Explaining the Railheading Travel Behaviour with Home Location, Park ‘N’ Ride Characteristics, and the Built Environment to Strengthen Multimodalism [pre-print]
Anthony Kimpton
24 September, 2020
[Please cite the version of this manuscript published in the Journal of Applied Spatial Analysis and Policy since this output is only intended to ensure research reproducibility. Likewise, the accompanying preprint text is included to provide context to the chunks of r script.]
library(Hmisc)## Loading required package: lattice## Loading required package: survival## Loading required package: Formula## Loading required package: ggplot2##
## Attaching package: 'Hmisc'## The following objects are masked from 'package:base':
##
## format.pval, unitslibrary(MASS)
library(sjPlot)
library(spdep)## Loading required package: sp## Loading required package: spData## To access larger datasets in this package, install the spDataLarge
## package with: `install.packages('spDataLarge',
## repos='https://nowosad.github.io/drat/', type='source')`## Loading required package: sf## Linking to GEOS 3.6.1, GDAL 2.2.3, PROJ 4.9.3library(knitr)
library(geosphere)## Warning: package 'geosphere' was built under R version 4.0.2library(rgdal)## rgdal: version: 1.4-8, (SVN revision 845)
## Geospatial Data Abstraction Library extensions to R successfully loaded
## Loaded GDAL runtime: GDAL 2.2.3, released 2017/11/20
## Path to GDAL shared files: D:/Users/uqakimpt/Documents/R/R-4.0.0/library/rgdal/gdal
## GDAL binary built with GEOS: TRUE
## Loaded PROJ.4 runtime: Rel. 4.9.3, 15 August 2016, [PJ_VERSION: 493]
## Path to PROJ.4 shared files: D:/Users/uqakimpt/Documents/R/R-4.0.0/library/rgdal/proj
## Linking to sp version: 1.4-1library(RColorBrewer)
library(grid)
library(lme4)## Loading required package: Matrixlibrary(janitor)##
## Attaching package: 'janitor'## The following objects are masked from 'package:stats':
##
## chisq.test, fisher.testlibrary(sf)
library(tidyverse)## -- Attaching packages ---------------------------------------------------------------------------- tidyverse 1.3.0 --## v tibble 3.0.1 v dplyr 0.8.5
## v tidyr 1.0.2 v stringr 1.4.0
## v readr 1.3.1 v forcats 0.5.0
## v purrr 0.3.4## -- Conflicts ------------------------------------------------------------------------------- tidyverse_conflicts() --
## x tidyr::expand() masks Matrix::expand()
## x dplyr::filter() masks stats::filter()
## x dplyr::lag() masks stats::lag()
## x tidyr::pack() masks Matrix::pack()
## x dplyr::select() masks MASS::select()
## x dplyr::src() masks Hmisc::src()
## x dplyr::summarize() masks Hmisc::summarize()
## x tidyr::unpack() masks Matrix::unpack()library(tmap)
library(tmaptools)
library(readxl)
library(GGally)## Registered S3 method overwritten by 'GGally':
## method from
## +.gg ggplot2##
## Attaching package: 'GGally'## The following object is masked from 'package:dplyr':
##
## nasalibrary(ggalluvial)
library(viridis)## Loading required package: viridisLiteoptions(scipen = 999, digits = 3)
opts_chunk$set(cache = FALSE, warning = FALSE, echo = FALSE, message = FALSE, include=TRUE, dpi = 600, out.width='100%', out.height = '100%')
#list.files()Abstract
Urban planning is transitioning away from the ‘Predict and Provide’ approach that accommodates automobility and towards the ‘Demand Management’ approach that prioritises alternatives that include active, shared, and public transport and restricts the convenience of automobility. While this transition could prove a sustainable solution for urban mobility, individuals already residing within auto-dependent settings may be unwilling or unable to relocate to high urban density where the alternatives are more viable. As such, restricting the automobility of these individuals potentially leaves them stranded throughout the urban form. The ‘Multimodalism’ approach is a pragmatic alternative that provides Park ‘n’ Rides, Kiss ‘n’ Rides, and feeder transit services that ensure everyone has access to rapid public transport yet the approach receives relatively little research attention. As such, researchers, policy makers, and planners are poorly equipped to influence intended multimodal travel behaviours or discourage the unintended such as ‘railheading’ towards more distantly located PnR. In this study, the transport planning and social psychology literature is examined to develop a conceptual model of travel behaviour, and for the first time, railheading behaviour is examined at the metropolitan-scale and explained using the conceptual model. The conceptual model and research findings strengthen the theoretical and empirical foundations for understanding travel behaviour, which in turn supports planning authorities and practitioners in promoting more sustainable transport behaviour, and in preparing for an urban future where Mobility-as-a-Service, ride-hailing, ride-sharing, eScooters, and autonomous vehicles become more integrated and commonplace.
Keywords
Multimodalism, Park‘n’ ride, Travel Behaviour, Parking, Transport, Landuse
Introduction
The Predict and Provide approach has dominated urban planning in developed countries for over fifty years and is regarded as contributing towards urban sprawl that can leave public transit impractical and private automobiles a necessity (i.e. auto-dependency; Goulden et al. 2014). Specifically, this approach entails predicting demand for road and parking capacity, and then providing accordingly. As a result, cities in more developed countries typically have sprawling urban forms. Given that the demand for faster, flexible, and unrestricted travel is potentially limitless, the demand for road and parking capacity is also potentially limitless thus the Demand Management approach is gaining popularity (Meyer 1999). Rather than accommodate motorists’ demands, the Demand Management approach intends to reduce the appeal of driving relative to alternatives such as active (e.g. walking and cycling), micro (e.g. eScooters), shared (e.g. ride hailing and ride sharing), or public transport (e.g. busses, ferries, and trains). Examples intended to reduce driving appeal include congestion pricing, parking pricing (e.g. parking meters), parking supply restrictions (e.g. parking maximums and unbundled parking), and Transit Orientated Development (TOD) that locates commuters within walkable range of rapid transit. While Demand Management could potentially provide a sustainable solution for urban mobility, it should not be regarded as a one-size-fits-all approach—at least in the short term—given that decades of Predict and Provide planning has shifted so many residents away from high density. As such, those unwilling or unable to relocate to within walking range of rapid transit or where public transit arrives at regular intervals have the potential to be left stranded in the urban periphery. The Multimodalism approach provides a middle ground by pragmatically accepting that private automobiles could be required to cover the ‘first mile’ of the public transit journey and so shifts the aim towards minimising the cost of transferring after the first mile from private automobiles to public transit (Turnbull 1995). Examples of this approach include feeder public transit services that deliver passengers to public transit terminals, Kiss ‘n’ Ride drop off bays, and Park `n’ Ride (PnR) parking facilities (Shaw and Walton 2001). Despite PnR emerging at least in the 1930s (Heggie and Papoulias 1976), motorists today still routinely drive past PnR with vacant parking spaces and therefore are choosing to not participate in multimodalism. The provider’s intent is clear, which is to maximise transit ridership and minimise the auto-network portion of the journey or gross Vehicle Miles Travelled (VTM) by locating facilities close to home, which together in turn minimise demand for road capacity (Arnott and Rowse 1999). In contrast, the passenger’s intent is relatively unclear given that their reasoned behaviour could include a complex decision hierarchy of criteria such as speed, affordability, comfort, reliability, or social prestige, and there could be further unreasoned behaviour such as habits and impulses (Van Acker, van Wee, and Witlox 2010). To better understand both the distinctions between provider and passenger intentions and between reasoned and unreasoned behaviour, this paper explores PnR facility choice and more specifically ‘rail-heading behaviour’ whereby passengers are choosing PnR other than their nearest PnR and therefore are defeating provider aims (Parkhurst and Meek 2014). The remainder of this paper is structured as follows. The next section explores the historical emergence of PnR, the transit provider and PnR user’s intention towards PnR, empirical findings predicting travel behaviour related to PnR use, and key psychological theories that can explain this travel behaviour. Section three outlines the study area, and describes the PnR, census, and revealed preference data, and the modelling approaches used to explain travel behaviour. Section four comprises the model results and section five concludes with the theoretical and practical implications for understanding railheading behaviour and multimodalism more broadly. Arguably, it is a critical period for understanding both of these dimensions so that researchers, officials, and practitioners can be better prepared for an urban future where ride-hailing, car-sharing, and autonomous vehicles become more commonplace (Sochor et al. 2015; Millard-Ball 2019), and where all become better integrated with public transit through Mobility as a Service (MaaS) applications and services (Kane and Whitehead 2018; Zhou et al. 2019).
