UBC Faculty Research and Publications

The Health Effects of Fixed-Guideway Transit Investment : A Review of Methods and Best Practice Devries, Danielle; Iroz-Elardo, Nicole, 1981-; Hong, Andy; Winters, Meghan; Brauer, Michael; Frank, Lawrence D. Apr 26, 2018

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
52383-Devries_Danielle_et_al_Heath_effects_Final_Report.pdf [ 1.31MB ]
Metadata
JSON: 52383-1.0368659.json
JSON-LD: 52383-1.0368659-ld.json
RDF/XML (Pretty): 52383-1.0368659-rdf.xml
RDF/JSON: 52383-1.0368659-rdf.json
Turtle: 52383-1.0368659-turtle.txt
N-Triples: 52383-1.0368659-rdf-ntriples.txt
Original Record: 52383-1.0368659-source.json
Full Text
52383-1.0368659-fulltext.txt
Citation
52383-1.0368659.ris

Full Text

 1  ! 	      Prepared For:  City of Vancouver  Vancouver Coastal Health Authority TransLink     Prepared By:  Danielle Devries, SFU Nicole Iroz-Elardo, UBC Andy Hong, UBC Meghan Winters, SFU Michael Brauer, UBC Lawrence Frank, UBC, Principal Investigator     April 26, 2018     TABLE OF CONTENTS  FIGURES .................................................................................................................................................... 4TABLES ..................................................................................................................................................... 5I. Introduction ............................................................................................................................................. 1II. Evidence Acquisition ............................................................................................................................. 21. Search Strategy ................................................................................................................................... 22. Eligibility ............................................................................................................................................ 33. Data Extraction ................................................................................................................................... 3III. Evidence Synthesis ............................................................................................................................... 41. Study Designs ..................................................................................................................................... 42. Study Characteristics .......................................................................................................................... 53. Methods Synthesis by Outcome ........................................................................................................ 113.1. Transportation Outcomes ........................................................................................................... 113.2. Intermediate and Health Behaviour Outcomes .......................................................................... 163.3. Environmental Outcomes ........................................................................................................... 223.4. Social and Economic Outcomes ................................................................................................ 243.5. Health Outcomes ........................................................................................................................ 294. Methodological Concerns ................................................................................................................. 294.1 Capturing Local and Regional Impacts ....................................................................................... 294.2. Isolating the Impacts of a Transit Investment ............................................................................ 314.3. Capturing impacts associated with vulnerable populations ....................................................... 314.4. Assessing behavioural change and environmental exposure ..................................................... 324.5. Location-specific objective data for exposure and behaviour ................................................... 324.6. Sample size and participant retention ........................................................................................ 32IV. Discussion ........................................................................................................................................... 331. Best Practices in Pre-Post Rail Studies ............................................................................................. 332. Research Gaps and Opportunities ..................................................................................................... 