0.004**
0.002 0.000 0.029 0.000 0.063 0.000
−0.001** −0.020** 9.009**
0.000 0.000 0.000 0.163
− 0.164 0.114 0.0063 0.076 − 33.40** − 0.002 − 0.897 0.0004 0.0009* 6.378**
0.844 0.912 0.327 0.259 0.002 0.785 0.751 0.820 0.080 0.000 0.06
− 0.095** − 0.293** 0.012** 0.108** − 1.995** − 0.001 1.239** 0.0009 0.0005* 5.725**
0.057 0.000 0.000 0.010 0.000 0.163 0.000 0.442 0.051 0.000 0.104
** Significant at the 95% confidence level. * Significant at the 90% confidence level.
has a worse goodness of fit (Deviance R2 of 0.06) and the model for stops outside of suburban employment centers has a deviance R2 of 0.104 (M4). Overall, these models have mediocre goodness of fit. That is, a large proportion of variation in walking distance is explained by unobserved factors. Driver's license is the only significant demographic variable in three models (M1, M2 and M4). It is negatively associated with walking distance. Transit users without a driver's license are transit-dependent. Thus, they tend to walk a longer distance than choice riders. In general, demographics have very weak correlations with walking distance as all bivariate correlations are smaller than 0.10. Household income has a lot of missing values. Adding household income in the model will reduce sample size from 7077 to 6530. Thus, we did not include it in our model. White and Black are insignificant in all models but M4. However, including them in the M4 makes the license variable insignificant. Thus, we decided not to include these two variables. In the models M1, M2 and M4, the dummy variable bus is negatively correlated with walking distance. Its negative coefficient indicates that transit users walk farther to light rail stations than to bus stops, consistent with previous studies (e.g. Daniels and Mulley, 2013; El-Geneidy et al., 2014). This is not surprising because light rail transit tends to have a larger stop spacing than bus. Note that we also developed a model for bus stops and a model for light rail stations, respectively, to examine whether the effects of land use variables differ between bus and rail transit. Although the effect sizes of these land use variables vary slightly, their sign and significance are consistent in the two models. Accordingly, we decide not to report the detailed model results in this paper. The effects of two trip characteristics on walking distance differ between downtown stops and stops outside of downtown and suburban centers. First, total trip length is positively correlated with walking distance in the M4. The larger the total trip length is, the smaller the share of walking distance of transit egress in the total trip is. Therefore, transit users are more likely to walk a longer distance when the total trip length is larger. This is consistent with previous studies (El-Geneidy et al., 2014). However, in the M2, total trip length has a negative correlation with walking distance of transit egress. In the sample, more than 90% of downtown transit users are commuters. Because suburban express services tend to have a larger total trip length and the siting of express stops is deliberated to serve very dense blocks, walking distance of transit egress tends to be shorter. Second, work-destination, a dummy variable for trip purpose, has a positive association with walking distance in the M4, but is negatively associated with walking distance in the M2. Commuters who use transit stops outside of downtown and suburban centers are more likely to be captive riders with limited choices. On the other hand, non-work trips are less
mandatory than commute trips, so individuals may choose different destinations or different means of transport if walking to destinations is too onerous. Therefore, it is plausible that commuters tend to walk longer to reach their destinations than non-commuters. By contrast, because the locations of stops within downtown areas are chosen mainly to serve commuters, it is reasonable that commuters tend to walk a shorter distance than non-commuters. Furthermore, these two variables are insignificant in the M3. After controlling for transit attributes, demographics and trip characteristics, job density is the only variable that is significant and has the same sign in all four models. It has a negative association with walking distance of transit egress. If a block has high job density, transit stops are likely to be located around the block. Accordingly, many commuters, as well as non-commuters, can walk short distances to reach their destinations. The number of jobs within a 30-minute walking distance (job accessibility) is positively associated with walking distance in the downtown model (M2). If jobs are a little far away from transit stops but are still within a walking distance, transit riders tend to walk longer to access those jobs. However, job accessibility is insignificant in the models for stops outside of downtown areas. It is worth noting that population density is insignificant for walking distance of transit egress, different from previous studies on walking distance of transit access (El-Geneidy et al., 2014; Jiang et al., 2012). In the M4, land use entropy is positively correlated with walking distance, suggesting that transit users are more likely to walk farther in a mixed-use environment outside of downtown and suburban centers. Combining with the literature (e.g. Ewing and Cervero, 2010), it seems that mixed-use around transit stops facilitates walking for both transit access and egress. However, land use entropy is insignificant for stops within suburban employment centers (M3). Because