Park ‘n’ Rides to Encourage Multimodalism
Detroit, MI was early to provide formal PnR during the 1930s at outer-city train stations. These PnR were provided so that train passengers that were parking outside stations (i.e. ‘fringe parking’) had dedicated off street parking and thus no longer exhausting the local parking supply and generating local social conflict (Heggie and Papoulias 1976; Noel 1988). As such, PnR were initially intended to alleviate social conflict due to fringe parking rather than encourage motorists to adopt multimodal travel behaviour (The Committee on Public Works 1970). PnR did not become commonplace throughout the US for several decades when the 1968 Federal Aid Highway Act enabled local and state planning authorities to request federal reimbursement for up to half the construction costs of their new PnR (Noel 1988). During this same period, PnR also began emerging overseas in countries such as the UK and Australia (Bureau of Roads 1969; Heggie and Papoulias 1976). Further, the intent shifted from accommodating fringe parkers towards encouraging multimodalism since this travel behaviour could reduce the transport network’s Vehicle Miles Travelled (VMT) by transferring motorists that share the same general route into shared public transit (Duncan and Christensen 2013). By reducing cumulative VMT, this would also reduce the cumulative carbon footprint of the city, and demands for road and parking capacity that become particularly expensive to provide as land values increase towards the inner city (Duncan and Cook 2014; Hounsell et al. 2011; Parkhurst 2000). ## The Provider’s Intent The transit provider’s intent could be summarised as ensuring that public transit is accessible to the largest possible population but this has become particularly challenging in modern times since automobility has reduced urban density (Kimpton et al. 2020b). For instance, origin-to-destination public transit services can reduce auto-dependency but providing enough services to emulate the convenience and flexibility of private automobiles may entail services operating at low capacity, which would likely prove financially infeasible. Likewise, origin-to-node public transit services that deliver passengers to Kiss `n’ Ride platforms so that they may ride rapid public transit to their final destination can reduce VTM but again emulating the convenience and flexibility of private automobiles would likely prove financially infeasible. As such, PnR that enable passengers to access rapid public transit by private automobile and Transit Oriented Development (TOD) that locates residents within walkable range of rapid public transit are two alternatives that continue to gain popularity (Parkhurst and Meek 2014; Singh, Lukman, Flacke, Zuidgeest, and Van Maarseveen 2017). The challenge as both approaches gain popularity is how PnR and TOD interact. For instance, when PnR and TOD are indistinguishable to motorists, TOD visitors may lose inclination to walk with PnR parking available thus undermining the demand management approach and also potentially undermining multimodalism if the TOD visitors exhaust or reduce the daily reliability of the PnR parking supply (Kimpton et al. 2020b). Likewise, motorists may fringe park throughout the TOD zone (e.g. informally park and ride) thus exhausting the already restricted TOD parking supply and generating social conflict with residents, traders, and service providers (i.e. parking overspill). Combining PnR and TOD is also challenging from an operational perspective given that placing PnR on cheap land minimises public overheads yet TOD typically raises nearby land values, which can leave PnR development and expansion prohibitively expensive (Mingardo 2013; Shirgaokar and Deakin 2005; Willson 2005; Dovey et al. 2014; Ginn 2009). Conversely, introducing PnR can reduce nearby residential land prices (Lieske et al 2019) thus suggesting that the housing market regards PnR as a local dis-amenity. Despite the global ubiquity of PnR, whether PnR are encouraging motorists to adopt multimodalism or are still just accommodating motorist that fringe park remains unclear (Christiansen et al. 2017; Dijk and Parkhurst 2014; Meek et al. 2009; Mingardo 2013; Parkhurst 2000; Zahabi et al. 2012). There is also evidence that PnR can induce additional trips to inner cities thus undermining outer city development and services (Meek, Ison, and Enoch 2009; Meek et al. 2008; Parkhurst 1995). Likewise, policy inequalities and development that provide free or under-priced parking at popular destinations such as the inner city can undermine the success of PnR thus metropolitan-scale coordination between parking supply and PnR is necessary to avoid policy inequalities and poor PnR performance (Hounsell et al. 2011). Game Theory is prominent theory within transportation research to explain the transit providers’ relationship with their current and potential ridership. Nash-type non-cooperative games are frequently employed to explain how individual passengers jointly determine their modal choices when there is a finite supply of transport elements such as road capacity, PnR parking, and transit capacity, and this understanding is typically modelled using traffic and transit simulations (Bell 2000). Through this lens, passengers are engaging in a zero-sum game and are assumed perfectly rational and informed agents that are seeking to minimise some specified criteria such as transfers, journey time, or driving portion of their journey which particularly aligns with the provider’s aims of reducing VMT. As such, modelling approaches can include Random Utility Maximisation models that seek the optimal outcome for the agent, and Random Regret Minimisation models that seek a good rather than optimal outcome thus improving computational tractability, and capture the human cognitive bias of ‘risk aversion’ where loss is perceived as outweighing an objectively commensurate or better gain thus leading to avoidance behaviour e.g. the regret of losing 10 dollars outweighs the elation of winning 20 thus the coin flip is never played (Sharma, Hickman, and Nassir 2019). Rather than regard the competition between passengers, Stockelberg-type non-cooperative games observe the metropolitan transport network where the zero-sum game is between providers and the travel demands of the metropolitan population (Fisk 1983; Sun and Gao 2006). For example, providers may seek to minimise road capacity demands by locating PnR closer to dwellings, and overheads by prioritising high residential density for transit coverage, repurposing unused PnR parking, and reducing service frequency for fewer vacant transit seats. While minimising overheads relative to ridership could ensure high operational efficiency, arguably maximising ridership is the actual game once the whole metropolitan transport network is considered. This is because maximising ridership minimises the public overheads of driving such as: increasing carbon emissions; exposing drivers, passengers, cyclists, and pedestrians to traffic collisions; and heightening demand for public roads and parking that also contribute to urban sprawl and further auto-dependency. The greater challenge is determining the passenger’s game objectives, or the voter’s for that matter given that roads, parking, and PnR can be tools for salving public disputes and winning votes, which has implications for providers (Huntley, 1993; Duncan and Christensen, 2013). ## The Passenger’s Intentions Relative to the provider’s intent, less is known about the breadth of passenger’s intentions. As such, the frequent focus of research is determining the optimal placement of PnR to hopefully persuade motorists to park and complete their journey by public transit (Kimpton et al. 2020b). Theory or simulation is typically the basis of optimal placement and will entail some specified criteria/objective from the passenger’s perspective such as the fastest or cheapest journey. Simulations also reveal what could be further criteria such as parking proximity to the platform given that this can minimise transfer times and therefore contribute towards a faster journey (Tsang et al. 2010). Notably, unlike theoretical or simulated agents, motorists lack perfect knowledge of the transport network and services, and likewise the intentions and behaviours of other motorists. As such, whether agents can approximate cognition and responses of motorists is questionable given that the ‘heuristic search patterns’ entail drawing from the imperfect knowledge gained through experience, and filtering this knowledge through various cognitive biases (Newell and Simon 1972). Indeed, compared to agents, motorists may subjectively weigh multiple criteria or carry grudges for the gambling, competing, and disappointment that public transit can entail that would be difficult to simulate (Kimpton et al. 2020b; Stieffenhofer et al. 2016). While both speed and cost are both relatively objective and therefore straightforward criteria to simulate, the weighting and comprehensiveness of these criteria remains difficult to determine and generally overlooked (Kimpton et al. 2020b). Intercept surveys are arguably better for clarifying and revealing further criteria that motorists/passengers perceive as influencing their travel behaviour. Such intercept surveys have revealed that PnR users perceive that parking capacity, platform facilities, nearby services, transit network connectivity, weather, travel purpose, time of the day, and social safety are criteria they consider when deciding