343. Limitations ........................................................................................................................................ 37V. Conclusions .......................................................................................................................................... 37Appendix 1 – Literature Review Search criteria ....................................................................................... 38  Appendix 2 – Methods commonly used in Quasi-experimental studies ................................................... 39Appendix 3 – Summary of All Articles Included in the Review .............................................................. 41References ................................................................................................................................................. 47  Figure 1: Linking Built Environment with Chronic Disease and Healthcare Costs (source Frank et al 2017) ........................................................................................................................................................... 1Figure 2: Search process for articles included in the Systematic Review .................................................. 4Figure 3. Search process for identifying studies based on research design .............................................. 38  Table 1. Study designs of the papers reviewed ........................................................................................... 5Table 2: Overview of study locations, rapid transit systems, methods and outcomes ................................ 6Table 3: Summary of Review Articles Addressing Travel Mode and Mode Shift as an Outcome .......... 12Table 4: Summary of Review Articles Addressing Vehicle Ownership as an Outcome .......................... 14Table 5: Summary of Review Articles Addressing Vehicle Miles Traveled as an Outcome ................... 14Table 6: Summary of Review Articles Addressing Accessibility as an Outcome .................................... 15Table 7: Summary of Review Articles Addressing Walking, Bicycling and Physical Activity as an Outcome .................................................................................................................................................... 16Table 8: Summary of Review Articles Addressing Body Mass Index, Overweight, and Obesity as an Outcome .................................................................................................................................................... 19Table 9: Summary of Review Articles Addressing Traffic Safety, Injuries, and Fatalities as an Outcome................................................................................................................................................................... 20Table 10: Summary of Review Articles Addressing Air Pollution & GHG Emissions as an Outcome .. 22Table 11: Summary of Review Articles Addressing Mental Health and Crime as an Outcome .............. 24Table 12: Summary of Review Articles Addressing Economic Development, Employment, and Personal Income as an Outcome .............................................................................................................................. 25Table 13: Summary of Review Articles Addressing Population Density and Land Value as Outcomes 27Table 14: Summary of Review Articles Addressing Health Outcomes ................................................... 29Table 15: Studies that fall under the type A category ............................................................................... 3312A growing body of evidence suggests that transportation and land use investments and policies can have far reaching implications for population health, access to economic opportunities, and greenhouse gas emissions (de Nazelle et al. 2011; Frank et al. 2008; Frank et al. 2010; Schweitzer & Zhou 2010). Transportation systems link people with social and health promoting resources, such as employment, education, food, recreation, social services, and health care (Caspi et al. 2012; Grengs 2015; Julien et al. 2015; Neutens 2015).  Public transportation provides mobility and access to opportunities for many people, especially for those living in more compact, mixed development corridors. Transit supports the concentration of activities that results in shorter distances to a broader range of opportunities. A compact urban form enables walking and biking between transit stops, home, work, and other destinations. Transit contrasts the private vehicle that provides a more dispersed form of access to activities and a less concentrated or “sprawling” urban form. The positive association between transit access, physical activity, and reduced chronic disease is becoming well documented (Besser & Dannenberg 2005; Edwards 2008; Freeland et al. 2013; Saelens et al. 2014; Lachapelle et al. 2009).   