land use entropy within downtown areas has strong correlations with job accessibility and the number of intersections, we manually removed it from the downtown model (M2) because of multi-collinearity. The number of intersections within half a mile of transit stops has a negative association with walking distance in the downtown model (M2). The result differs from some studies on transit access (e.g. ElGeneidy et al., 2014). The number of intersections is an indicator of street connectivity. Streets with higher connectivity would provide more route options for pedestrians and reduce actual walk distance (e.g. Townsend and Zacharias, 2010). However, this variable is insignificant for stops outside of downtown areas. The number of transit stops is negatively correlated with walking distance in the downtown model (M2) but has a positive correlation in the two non-downtown models (M3 and M4). This variable measures the quality of transit service around destinations. In downtown areas, the higher the number of transit stops is, the more robust transit service 136
Journal of Transport Geography 64 (2017) 132–138
J. Wang, X. Cao
5. Conclusion
Table 4 Average elasticities of independent variables.
Licensea Busb Total trip length Work-destinationc Job density Job accessibility # of intersections # of transit stops
All stops (M1)
Downtowns (M2)
Suburban employment centers (M3)
Nondowntown and nonsuburban centers (M4)
− 0.17** − 0.29** − 0.06* − 0.07** − 0.07** 0.31** − 0.48** − 0.16**
− 0.17** − 0.30** − 0.06** − 0.30** − 0.02* 0.61** − 2.50** − 0.28**
−0.01 −0.01 0.05 0.08 −0.15** −0.04 −0.01 0.05*
− 0.11** − 0.26** 0.06** 0.14** − 0.05** 0.15 − 0.16 0.02*
This study examines the effects of the built environment around transit stops on walking distance of transit egress and develops four models to compare their effects between stops within and outside of downtown areas in the Twin Cities. We found that for transit egress, it is jobs rather than population that affect transit riders' walking distance, because population density is not significantly associated with walking distance, and that job density is negatively associated with walking distance in all four models. However, other built environment variables show different effects in the four models. In particular, the number of intersections has a significant and negative association with walking distance for stops within downtown areas while it is not significantly correlated with walking distance for stops outside of downtown areas. The number of transit stops has a positive association for stops within downtown areas but is negatively correlated for stops outside of downtown areas. Land use mix around transit stops has a significant and positive association with walking distance for stops outside of downtown and suburban centers but does not have any significant association with walking distance for stops within suburban employment centers. Moreover, these variables show different magnitudes of the impacts for stops within and outside of downtown areas. Specifically, the elasticity of intersection intensity is 2.5 for stops within downtown areas but is very small for stops outside of downtown areas. Job density and accessibility are inelastic. The effect of job accessibility on walking distance within downtown areas is much larger than that outside of downtown areas. Job density seems to have a larger effect within suburban employment centers than within downtown areas. This may reflect diminishing returns when job density is very high in downtown areas. Overall, the models yield interesting insights on stop area planning and the siting of transit stops. Downtown areas usually have a pedestrian-friendly environment while non-downtown areas have a great variation in pedestrian environments. Thus, land use policies should focus on transit stops outside of downtown areas, especially the stops located at secondary activity or job centers, to facilitate transit users' walking to their destinations. However, the different impacts of built environment elements outside of downtown areas on walking distance within suburban employment centers and outside of suburban employment centers provided different land use policy implications. Land use mix around transit stations outside of suburban employment centers has a positive association with walking distance. Therefore, land use policies should promote mixeduse development to encourage transit users to walk longer. On the other hand, job density improves transit users' access to transit stops and reduces their walking distance and the effect is larger within suburban employment centers. Thus, planners should reinforce job concentration in suburban centers. Moreover, offering more transit service also encourages transit users to walk farther, particularly in suburban employment centers. When choosing stop locations for new transit services within downtown areas, the priority should be given to blocks with good street connectivity, as the number of intersections has the largest effect on minimizing walking distance. This choice provides transit users a direct and convenient path to reach their destinations. Besides, job accessibility around stops within downtown areas has a positive association with walking distance. Therefore, transit stops should be located at the job-rich areas of the CBD to increase users' walking distance and serve more people by enlarging catchment areas. Planners should also locate transit stops around the blocks with high job density although its effect on walking distance reduction is not large.