whether to use PnR and which PnR (Bos et al. 2004; Frank and Pivo, 1994). Further research suggests that they intend to minimise diversions and double-backs to avoid introducing further loss when they encounter fully-occupied PnR; and confessions that they engage in ‘comfort seeking’ behaviour that entails ‘railheading’ past the nearest PnR to minimise the period within the relative discomfort of public transit (Parkhurst and Buxton 2005). As such, intercept surveys have proven effective in expanding the list of criteria but arguably remain poor at capturing the cognitive and affective biases that influence how travel attitudes and intentions manifest as travel behaviour, and are potentially limited by the ‘Hawthorne Effect’ whereby responses and observed behaviour are influenced the observer. Revealed preference surveys (e.g. licence plate surveys and gate counters) are the least common PnR empirical approach (Sharma, Hickman, and Nassir 2019) but have advantages including that it reveals actual travel behaviour without influencing the findings, and can reveal criteria unknown by who is being observed. Criteria that can predict PnR travel behaviour include: abundant parking, frequent services, minimal transit times (Debrezion et al. 2009), easy access and egress by automobiles (Wardman et al. 2007; Zhao et al. 2019), and proximity to the general driving route to avoid diverting or doubling back which extends the journey (Mahmoud, Habib, and Shalaby 2014; Spillar 1997). A PnR gate survey throughout Perth, Australia revealed that most PnR are fully occupied by 7:30am (Lin and Robinson 2014), which suggests that the typical worker starting at 9am would need to arrive at least 90 minutes prior to their shift to utilize PnR. Similar survey findings could explain why PnR parking fees and permits have been introduced throughout some cities to discourage PnR misuse, encourage car-pooling, and improve parking reliability for those willing or able to pay (Dijk and Parkhurst 2014; Meek et al. 2008; Stieffenhofer et al. 2015). Through a critical lens, fees and permits could be adding further financial burdens to already auto-dependent households. Indeed, even PnR with free parking have been criticised as exclusionary towards households that are financially and/or physically unable to drive (Duncan and Christensen 2013), which could explain why PnR users are typically from wealthier households (Foote 2000). ## Unintended Travel Behaviour Despite the breadth of approaches used to explain PnR travel behaviour, vast numbers of motorists worldwide continue to drive past vacant PnR each day and adopt travel behaviours unintended by providers such as ‘railheading’ which entails choosing PnR other than their nearest (Parkhurst and Meek 2014). Aside from undermining their provider’s attempts to minimise VMT and public expenses such as road capacity and parking, an inability to explain such unintended travel behaviours has the potential to constrain the success of multimodalism more broadly. As such, examining and explaining railheading is the focus for the remainder of this study. As discussed previously, the provider’s intent is to maximise transit ridership and minimise the VMT within the transport network while the passenger’s intentions have greater dimensionality given the range of criteria and ‘trade-offs’ to be considered when searching and selecting a PnR (Figure 1; Sharma et al. 2017). Assuming that passengers lack real time information about parking vacancy across their potential PnR, then their driving route presents PnR in a linear manner resembling an ‘optimal stopping problem’ that was first popularised by Flood as the ‘fiancé problem’ since suitors must decide whether they have: (a) just landed the best catch or (b) should throw it back and keep fishing even if a better catch never comes along (“the one that got away”; Feguson 1989). Wardrop (1952) introduced this optimal stopping problem to transport research and it has proven influential although it is increasingly framed as the ‘utility-maximization framework’ (Hamer 2010; Hendricks and Outwater 1998; Qin et al. 2013), which draws upon several key theories to explain travel behaviour. For instance, ‘Motivational Theory’ posits that individuals seek to maximise the utility gained by activities (Atkinson and Birch 1970; Axhausen and Garling 1992) and ‘Time Allocation Theory’ posits that individuals and households have ‘time allocation budgets’ that constrain the activities and behaviours (Becker 1965).
Figure 1. Railheading as unintended travel behaviour
Having explored the provider and passenger’s intentions and identified decision-making criteria, a theoretical framework is required to explain how intentions and criteria influence the passenger’s travel behaviour. Salomon and Ben-Akiva (1983) posit that individuals have a ‘hierarchical decision structure’ for selecting and ranking criteria for: short-term decisions such as travel mode or between available PnR on a daily basis; and long-term lifestyle decisions such as choosing whether to relocate within walking distance of public transit or where private automobile is a necessity (Van Acker et al. 2010). Lin, Wang, and Guan’s (2017) theoretical framing emphasises that these influences are directional whereby ‘residential determination’ occurs when the built environment influences travel attitudes and behaviours and ‘residential self-selection’ occurs when travel attitudes influence the built environment and travel behaviour. Further, ‘social equity’ has a role given that residential selection would be less influential if every location had similar access to infrastructure (e.g. PnR). Kimpton (2017) identifies that distinct operationalisations of social equity have emerged in research and policy that include: (1) ‘provision’ that ensures infrastructure is evenly dispersed throughout the transport network to minimise the behavioural influence of urban centrality and route direction (e.g. PnR and transit coverage); (2) ‘accessibility’ that ensures the cost of access and egress to infrastructure is similar throughout the transport network to minimise the behavioural influence of location and automobile ownership; and ‘population pressure’ that is the (3) anticipated demand for infrastructure and ensures that provision and accessibility are sensitive to residential concentrations that heighten demand for finite resources (e.g. PnR parking and transit seating). Van Acker, van Wee, and Witlox (2010) employ the Theory of Planned Behaviour to explain how travel attitudes influence travel behaviour. They posit that there are cognitive ‘reasoned influences’ whereby knowledge, experiences, and perceptions inform travel behaviour and affective ‘unreasoned influences’ whereby habits and impulses influence travel behaviour. Last, each conceptualisation just discussed highlights that there is socio-economic variability between all individuals and households that constrain where they work and can afford to reside (e.g. housing affordability), cultural expectations that inform attitudes (e.g. from an auto-centric culture), and behavioural constraints (e.g. mobility impairment or household budget for private automobility). Combined, these conceptualisations provide a conceptual framework of transport behaviour (Figure 2) for explaining specific behaviours such as railheading, which is the focus of the remaining sections.
Figure 2. A Conceptual Framework of Transport Behaviour (adapted from Lin, Wang, and Guan 2017; Kimpton 2017; and van Acker, van Wee, and Witlox 2010)
Data and Method
Study Frame
Brisbane is the third most populated Australian city ranked behind Sydney and Melbourne (i.e. 2,270,807, 4,823,993, and 4,485,210 respectively). Arguably, Brisbane is an exceptional study frame for metropolitan spanning transport and land use planning given that the Brisbane City Council population exceeds the City of Sydney and City of Melbourne inner-city local councils by a factor of 5 (i.e. 1,131,155, 135,964, and 208,376 respectively; ABS, 2016), and jurisdictions by a factor of 36 (i.e. 134,166ha, 2,674ha and 3,779ha respectively (Australian Bureau of Statistics, 2016). Located at the south-eastern coast of the state of Queensland (Figure 3), Brisbane’s earlier urban form was a linear-city where residents and industry were concentrated within the bends along the Brisbane River. While its contemporary urban form remains a linear-city, the rapidly expanding population—18 percent from 2006 to 2016—is fanning out along the t-shaped highway system that connects three neighbouring cities: Gold Coast, the Sunshine Coast, and Ipswich. Cross-river travel typically entails following one of multiple public ferry routes or crossing one of fifteen major bridges, however it is notable that seven of these bridges exclude private cars, and four exclude pedestrians and cyclists thus cross-river travel is particularly multimodal within Brisbane. Despite the difficulties of cross-river travel, 67 percent of the Greater Brisbane Statistical Division—the study frame—are remaining within their private automobiles for the entire journey and just eight percent are journeying by both private automobile and public transit, which suggests they are using one of the 135 PnR available or at least fringe parking (Australian Bureau of Statistics 2016). In sum, Brisbane has the hallmarks of a highly auto-dependent city but also a rare opportunity for transport and land use planning that is coordinated at the major metropolitan-level due to having an exceptionally large local government.
## OGR data source with driver: ESRI Shapefile
## Source: "D:\Projects\Anthony\Parking\Data\Australia\Census\2016_STE_shape", layer: "STE_2016_AUST"
## with 9 features
## It has 4 fields## OGR data source with driver: ESRI Shapefile
## Source: "D:\Projects\Anthony\Parking\Data\Australia\Census\2016_GCCSA_shape", layer: "GCCSA_2016_AUST"
## with 34 features
## It has 6 fields## OGR data source with driver: ESRI Shapefile
## Source: "D:\Projects\Anthony\Parking\Documents\Drafts\Applied Spatial Analysis and Policy", layer: "SA2_2016_AUST"
## with 2310 features
## It has 12 fields## OGR data source with driver: ESRI Shapefile
## Source: "D:\Projects\Anthony\Parking\Documents\Drafts\Applied Spatial Analysis and Policy", layer: "Baseline_roads_and_tracks"
## with 563747 features
## It has 7 fields
## Integer64 fields read as strings: ROADTYPE
Figure 3. Park ‘n’ Ride throughout the Greater Brisbane Statistical Division in Australia and the direct paths between PnR users and their suburb origins
Outcome Variables In the interest of reproducibility, all data cleaning, geo-processing, visualising, and statistical modelling is conducted using the open source R programming language and rStudio Integrated Development Environment, and the script is uploaded to a public repository (http://rpubs.com/[temporarily redacted]). For examining travel behaviour, the State Department of Transport and Main Roads collected revealed preference surveys in 2015, 2017, and 2018 that include the spatial coordinates of where PnR users were parked, whether this location is inside the PnR or overflow, and the suburb where the automobile is registered that is available upon request (TMR 2019b). Unfortunately change to railheading behaviour throughout the study period cannot be examined as the licence plates were redacted to preserve PnR user anonymity. TMR can also upon request provide the platform location of the 135 PnR and the parking bay counts for 2011 and 2016. To examine the built environment where PnR users reside and have chosen PnR, the Australian Bureau of Statistics’ (ABS 2016) 102 Statistical Area Two (SA2) neighbourhood units located throughout the Greater Brisbane statistical division, and their 2016 Census of Population and Housing are used. SA2 neighbourhood units are used since these are the smallest units available describing both dwellings and workplaces, and although SA2 have similar residential populations, it is notable that parking and riding is more common towards the inner city (Figure 4). The remaining data source is a road network that can be downloaded from the Queensland Government Open Data portal (TMR, 2019c).
Figure. 4 Multimodal travel, railheading behaviour and its distance aggregated to the suburb origin
##
## Moran I test under randomisation
##
## data: drop_na(temp, pnr.ers)$pnr.ers
## weights: lw n reduced by no-neighbour observations
##
##
## Moran I statistic standard deviate = 13, p-value <0.0000000000000002
## alternative hypothesis: greater
## sample estimates:
## Moran I statistic Expectation Variance
## 0.45775 -0.00345 0.00135##
## Moran I test under randomisation
##
## data: drop_na(temp, railheaded.percent)$railheaded.percent
## weights: lw n reduced by no-neighbour observations
##
##
## Moran I statistic standard deviate = 4, p-value = 0.0002
## alternative hypothesis: greater
## sample estimates:
## Moran I statistic Expectation Variance
## 0.12739 -0.00345 0.00139##
## Moran I test under randomisation
##
## data: drop_na(temp, railheading.km.average)$railheading.km.average
## weights: lw n reduced by no-neighbour observations
##
##
## Moran I statistic standard deviate = 24, p-value <0.0000000000000002
## alternative hypothesis: greater
## sample estimates:
## Moran I statistic Expectation Variance
## 0.89796 -0.00345 0.00141The railheaded outcome variable is coded ‘0’ for motorists that were observed using their nearest PnR—using straight-line distance—and ‘1’ if they were using another PnR (Figure 5), and has declined overall from 88 percent in 2015 to 83 in 2018 (Table 1). While journeys rather than neighbourhoods are the unit of analysis for this study, it is notable that railheading behaviour typically starts from neighbourhoods located further from PnR and Central Station (Figure 4) although the weak spatial auto-correlation between neighbourhoods suggests that start location has minimal influence over this travel behaviour (Global Moran’s I = 0.1; p < 0.001) and that a spatial model is unnecessary for explaining engagement in railheading behaviour. The railheading distance outcome variable is the kilometres that the chosen PnR is past the nearest PnR (Figure 5), which has declined from nine kilometres on average in 2015 to six in 2018. Further, railheading distance lessens towards the inner city and along the PnR network, and the standard deviation exceeds the mean for each period, which reveals over-dispersion and that a negative binomial model is more appropriate than a Poisson model for this outcome variable.
Figure 5. The Park `n’ Ride chosen from six suburb origins
Table 1. Data Summary
| var | year | n | percentage | mean | sd |
|---|---|---|---|---|---|
| railheaded1 | 2015 | 12808 | 88.5 | 0.88 | 0.32 |
| railheaded2 | 2017 | 17932 | 85.3 | 0.85 | 0.35 |
| railheaded3 | 2018 | 21540 | 83.0 | 0.83 | 0.38 |
| origin.centrality.km1 | 2015 | 12808 | 2517.6 | 25.18 | 16.82 |
| origin.centrality.km2 | 2017 | 17932 | 2389.6 | 23.90 | 15.69 |
| origin.centrality.km3 | 2018 | 21540 | 2305.9 | 23.06 | 14.91 |
| diversion.degrees1 | 2015 | 12808 | 3048.2 | 30.48 | 28.48 |
| diversion.degrees2 | 2017 | 17932 | 3358.2 | 33.58 | 29.70 |
| diversion.degrees3 | 2018 | 21540 | 3530.1 | 35.30 | 30.79 |
| parking.to.platform.m1 | 2015 | 12808 | 13315.8 | 133.16 | 82.00 |
| parking.to.platform.m2 | 2017 | 17932 | 14026.4 | 140.26 | 84.25 |
| parking.to.platform.m3 | 2018 | 21540 | 14601.7 | 146.02 | 87.35 |
| parked.outside1 | 2015 | 12808 | 26.8 | 0.27 | 0.44 |
| parked.outside2 | 2017 | 17932 | 31.6 | 0.32 | 0.46 |
| parked.outside3 | 2018 | 21540 | 35.9 | 0.36 | 0.48 |
| railheading.km1 | 2015 | 11330 | - | 9.47 | 12.20 |
| railheading.km2 | 2017 | 15305 | - | 7.35 | 10.54 |
| railheading.km3 | 2018 | 17873 | - | 6.18 | 9.19 |
| bays.change | - | 102 | - | 37.14 | 152.42 |
| transit.km | - | 102 | - | 16.14 | 10.06 |
| destination.road.m.per.ha | - | 102 | - | 98.13 | 36.94 |
| destination.land.use.diversity | - | 102 | - | 0.81 | 0.15 |
| destination.worker.density | - | 102 | - | 5.86 | 7.27 |
| destination.pow.private.auto.only | - | 102 | - | 84.80 | 5.13 |
| destination.dwelling.density | - | 102 | - | 5.08 | 3.57 |
| destination.average.hh.cars | - | 102 | - | 1.73 | 0.24 |
| destination.ur.private.auto.free | - | 102 | - | 17.60 | 9.89 |
Explanatory Variables
The TMR’s PnR and road network data provide the PnR characteristics and built environment, and the ABS’ 2016 Census of Population and Housing provides the characteristics of the workplaces and households, and Meshblock land use as described previously. Origin centrality is the straight-line distance between the suburb origin and Central Station that decreased from 25 kilometres on average in 2015 to 23 in 2018. Route diversion is the degrees between the direct route to Central Station and the chosen PnR from the suburb origin, which increased from 30 degrees in 2015 to 35 in 2018 thus willingness to divert travel to use PnR is rising. Transfer distance is the meters between where the automobile was parked and the PnR centroid, which increased from 133 meters in 2015 to 146 in 2018. Parking outside is coded ‘1’ when the automobile is located outside the PnR and ‘0’ when parked within, and this has increased from 27 percent in 2015 to 36 percent in 2018. Change to PnR parking is the 2016 PnR bay counts minus the 2011 counts, which on average increased by 37 bays throughout this period. Last among the spatial and PnR characteristics, transit distance is the straight-line kilometres between each PnR and Central Station, which on average is 16 kilometres. The remaining explanatory variables describe the neighbourhood context of the PnR. Destination road density is the meters of road network per hectare (mean = 98 meters) and destination land use diversity is an index revealing the probability that two randomly located points will not fall onto the same land use type (mean = 0.8; Blau 1977). Destination worker density is the number of people working within the same SA2 as the PnR divided by SA2 hectares given that this is a proxy for the degree to which a PnR is also a potential destination (mean = 6). Destination workers driving is the percent of workers that drove their automobiles to work (mean = 85%) thus revealing local competition for nearby parking between PnR users and nearby workers. Destination residential density is the number of dwellings per hectare in suburbs that contain PnR as a proxy for TOD (mean = 5). Likewise, destination residential cars is average cars per dwelling (mean = 1.73) and destination auto-free dwellings is the percent of dwellings without cars (mean = 18%) for identifying TOD zones where residents could potentially be competing for the on-street parking that often contains the PnR parking overspill.