** Significant at the 95% confidence level in Table 3. * Significant at the 90% confidence level in Table 3. a Proportion reduction in walking distance relative to people without license. b Proportion reduction in walking distance relative to light rail stations. c Proportion reduction in walking distance relative to non-work destination.
around destinations is. The presence of multiple transit alternatives allows transit users to choose the stop closer to their destinations. This finding is consistent with Alshalalfah and Shalaby (2007). However, in the non-downtown areas, people will use public transit to reach destinations served by fewer transit alternatives only if the actual walking distance is acceptable. Thus, the positive association makes sense in the non-downtown models. Taken together, the effects of land use variables on walking distance vary by stop location. Job density and the number of transit stops are the only significant variables in the model for suburban employment centers. Land use mix is significant in the model for stops outside of downtown and suburban centers. In addition, job accessibility and the number of intersections are significant in the downtown model. Furthermore, the number of transit stops has opposite effects in the downtown and non-downtown models. If we do not differentiate stop location, we will end up with a general model M1, which is mostly consistent with the downtown model M2. Accordingly, we are likely to misinform planning practice for stops outside of downtown areas.
4.2. Average elasticity To understand practical significance of built environment variables and make models comparable, we used a consistent model specification for the four models (by excluding land use entropy) and calculated their elasticities (Table 4). Generally, built environment characteristics tend to have effects equivalent to other variables, and their effects within downtown areas seem to be larger than those outside of downtown areas. Job density has a small effect on walking distance for stops within downtown areas and for stops outside of downtown and suburban centers. However, its effect on walking distance for stops within suburban employment centers is much larger. Particularly, a 1% increase in job density around transit stops shortens walking distance by 0.15%. Job accessibility has a larger effect on walking distance of transit egress than job density for stops within downtown areas. In particular, a 1% increase in job accessibility around transit stops increases walking distance by 0.61%. The number of intersections has the largest effect size for downtown stops (M2): increasing street connectivity by 1% around downtown stops reduces transit users' walking distance to reach their destinations by 2.5%. Furthermore, a 1% increase in the number of transit stops decreases walking distance by 0.28% for stops within downtown areas, whereas its effect size reduces to 0.05% for stops within suburban employment centers and 0.02% outside of downtown and suburban centers.
References Alshalalfah, B.W., Shalaby, A.S., 2007. Case study: relationship of walk access distance to transit with service, travel, and personal characteristics. J. Urban Plan. Dev. 133 (2), 114–118.