Analytical Approach
A logistic regression model is used to explain the adoption of railheading travel behaviour as a binary outcome variable. Specifically, the R programming language is used (R Core Team 2019) and ‘glm’ function from the lme4 library (Bates et al. 2015), which has the following formula:
Where P is the probability that railheading will occur,β_0 the base of the logarithm, and 〖βx〗_i is a predictor. As discussed previously, railheading distance is over-dispersed thus a negative binomial regression is more appropriate than a Poisson model for explaining each kilometre of railheading distance according to the explanatory variables. The R programming language is again used however this time the ‘glm.nb’ function from the MASS library is used (Venables and Ripley 2013), which has the following formula:
where μ_i is railheading one kilometre, and exp(ln(t_i )+ 〖βx〗_i ) the exponential of a natural logarithm of t_i for each exposure or incident that is added to 〖βx〗_i for each predictor.
Results
Engagement in Railheading Behaviour
A binomial logistic regression model is employed to determine whether PnR users’ home origin, their route, and the characteristics and context of their chosen PnR can predict railheading behaviour (Table 2). While the model coefficients are Log-Odds, these findings have also been converted to Odds Ratios (O.R.) to simplify interpretation and order of magnitude that has been visualised as a Forrest plot (Figure 6).
| Logistic Model | ||
|---|---|---|
| Explanatory Variables | Log-Odds | s.e. |
| origin centrality [km] | 0.18 *** | 0.01 |
| diverting to PnR from destination [degrees] | -0.02 *** | 0.00 |
| PnR parking type [overflow] | -0.11 ** | 0.03 |
| PnR parking to platform [m] | 0.00 | 0.00 |
| PnR survey year [n] | -0.07 *** | 0.01 |
| PnR bays change [n] | -0.00 *** | 0.00 |
| PnR centrality [km] | -0.21 *** | 0.01 |
| road density near PnR [m per ha] | -0.01 *** | 0.00 |
| land use diversity near PnR [index] | -0.78 *** | 0.21 |
| worker density near PnR [per ha] | 0.05 *** | 0.00 |
| workers arriving by car near PnR [%] | -0.03 *** | 0.00 |
| dwelling density near PnR [per ha] | 0.10 *** | 0.01 |
| cars per dwelling near PnR [n] | 1.18 *** | 0.10 |
| auto-free households near PnR [%] | -0.00 | 0.00 |
| Observations | 52280 | |
| R2 Tjur | 0.199 | |
| log-Likelihood | -16919.504 | |
|
||
| Logistic Model | ||
|---|---|---|
| Explanatory Variables | Odds Ratios | s.e. |
| origin centrality [km] | 1.19 *** | 0.01 |
| diverting to PnR from destination [degrees] | 0.98 *** | 0.00 |
| PnR parking type [overflow] | 0.90 ** | 0.03 |
| PnR parking to platform [m] | 1.00 | 0.00 |
| PnR survey year [n] | 0.93 *** | 0.01 |
| PnR bays change [n] | 1.00 *** | 0.00 |
| PnR centrality [km] | 0.81 *** | 0.01 |
| road density near PnR [m per ha] | 0.99 *** | 0.00 |
| land use diversity near PnR [index] | 0.46 *** | 0.21 |
| worker density near PnR [per ha] | 1.05 *** | 0.00 |
| workers arriving by car near PnR [%] | 0.97 *** | 0.00 |
| dwelling density near PnR [per ha] | 1.11 *** | 0.01 |
| cars per dwelling near PnR [n] | 3.26 *** | 0.10 |
| auto-free households near PnR [%] | 1.00 | 0.00 |
| Observations | 52280 | |
| R2 Tjur | 0.199 | |
| log-Likelihood | -16919.504 | |
|
||
| Negative Binomial Model | ||
|---|---|---|
| Explanatory Variables | Log-Mean | s.e. |
| origin centrality [km] | 0.05 *** | 0.00 |
| diverting to PnR from destination [degrees] | -0.02 *** | 0.00 |
| PnR parking type [overflow] | 0.10 *** | 0.01 |
| PnR parking to platform [m] | 0.00 *** | 0.00 |
| PnR survey year [n] | -0.11 *** | 0.00 |
| PnR bays change [n] | -0.00 *** | 0.00 |
| PnR centrality [km] | -0.07 *** | 0.00 |
| road density near PnR [m per ha] | -0.00 *** | 0.00 |
| land use diversity near PnR [index] | 0.18 *** | 0.05 |
| worker density near PnR [per ha] | 0.02 *** | 0.00 |
| workers arriving by car near PnR [%] | -0.01 *** | 0.00 |
| dwelling density near PnR [per ha] | 0.02 *** | 0.00 |
| cars per dwelling near PnR [n] | 0.07 * | 0.04 |
| auto-free households near PnR [%] | -0.03 *** | 0.00 |
| Observations | 44508 | |
| R2 Nagelkerke | 0.670 | |
| log-Likelihood | -119956.137 | |
|
||
| Negative Binomial Model | ||
|---|---|---|
| Explanatory Variables | Incidence Rate Ratios | s.e. |
| origin centrality [km] | 1.05 *** | 0.00 |
| diverting to PnR from destination [degrees] | 0.98 *** | 0.00 |
| PnR parking type [overflow] | 1.11 *** | 0.01 |
| PnR parking to platform [m] | 1.00 *** | 0.00 |
| PnR survey year [n] | 0.90 *** | 0.00 |
| PnR bays change [n] | 1.00 *** | 0.00 |
| PnR centrality [km] | 0.94 *** | 0.00 |
| road density near PnR [m per ha] | 1.00 *** | 0.00 |
| land use diversity near PnR [index] | 1.20 *** | 0.05 |
| worker density near PnR [per ha] | 1.02 *** | 0.00 |
| workers arriving by car near PnR [%] | 0.99 *** | 0.00 |
| dwelling density near PnR [per ha] | 1.02 *** | 0.00 |
| cars per dwelling near PnR [n] | 1.07 * | 0.04 |
| auto-free households near PnR [%] | 0.97 *** | 0.00 |
| Observations | 44508 | |
| R2 Nagelkerke | 0.670 | |
| log-Likelihood | -119956.137 | |
|
||
Figure 6. Predictors of railheading behaviour
Railheading Behaviour Extent
A negative binomial regression model is employed to determine whether home origin, route, and the characteristics and context of the PnR can predict the extent of railheading in kilometres (Table 1). Although log-means are the coefficients for this model, these findings are also converted to Incidence Rate Ratios (I.R.R.) to simplify interpretation and are again visualised using a Forrest plot to order by magnitude (Figure 7). Residing further from the urban core increases the likely extent of railheading (I.R.R. = 1.05; p < 0.001) but this likelihood declines when access entail diverting away from the general path of the destination (I.R.R. = 0.98; p < 0.001). When the context of their chosen PnR is examined, each additional car per dwelling (I.R.R. = 1.07; p < 0.05) increases the likely extent of railheading by 7 percent although the confidence intervals are wide (Figure 7), and dwellings and workers per hectare each by 2 percent (I.R.R. = 1.02; p < 0.001; I.R.R. = 1.02; p < 0.001) while road density has a weak influence (I.R.R. = 1.00; p < 0.001) but unlike engagement in railheading, land use diversity increases the likely extent of railheading (I.R.R. = 1.20; p < 0.001) although again the confidence intervals are wide (Figure 7). These findings indicate that once railheading has begun, PnR users will extend their travel further to reach mixed land use thus suggesting that they intend to access nearby services. The likely extent of railheading increases to reach PnR where fewer nearby workers are arrive by car (I.R.R. = 0.97; p < 0.001) and where fewer residents own cars (I.R.R. = 0.97; p < 0.001), which suggests less auto-dependent locations are preferred to minimise competition for parking. Likewise, those parking as overflow are railheading further (I.R.R. = 0.9; p < 0.01) thus suggesting that overflow parking is the last resort after exhausting alternate and more closely located PnR in their search. Last, railheading distances decline towards the urban core (I.R.R. = 0.94; p < 0.001) and have since 2015 (I.R.R. = 0.90; p < 0.001), which are both positive findings since cruising for PnR will be less common where land is most expensive and it is declining thus suggesting that PnR development strategies are proving successful.
Figure 7. Predictors of railheading extent
Discussion and Conclusions
This study started by exploring the three dominant transport planning approaches to clarify the intent of the Multimodalism planning approach. Namely, ensuring that people residing within auto-dependent settings are not left stranded as cities increasingly prioritise active, public, shared, and eventually autonomous transport. Following, the focus narrowed to Park `n’ Ride facilities since these provide the auto-dependent with the opportunity to access public transit and in doing so, minimise their demands for road and inner-city parking capacity. Next, the empirical PnR literature was examined to clarify the distinction between transit provider and transit passenger intentions, and the unintended travel behaviours that can emerge such as railheading beyond the nearest PnR, which was the focus of the remaining study. The transport and social psychology literature was also examined for developing a novel conceptual framework of transport behaviour for explaining unintended travel behaviours. Last, a revealed preference survey of PnR users was examined, which afforded a unique opportunity to empirically examine railheading behaviour for the first time. Specifically, engagement in railheading behaviour was examined using a logistic regression model, and the extent of railheading behaviour using a negative binomial regression model. The findings of these models will now be discussed in greater depth, and followed by a concluding discussion of the policy implications. Despite the transit provider’s intentions to minimise VMT, 85 percent of Brisbane Statistical Division PnR users are engaging in railheading behaviour although this behaviour is in decline thus suggesting the Multimodalism planning approach is encouraging the intended travel behaviour. While it is not possible to distinguish residential determination from residential self-selection with the available data (Lin, Wang, and Guan 2017), there is evidence that place of origin influences engagement in railheading behaviour as this behaviour was most common from residents residing towards the urban periphery. While there are studies suggesting that PnR users intend to minimise diverting their route laterally or backwards to access the PnR (Mahmoud, Habib, & Shalaby 2014; Spillar 1997), an interesting finding is that railheading behaviour becomes more likely when there is the opportunity to encounter further PnR along the direct route. This suggests PnR users have greater commitment to their first encountered PnR upon deciding to divert their route. There was also evidence that PnR users are less inclined to railhead if their nearest PnR is surrounded by mixed land use and there are fewer nearby workers arriving by private automobile thus confirming that having the opportunity to access nearby services and minimise competition for parking can influence travel behaviours such as railheading (Bos et al. 2004; Frank and Pivo, 1994). These findings are particularly problematic given that mixed land use and arrival by modes other than private automobile are indicators of TOD, and TOD and PnR historically have had an antagonistic relationship since PnR require cheap land and should prioritise automobile access, while TOD raises nearby land values and should prioritise walking and cycling (Mingardo 2013; Shirgaokar and Deakin 2005; Willson 2005). While the Theory of Planned Behaviour suggests that stopping for these PnR that resemble TODs could comprise reasoned decisions or unreasoned habits and impulses (Van Acker, van Wee, and Witlox 2010). While the reasoning cannot be determined, this theory suggest that familiarity with the PnR has a role. As such, PnR users may more readily recall PnR resembling TOD because they visit these for purposes other than community, while PnR located away from further destinations or where land use is homogenous could fall outside of the PnR user’s awareness space altogether.
Examining the extent of railheading travel also provided further insight into this behaviour. For instance, PnR users were railheading further to reach PnR surrounded by greater land use diversity thus further supporting that they are persuaded by nearby services (Bos et al. 2004; Frank and Pivo, 1994). This could suggest that utility-maximization has a role (Hamer 2010; Hendricks and Outwater 1998; Qin et al. 2013) since these PnR have greater potential utility for PnR users if they require access to particular services before or after work (e.g. collecting groceries after work). Whether this influence is reasoned or unreasoned cannot be determined yet the empirical implications are important given that PnR simulations typically specify speed or affordability as the full extent of utility without considering further forms of utility such as daily access to services. Last among the most salient findings, PnR users are railheading further before parking as overflow. This finding suggests that overflow parking is a last resort once earlier PnR alternatives have been exhausted. This confirms that railheading is at least in part an optimal stopping problem (Feguson 1989; Wardrop 1952), and a potential strategy could be to minimise other forms of utility towards the inner city—such as nearby services—so that false hope for a better PnR does generate railheading behaviour. Given that this study is the first examination of railheading behaviour at the metropolitan transport network-scale and that secondary data is used, these findings are exploratory in nature but point the way to further research. For instance, primary data collection across multiple years could examine whether residential relocation influences PnR travel behaviour to clarify the roles of residential self-selection, residential determination, and travel habits. Similarly, primary data collection across consecutive days could explore the stability of PnR travel behaviour alongside the role of travel habits and impulses. A final recommendation could include capturing licence plates at entry to the PnR for a more direct examination of how PnR users respond to PnR parking vacancies.
## Policy Implications Urbanisation is accelerating worldwide (United Nations 2011) and so urban traffic congestion is worsening given that there are larger urban populations attempting to move about while still confined to the same two-dimensional urban surface (Stout 2015). As such, there are practical limits to private automobility by increasing road and parking capacity unless city officials, developers, and tax payers are willing to pay more by building up. For this reason, the Predict and Provide planning approach is losing financial viability and relevance in favour of more progressive planning approaches such as the Demand Management planning approach that prioritises active (e.g. walking and cycling), micro (e.g. eScooters and eBikes), public (e.g. busses, trams, and trains), and shared transport (e.g. ride-hailing and ride-sharing). While the end goal of this Demand Management urban reform is clear and may prove a sustainable solution to urban transport demands, the transition needs to be inclusive of those already residing within highly auto-dependent settings (Banister 2005; Ferguson 1990). This way, transit providers and planning authorities can avoid compounding transport disadvantage and the potential to strand aging-populations out towards the urban periphery. As such, the Multimodalism planning approach has the potential to smooth the transition by ensuring that the auto-dependent can still access public transit (Turnbull 1995; Shaw and Walton 2001). Yet the difficulty associated with explaining why motorists worldwide regularly ignore PnR with vacant parking and why they engage in unintended travel behaviours such as railheading beyond their nearest PnR is currently limiting Multimodalism outcomes. This study reveals that residential location, the built environment, the social equity of services, PnR characteristics and placement, and socio-demographics can each in part explain travel behaviour. Further, it reveals that PnR users seek to minimise how far they must divert their route to access PnR, that they are railheading further to reach PnR with nearby services, and that overflow parking generally occurs as a last resort after railheading greater distances. As such, strategies could include placing the PnR for more centrally located PnR users away from nearby services so railheading has less utility for more distantly located PnR users. This reduces the potential for parking conflict closer to the inner city and would reduce VMT, road capacity demand, and inner-city parking demand. In closing, there is an urgent need to better understand travel behaviour. After decades of automobile-fuelled urban expansion and suburbanisation, the period of ‘peak-car’ could be approaching its end within at least the more economically developed world (Goodwin 2012). The demand for road and parking capacity, and for faster and longer travel appears potentially limitless (Meyer 1999), and congestion the inevitable expression of automobility (Goodwin 1996; Hills 1996). As planning authorities and practitioners race to remain abreast of this urban transition, disruptive transportation technologies are emerging at an accelerating rate that introduce further complexity to the task (e.g. ride-hailing, ride-sharing, micro-mobility, Mobility-as-a-Service, and autonomous vehicles). As such, planning authorities and practitioners are forced to anticipate, prepare, and urgently adapt for an era of urban mobility without precedent (Anderson et al. 2016; Marshall 2018; McLaren and Agyeman 2015). It is plausible urban populations will resume their collective push towards the urban periphery as transport diversifies, becomes better integrated, and eventually autonomous since attentions will be freed from the roads and the frustrations of traffic congestion and locating parking (Anderson et al. 2016). As such, research that deepens our present understanding of how to influence travel behaviour through policy and design has the potential to strengthen the Multimodalism and Demand Management planning approaches, and divert the future of urban mobility away from scenarios where swarms of autonomous vehicles have the potential to clog city roads and increase VMT. With this concern in mind, this study has developed a novel conceptual model of travel behaviour, which was employed to examine and explain railheading behaviour. These research findings will interest urban researchers, policy makers, and planners aiming to understand and improve multimodalism and ensure that public transit is inclusive. By pursuing these aims, they are ensuring that future generations can enjoy urban mobility of at least similar quality as their forbearers or in other words, urban mobility that is sustainable. ## Acknowledgement This research is conducted through a project funded by the Australian Research Council Linkage Project grant LP160100031 with additional support from the industry partner the Queensland Department of Transport and Main Roads. Notably, the interpretations of the analysis are solely those of the authors and do not necessarily reflect the views and opinions of the Queensland Department of Transport and Main Roads or any of its employees.