137
Journal of Transport Geography 64 (2017) 132–138
J. Wang, X. Cao
Hsiao, S., Lu, J., Sterling, J., Weatherford, M., 1997. Use of geographic information system for analysis of transit pedestrian access. Transp. Res. Rec. 1604, 50–59. Jiang, Y., Zegras, P.C., Mehndiratta, S., 2012. Walk the line: station context, corridor type and bus rapid transit walk access in Jinan, China. J. Transp. Geogr. 20 (1), 1–14. Kim, S., Ulfarsson, G.F., Hennessy, J.T., 2007. Analysis of light rail rider travel behavior: impacts of individual, built environment, and crime characteristics on transit access. Transp. Res. A Policy Pract. 41 (6), 511–522. King, G., 1988. Statistical models for political science event counts: bias in conventional procedures and evidence for the exponential Poisson regression model. Am. J. Polit. Sci. 838–863. Loutzenheiser, D., 1997. Pedestrian access to transit: model of walk trips and their design and urban form determinants around Bay area rapid transit stations. Transp. Res. Rec. 1604, 40–49. Maghelal, P.K., 2011. Walking to transit: influence of built environment at varying distances. Institute of transportation engineers. ITE J. 81 (2), 38. Metropolitan Council, 2011. Metropolitan Council Travel Behavior Inventory Transit Onboard Survey., Cambridge Systematics, Inc. http://www.metrocouncil.org/ Transportation/Publications-And-Resources/TBI-OnBoard-Survey-Final-Report.aspx. Murray, A.T., Davis, R., Stimson, R.J., Ferreira, L., 1998. Public transportation access. Transp. Res. Part D: Transp. Environ. 3 (5), 319–328. Olszewski, P., Wibowo, S., 2005. Using equivalent walking distance to assess pedestrian accessibility to transit stations in Singapore. Transp. Res. Rec. 1927, 38–45. O'Sullivan, S., Morrall, J., 1996. Walking distances to and from light-rail transit stations. Transp. Res. Rec. 1538, 19–26. Silva, S.J.M.C., Tenreyro, S., 2006. The log of gravity. Rev. Econ. Stat. 88, 641–658. Townsend, C., Zacharias, J., 2010. Built environment and pedestrian behavior at rail rapid transit stations in Bangkok. Transportation 37 (2), 317–330. Weinstein Agrawal, A., Schlossberg, M., Irvin, K., 2008. How far, by which route and why? A spatial analysis of pedestrian preference. J. Urban Des. 13 (1), 81–98. Zhao, F., Chow, L.F., Li, M.T., Ubaka, I., Gan, A., 2003. Forecasting transit walk accessibility: regression model alternative to buffer method. Transp. Res. Rec. 1835, 34–41.
Brown, J.R., Thompson, G.L., 2008. Examining the influence of multi-destination service orientation on transit service productivity: a multivariate analysis. Transportation 35 (2), 237–252. Cervero, R., 2001. Walk-and-ride: factors influencing pedestrian access to transit. J. Publ. Transp. 7 (3). Chalermpong, S., Wibowo, S.S., 2007. Transit Station Access Trips and Factors Affecting Propensity to Walk to Transit Stations in Bangkok, Thailand. In Proceedings of the Eastern Asia Society for Transportation Studies. The 7th International Conference of Eastern Asia Society for Transportation Studies (Vol. 2007, No. 0, pp. 232-232). Eastern Asia Society for Transportation Studies. Chia, J., Lee, J., Kamruzzaman, M., 2016. Walking to public transit: exploring variations by socioeconomic status. Int. J. Sustain. Transp. 10 (9), 805–814. Daniels, R., Mulley, C., 2013. Explaining walking distance to public transport: the dominance of public transport supply. J. Transp. Land Use 6 (2), 5–20. El-Geneidy, A., Grimsrud, M., Wasfi, R., Tétreault, P., Surprenant-Legault, J., 2014. New evidence on walking distances to transit stops: identifying redundancies and gaps using variable service areas. Transportation 41 (1), 193–210. Ewing, R., Cervero, R., 2010. Travel and the built environment: a meta-analysis. J. Am. Plan. Assoc. 76 (3), 265–294. Frank, L.D., Schmid, T.L., Sallis, J.F., Chapman, J., Saelens, B.E., 2005. Linking objectively measured physical activity with objectively measured urban form: findings from SMARTRAQ. Am. J. Prev. Med. 28 (2), 117–125. García-Palomares, J.C., Gutiérrez, J., Cardozo, O.D., 2013. Walking accessibility to public transport: an analysis based on microdata and GIS. Environ. Plan. Plan. Des. 40 (6), 1087–1102. Gould, W., 2011. Use Poisson Rather than Regress; Tell a Friend. Stata Blog. http://blog. stata.com/2011/08/22/use-poisson-rather-than-regress-tell-a-friend/. Guo, Z., 2009. Does the pedestrian environment affect the utility of walking? A case of path choice in downtown Boston. Transp. Res. Part D: Transp. Environ. 14 (5), 343–352. Gutiérrez, J., García-Palomares, J.C., 2008. Distance-measure impacts on the calculation of transport service areas using GIS. Environ. Plan. Plan. Des. 35 (3), 480–503.
138