References
Anderson, J. M., Kalra, N., Stanley, K. D., Sorensen, P., Samaras, C., & Oluwatola, O. A. (2016). Autonomous Vehicle Technology: A Guide for Policymakers. Santa Monica, C.A.: RAND Corporation. https://doi.org/10.7249/RR443-2
Atkinson, J. W., & Birch, D. (1970). The Dynamics of Action.
Arnott, R., & Rowse, J. (1999). Modelling parking. Journal of Urban Economics, 45(1), 97-124.
Australian Bureau of Statistics. (2016). Census of Population and Housing. (ABS, Ed.). Canberra: Author. http://www.abs.gov.au/websitedbs/censushome.nsf/home/datapacks
Axhausen, K.W., and Gärling, T. (1992) Activity‐based approaches to travel analysis: conceptual frameworks, models, and research problems. Transport Reviews 12.4 (1992): 323-341.
Banister, D. (2005). Unsustainable Transport : City Transport in the New Century. Routledge. https://doi.org/10.4324/9780203003886
Bates, D., Machler, M., Bolker, B., & Walker, S. (2015). Fitting Linear Mixed-Effects Models Using lme4. Journal of Statistical Software, 67(1), 1–48.
Becker, G. S. (1965). A Theory of the Allocation of Time. The Economic Journal, 493-517.
Bell, M. G. (2000). A game theory approach to measuring the performance reliability of transport networks. Transportation Research Part B: Methodological, 34(6), 533-545.
Blau, P. M. (1977). Inequality and Heterogeneity. New York: Free Press.
Bos, I. D. M., Van der Heijden, R. E. C. M., Molin, E. J. E., & Timmermans, H. J. P. (2004). The choice of park and ride facilities: An analysis using a context-dependent hierarchical choice experiment. Environment and Planning A. https://doi.org/10.1068/a36138
Bureau of Roads. (1969). Report on Australia road systems - 1968, Part 14. Canberra: Author.
Christiansen, P., Engebretsen, Ø., Fearnley, N., & Hanssen, J. U. (2017). Parking facilities and the built environment: Impacts on travel behaviour. Transportation Research Part A: Policy and Practice, 95, 198–206.
Debrezion, G., Pels, E., & Rietveld, P. (2009). Modelling the joint access mode and railway station choice. Transportation Research Part E: Logistics and Transportation Review, 45(1), 270–283. https://doi.org/10.1016/j.tre.2008.07.001
Dijk, M., & Parkhurst, G. (2014). Understanding the mobility-transformative qualities of urban park and ride polices in the UK and the Netherlands. International Journal of Automotive Technology and Management, 14(3/4), 246. https://doi.org/10.1504/IJATM.2014.065292
Dovey, K., Woodcock, I., & Pike, L. (2014). Intensifying Melbourne: Transit-oriented urban design for resilient urban futures. Planning News, 41(2), 10.
Duncan, M., & Christensen, R. K. (2013). An analysis of park-and-ride provision at light rail stations across the US. Transport Policy, 25, 148–157. https://doi.org/10.1016/J.TRANPOL.2012.11.014
Duncan, M., & Cook, D. (2014). Is the provision of park-and-ride facilities at light rail stations an effective approach to reducing vehicle kilometers traveled in a US context? Transportation Research Part A: Policy and Practice, 66, 65–74. https://doi.org/10.1016/J.TRA.2014.04.014
Ferguson, T. S. (1989). Who solved the secretary problem? Statistical Science, 4(3), 282-289.
Ferguson, E. (1990). Transportation demand management planning, development, and implementation. Journal of the American Planning Association, 56(4), 442–456. https://doi.org/10.1080/01944369008975448
Fisk, C. S. (1984). Game theory and transportation systems modelling. Transportation Research Part B: Methodological, 18(4-5), 301-313.
Foote, P. J. (2000). Chicago transit authority weekday park-and-ride users: Choice market with ridership growth potential. Transportation Research Record. https://doi.org/10.3141/1735-19
Frank, L. D., & Pivo, G. (1994). Impacts of mixed use and density on utilization of three modes of travel: single-occupant vehicle, transit, and walking. Transportation Research Record.
Ginn, S. (2009). The Application of the Park & Ride and TOD Concepts to Develop a New Framework that Can Maximise Public Transport Patronage. Queensland University of Technology: Brisbane.
Goodwin, P. B. (1996). Empirical evidence on induced traffic: A review and synthesis. In Transportation 23.
Goodwin, P. B. (2012). Peak Travel, Peak Car and the Future of Mobility.
Goulden, M., Ryley, T., & Dingwall, R. (2014). Beyond ‘predict and provide’: UK transport, the growth paradigm and climate change. Transport Policy, 32, 139–147. https://doi.org/https://doi.org/10.1016/j.tranpol.2014.01.006
Hamer, P. (2010). Analysing the effectiveness of park and ride as a generator of public transport mode shift. Road & Transport Research: A Journal of Australian and New Zealand Research and Practice, 19(1), 51.
Heggie, I. G., & Papoulias, D. B. (1976). Operational Performance of Park and Ride: Objectives and Achievements in Oxford. Transport Studies Unit.
Hendricks, S., & Outwater, M. (1998). Demand forecasting model for Park-and-Ride lots in King County, Washington. Transportation Research Record: Journal of the Transportation Research Board, 1623(1), 80–87. https://doi.org/10.3141/1623-11
Hounsell, N., Shrestha, B., & Piao, J. (2011). Enhancing Park and Ride with access control: A case study of Southampton. Transport Policy, 18(1), 194–203. https://doi.org/10.1016/j.tranpol.2010.08.002
Huntley, P. (1993). Park and ride: strategic modal shift or cheap overflow. Local Transport Today, 22(July), 12-13.
Kane, M., & Whitehead, J. (2018). How to ride transport disruption –a sustainable framework for future urban mobility. Australian Planner, 54(3), 177–185. https://doi.org/10.1080/07293682.2018.1424002
Kimpton, A., Pojani, D., Ryan, C., Ouyang, L., Sipe, N., & Corcoran, J. (2020a). Contemporary parking policy, practice, and outcomes in three large Australian cities. Progress in Planning, 100506. https://doi.org/10.1016/j.progress.2020.100506
Kimpton, A., Pojani, D., Sipe, N., & Corcoran, J. (2020b). Parking Behavior: Park ‘n’ Ride (PnR) to encourage multimodalism in Brisbane. Land Use Policy 91(February)104304.
Lin, T., Xia, J., & Robinson, T. P. (2014). Spatial analysis of Park and Ride users’ dynamic accessibility to train station: A Case Study in Perth. https://pdfs.semanticscholar.org/7885/09b98b97d4bc2057eaf81a8727e92dd15e6a.pdf. Accessed 5 February 2019
Lin, T., Wang, D., & Guan, X. (2017). The built environment, travel attitude, and travel behavior: Residential self-selection or residential determination? Journal of Transport Geography, 65, 111-122.
Mahmoud, M. S., Habib, K. N., & Shalaby, A. (2014). Park-and-Ride access station choice model for cross-regional commuting: case study of Greater Toronto and Hamilton Area, Canada. Transportation Research Record, 2419(1), 92–100.
Marshall, A. (2018). As shared scooters invade, San Francisco decides who belongs where. Wired.
McLaren, D., & Agyeman, J. (2015). Sharing Cities: a Case for Truly Smart and Sustainable Cities. Cambridge, MA : MIT Press.
Meek, S., Ison, S. G., & Enoch, M. P. (2008). Park and ride: Lessons from the UK experience. Transportation Research Board 87th Annual Meeting, 13-17 January 2008,. Washington DC.
Meek, S., Ison, S. G., & Enoch, M. (2009). Stakeholder perspectives on the current and future roles of UK bus-based Park and Ride. Journal of Transport Geography, 17(6), 468–475. https://doi.org/10.1016/J.JTRANGEO.2008.10.007
Millard-Ball, A. (2019). The autonomous vehicle parking problem. Transport Policy, 75, 99–108. https://doi.org/10.1016/J.TRANPOL.2019.01.003
Mingardo, G. (2013). Transport and environmental effects of rail-based Park and Ride: evidence from the Netherlands. Journal of Transport Geography, 30, 7–16. https://doi.org/10.1016/J.JTRANGEO.2013.02.004
Meyer, M. D. (1999). Demand management as an element of transportation policy: using carrots and sticks to influence travel behavior. Transportation Research Part A: Policy and Practice, 33(7-8), 575-599.
Newell, A., & Simon, H. A. (1972). Human Problem Solving (Vol. 104, No. 9). Englewood Cliffs, NJ: Prentice-Hall.
Noel, E. C. (1988). Park-and-ride: alive, well, and expanding in the United States. Journal of Urban Planning and Development, 114(1), 2–13.
Parkhurst, G. (1995). Park and ride: could it lead to an increase in car traffic? Transport policy, 2(1), 15–23. https://doi.org/10.1016/0967-070X(95)93242-Q
Parkhurst, G. (2000). Influence of bus-based park and ride facilities on users’ car traffic. Transport Policy, 7(2), 159–172. https://doi.org/10.1016/S0967-070X(00)00006-8
Parkhurst, G., & Buxton, S. (2005). Towards a Truly Strategic Appraisal Process for Evaluating Park and Ride Schemes: The Development of an Assessment Framework for London.
Parkhurst, G., & Meek, S. (2014). The effectiveness of park-and-ride as a policy measure for more sustainable mobility. In Parking Issues and Policies (pp. 185–211). Emerald Group Publishing Limited. https://doi.org/10.1108/S2044-994120140000005020
Qin, H., Guan, H., & Wu, Y.-J. (2013). Analysis of park-and-ride decision behavior based on Decision Field Theory. Transportation Research Part F: Traffic Psychology and Behaviour, 18, 199–212. https://doi.org/10.1016/J.TRF.2013.02.001
R Core Team. (2019). R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing. https://www.r-project.org/
Salomon, I., & Ben-Akiva, M. (1983). The use of the life-style concept in travel demand models. Environment and Planning A, 15(5), 623-638.
Sharma, B., Hickman, M., & Nassir, N. (2019). Park-and-ride lot choice model using random utility maximization and random regret minimization. Transportation, 46(1), 217-232.
Shaw, J., & Walton, W. (2001). Labour’s new trunk-roads policy for England: an emerging pragmatic multimodalism? Environment and Planning A, 33(6), 1031–1056. https://doi.org/10.1068/a33202
Shirgaokar, M., & Deakin, E. (2005). Study of park-and-ride facilities and their use in the San Francisco Bay Area of California. Transportation Research Record: Journal of the Transportation Research Board, (1927), 46–54.
Singh, Y. J., Lukman, A., Flacke, J., Zuidgeest, M., & Van Maarseveen, M. F. A. M. (2017). Measuring TOD around transit nodes-Towards TOD policy. Transport Policy, 56, 96-111.
Sochor, J., Strömberg, H., & Karlsson, I. C. M. (2015). Implementing Mobility as a Service. Transportation Research Record: Journal of the Transportation Research Board, 2536(1), 1–9. https://doi.org/10.3141/2536-01
Sun, L. J., & Gao, Z. Y. (2007). An equilibrium model for urban transit assignment based on game theory. European Journal of Operational Research, 181(1), 305-314.
Spillar, R. J. (1997). Park-and-ride planning and design guidelines. 1995 William Barclay Parsons Fellowship Parsons Brinckerhoff Monograph 11. New York, NY: Parsons Brinckerhoff Inc.
Stieffenhofer, K. E., Barton, M., & Gayah, V. V. (2016). Assessing park-and-ride efficiency and user reactions to parking management strategies. Journal of Public Transportation. https://doi.org/10.5038/2375-0901.19.4.5
Stieffenhofer, K. E., Barton, M., & Gayah, V. V. (2015). Assessing Park-and-Ride Use and User Reactions to Parking Management Strategies: A Case Study in Puget Sound, Washington. Journal of Public Transportation.
Stout, F. (2015). The automobile, the city, and the new urban mobilities. In The City Reader (pp. 752–762). Routledge.
The Committee on Public Works. (1970). 1970 Annual Report on the Demonstration Fringe Parking Program. Washington, D.C.
Transport and Main Roads. (2019a). PnR parking supply for 2011 and 2016. Brisbane: Author.
Transport and Main Roads. (2019b). Park `n’ Ride Number Plate Survey. Brisbane: Author.
Transport and Main Roads. (2019c). State Controlled Roads - Queensland. https://www.data.qld.gov.au/dataset/state-controlled-roads-queensland. Accessed 1 December 2019
Tsang, F. W. K., Shalaby, A. S., & Miller, E. J. (2010). Improved modelling of park-and-ride transfer time: Capturing the within-day dynamics. Journal of Advanced Transportation, 39(2), 117–137. https://doi.org/10.1002/atr.5670390202
Turnbull, K. F. (1995). Effective Use of Park-and-Ride Facilities. Texas, TN: National Cooperative Highway Research Council.
United Nations. (2011). World Urbanization Prospects, the 2011 Revision. New York: Author.
Van Acker, V., Van Wee, B., & Witlox, F. (2010). When transport geography meets social psychology: toward a conceptual model of travel behaviour. Transport Reviews, 30(2), 219-240.
Venables, W. N., & Ripley, B. D. (2013). Modern Applied Statistics with S-Plus (4th ed.). New York, NY: Springer Science & Business Media. https://doi.org/10.2307/2685660
Wang, H., Meng, Q., & Zhang, X. (2014). Park-and-ride network equilibrium with heterogeneous commuters and parking space constraint. Transportation Research Record: Journal of the Transportation Research Board, (2466), 87–97.
Wardman, M., Lythgoe, W., & Whelan, G. (2007). Rail Passenger Demand Forecasting: Cross-Sectional Models Revisited. Research in Transportation Economics. https://doi.org/10.1016/S0739-8859(07)20005-8
Wardrop, J. G. (1952). Road paper: some theoretical aspects of road traffic research. Proceedings of the Institution of Civil Engineers, 1(3), 325-362.
Willson, R. (2005). Replacement Parking for Joint Development: an access policy methodology.
Zahabi, S. A. H., Miranda-Moreno, L. F., Patterson, Z., & Barla, P. (2012). Evaluating the effects of land use and strategies for parking and transit supply on mode choice of downtown commuters. Journal of Transport and Land Use, 5(2), 103–119.
Zhao, X., Chen, P., Jiao, J., Chen, X., & Bischak, C. (2019). How does ‘park and ride’ perform? An evaluation using longitudinal data. Transport Policy, 74, 15–23. https://doi.org/10.1016/j.tranpol.2018.11.004
Zhou, Y., Li, Y., Hao, M., & Yamamoto, T. (2019). A System of Shared Autonomous Vehicles Combined with Park-And-Ride in Residential Areas. Sustainability, 11(11), 3113.