As illustrated in Figure 1, transportation investments and associated land use actions may impact health outcomes through two primary paths: one is through encouraging or discouraging health promoting behaviours and the other through exposure to harmful substances and stressors. The linkages identified in the diagram are not exhaustive and have been shown to have varying degrees of influence depending on physical, socio-demographic, and context.   !-*D5!&#!&-!$,&.!*'&%&,/!,  *'&!!++&	$, *'+,+<+'-**&#,$ECDJ=   2  Both behavioural and exposure based pathways linking the built environment with health outcomes are important; further, the relationships between behavioural and exposure impacts of the built environment on health is complex.  For example, public transit investments likely contribute to regional reductions in traffic and air pollutant emissions, however localized air pollution is generally higher near roads and even higher still in “street canyons” (Beaudoin et al. 2015; McAndrews et al. 2017). Thus exposure may actually increase for those who live where transit investments and related densification occur.  Further, for the segments of the population which actually engage in active travel, this behaviour may result in increased exposure to air pollution, intake of inhaled pollutants, and injury risk (Bigazzi & Figliozzi, 2014; de Nazelle et al., 2011; Jerrett et al., 2009; Jerrett et al., 2014). For others, increased air pollution exposure related to densification may lead to increased risk. At an individual level, while there is theoretically some point at which the harms of pollution outweighs physical activity benefits (Tainio et al. 2016), higher inhalation rates during physical activity and exposure to localized pollution without the barrier of a cabin suggest caution (Bigazzi & Figliozzi 2014; Ham et al. 2017).   The Millennium Line Broadway Extension (MLBE) is a rail rapid transit project that extends an existing SkyTrain line in the City of Vancouver. It runs along the Broadway Corridor, with six underground stations spanning six km. The MLBE will serve the second largest job centre (the UBC-Broadway corridor) in the province, and the largest hospital in Western Canada. MLBE presents an opportunity to assess both behavioural and exposure-based health effects; deepen our understanding of these mechanisms and pathways; and build upon the growing body of intervention based research summarized in this report.  This report summarizes evidence and methodologies from longitudinal studies of light rail, rail rapid transit, and bus rapid transit lines.  Given the diversity of studies to date and the need to carefully plan a pre-post study, this literature review and analysis of methodological options provide the City of Vancouver, partner agencies, and researchers the basis to make informed decisions about study design for future pre-post assessments of MLBE.  A careful assessment of the relative differences in health impacts of transit investment for “choice” versus “dependent” riders will be an important step forward for future studies (Lachapelle et al. 2016).   2 92( *(*/Studies were selected for inclusion in this report if they focused on an exclusive right-of-way or fixed guideway transit intervention (rail or bus rapid transit (BRT)) with a pre-test and post-test study design. Four sets of search terms were used to identify articles: 1) rapid transit interventions 2) health-related and other1 outcomes 3) longitudinal impacts and 4) specific populations (See Appendix 1). Several databases were searched including TRID, GEOBASE, Web of Science, Academic Search Complete, Medline, CINAHL, PsycINFO, and Dissertations and Theses Abstracts and Index. Searches were                                                 1 Several bodies of literature were not included in the selected articles, since they did not add to the main goal of this review: health impacts on the transit operator, air pollution inside buses or subways, and impacts of construction, such as vibration, not the opening of new transit.  Additionally, there is a large literature (32 articles and 5 theses) on property value that have been previously reviewed for BRTs (Stokenberga 2014) and were thus excluded from this review.        3 limited to those published in English over the last 10 years (2007-2017), and inclusive to articles up to August 31, 2017. After the title and abstract scan, we checked for unique articles in the reference lists and Google Scholar. The search strategy is demonstrated in figure 2.   :2#!!!#!*/We initially reviewed titles and abstracts for inclusion in the full text assessment using two main inclusion criteria: 1. The article must study a rapid transit intervention; and 2. The intervention must be a right-of-way or fixed guideway system.  In the full text assessment of the articles, two additional criteria were used: 1. The study must include a pre-test and post-test measurement of the outcome and 2. The study outcome cannot be property value1.  ;2*.*(*!&%For each article, information was extracted on study aim, study design, city, type of intervention (i.e. light rapid transit (LRT), bus rapid transit (BRT), etc.), year of intervention, outcome variables, data collection methods, study groups, unit of analysis, spatial scale, statistical methods, and findings.  4!-*E5* (*'++'**,!$+!&$-!&, 1+,%,!.!/2	92*+/)!%)Since this review focussed on pre- and post- test studies, all articles assess the implementation of a rapid transit system. These studies are also called “natural experiment studies”, because they capitalize on something that occurs in policy or practice.  The intervention or investment is viewed by researchers as as the basis of of a study’s experimental design. To carry out these studies in a truly experimental way, the researchers must include a comparison group that is not exposed to the intervention, in this case a new transit investment. Such studies may be deemed “quasi-experimental” because they mimic a true experiment. The comparison group is crucial to determine if the results observed are indeed because of the transit investment. However, in practice, it is not always feasible to include a comparison group. In the review we found half of the studies (n=26) did include a comparison group. Another major design consideration for natural experimental studies is data collection. There are two main ways to collect data: longitudinal and repeat cross-sectional. In a longitudinal study the data collection is longitudinal, so the same individuals are followed over time and measured before and after (pre versus post) transit investments are implemented. In a repeat-cross-sectional design, a different sample is measured pre versus post the transit investment is made.  While a longitudinal study removes some error in comparing pre versus post, it requires more coordination and budget and retaining    5 participants can be difficult. Often researchers do not have the capacity to collect their own primary data; therefore, most of them rely on secondary data sources, such as the Canadian Community Health Survey, California Health Interview Survey, Electronic Medical Record / PopData), which are regularly collected for repeat cross-sectional surveys. In this review we found about a third of the studies (n=18) used a longitudinal data.   All study designs in this review fall into one of four categories:   A) Longitudinal study with a comparison group B) Repeat cross-sectional with a comparison group C) Longitudinal study without a comparison group, and  D) Repeat cross-sectional without a comparison group.   For the purpose of reviewing best practice in pre versus post test study methodology, we have ranked them A to D. Studies with a comparison group (A & B) are more rigorous than those without (C & D).  They allow researchers to determine if the results are due to the exposure to new rapid transit versus something else that also impacts a given outcome. Studies using a longitudinal research data (A & C) are more rigorous than those using repeat cross-sectional (B & D), because they allow researchers to have better certainty that the changes occurred because of the intervention, not external factors from measuring a different sample. Table 1 outlines the four study design categories.  $D6,-1+!&+', ((*+*.!/ Longitudinal Repeat Cross Sectional YES Comparison Group A (5 studies) B (21 studies) NO  Comparison Group C (13 studies) D (13 studies)  :2*+/ (*(!)*!)A total of 52 studies were included in the review: 50 published journal articles and 2 theses and dissertations. These studies described 33 unique rapid transit systems in 31 cities including 8 bus rapid transit, 25 rail (light rail transit, rail rapid transit, or tram/street car system) (see Table 2). To avoid confusion, the rapid transit systems were categorized into three groups as follows:  • Bus Rapid Transit (BRT): Bus system with separated guideway. Examples include TransMilenio in Bogota, Colombia.  • Light Rail Transit (LRT): Low capacity rail system with separated guideway and priority through intersections. It may be in street or in a fully grade separated right of way. Trams and streetcars are included this category. Examples include MAX in Portland, Oregon and TRAX in Salt Lake City, Utah.  • Rail Rapid Transit (RRT): A fully grade separated rail system. It may be underground or elevated. Examples include metros, subways and SkyTrain in Metro Vancouver.  In terms of geographic coverage, both high-income (n=35 studies) and low- and mid-income countries (n=17) are included, with studies from Canada, Denmark, Norway, Spain, USA, and UK, to Chile,    6 China, Columbia, Ecuador, India, Mexico, Pakistan, Singapore, and Taiwan.  There were examples of multiple studies on the same systems: Bogota, CO (n=7), Cambridge (n=3), Denver (n=3), Los Angeles, US (n=5), Salt Lake City, US (n=7), and Toronto, CA (n=4). Table 1 outlines studies reviewed by city and rapid transit system.  $E5.*.!/'+,-1$',!'&+3*(!,*&+!,+1+,%+3%, '+&'-,'%+City Transit System Image Methodology (Study Author, Date)  Outcomes Beijing, CN Subway (RRT)  '*(&))4)*!&%#-!* &$'(!)&% (Xie 2016)  Transportation (mode) Bergen, NO Bergen Light Rail (LRT)  '*(&))4)*!&%#-!* &$'(!)&% (Engebretsen et al. 2017)  Transportation (mode) Bogota, CO TransMilenio (BRT)  '*(&))4)*!&%#-!* &$'(!)&%   (Bocarejo et al. 2013; Bocarejo et al. 2016; Combs & Rodriguez 2014; Combs 2017; Heres et al. 2009; Rodriguez et al. 2016)  '*(&))4)*!&%#1%&&$'(!)&%<'*"',$6ECDE=   Transportation (accessibility, vehicle ownership), behavioural (injuries), social & economic (density, income, land use, land value)  Cali, CO MIO (BRT)  '*(&))4)*!&%#1%&&$'(!)&%  (Delmelle & Casas 2012)  Transportation (accessibility) Cambridge, UK Cambridgeshire Guided Busway (BRT)  &%!*+!%#1%&&$'(!)&% (Panter et al. 2016; Heinen & Ogilvie 2016; Heinen et al. 2017) Transportation (mode), behavioural (physical activity)     7 Charlotte, US LYNX Blue Line (LRT)  '*(&))4)*!&%#-!* &$'(!)&%  (Billings 2011); #&%!*+!%#1%&&$'(!)&%6'&)*4 &&%*(&#-!* '(&'%)!*/$* !%7 (MacDonald et al. 2010) Behavioural (BMI, physical activity), social & economic (crime) Chicago, US CTA Rail (RRT)  '*(&))4)*!&%#-!* &$'(!)&% (Shen 2013)  Social & economic (density) Copenhagen, DK Metro (RRT)  '*(&))4)*!&%#-!* &$'(!)&% (Rotger & Nielsen 2015) Transportation (commute distance), social & economic (earnings) Croydon, UK Tramlink (LRT)  '*(&))4)*!&%#-!* &$'(!)&% (Lee & Senior 2013) Transportation (car ownership, mode) Dallas, US A-Train (RRT)  '*(&))4)*!&%#-!* &$'(!)&%<**'- ECDG=  Social & economic (retail location) Delhi, IN Metro (RRT)  '*(&))4)*!&%#-!* &$'(!)&% (Goel & Gupta 2015) Environmental (air pollution) Denver, US RTD Light Rail (LRT)  '*(&))4)*!&%#-!* &$'(!)&% (Bhattacharjee & Goetz 2016; Shen 2013) '*(&))4)*!&%#1%&&$'(!)&% (Bardaka et al. 2016)  Social & economic (density, income, land use)    8 Eugene, US EmX (BRT)  '*(&))4)*!&%#1%&&$'(!)&% (Nelson et al. 2013) Social & economic (new jobs) Houston, US METROrail (LRT)  &%!*+!%#-!* &$'(!)&% (Park & Sener 2017) &%!*+!%#1%&&$'(!)&% (Durand et al. 2016)  Transportation (mode), behavioural (physical activity), environmental (air pollution), health (stroke mortality)  Lahore, PK Metrobus (BRT)  '*(&))4)*!&%#1%&&$'(!)&% (Mansoor et al. 2016) Social & economic (public perspectives) Los Angeles, US Metro (LRT)  &%!*+!%#-!* &$'(!)&% (Spears et al. 2017; Hong et al. 2016);  '*(&))4)*!&%#-!* &$'(!)&%< &ECDF=; '*(&))4)*!&%#1%&&$'(!)&%6'&)*4 &&%*(&#-!* '(&'%)!*/$* !%7 (Ridgeway & MacDonald 2017)  Transportation (vehicle miles travelled), behavioural (physical activity), environmental (GHG emissions), social & economic (density, crime) Madrid, ES Metro (RRT)  '*(&))4)*!&%#-!* &$'(!)&% (Calvo et al. 2013) Social & economic (population density Mexico City, MX Metrobus (BRT)  '*(&))4)*!&%#1%&&$'(!)&% (Chang et al. 2017) Behavioural (physical activity) Metro Rail (RRT)  '*(&))4)*!&%#1%&&$'(!)&% (Guerra 2014) Transportation (ridership, travel time), social & economic (costs, land use)    9 Manchester, UK Metrolink (LRT)  '*(&))4)*!&%#-!* &$'(!)&% (Lee & Senior 2013) Transportation (car ownership, mode) Minneapolis, US METRO (LRT)  '*(&))4)*!&%#-!* &$'(!)&% (Hurst & West 2014) '*(&))4)*!&%#1%&&$'(!)&% (Fan et al. 2012)  Transportation (accessibility), social & economic (land use) Portland, US MAX (LRT)  '*(&))4)*!&%#-!* &$'(!)&% (Ewing & Hamidi 2014) Transportation (vehicle miles traveled) Quito, EC Trolebus (BRT)  '*(&))4)*!&%#-!* &$'(!)&%  (Rodriguez et al. 2016) Social & economic (land use) Salt Lake City, US TRAX (LRT)  &%!*+!%#-!* &$'(!)&%  (Werner et al. 2016) &%!*+!%#1%&&$'(!)&% (Brown & Werner 2007; Brown & Werner 2011; Brown et al. 2017; Brown & Werner 2008; Brown et al. 2015; Miller et al. 2016)  Transportation (mode, ridership), behavioural (BMI, physical activity), social & economic (perspectives) Santiago, CL Transantiago (BRT)  &%!*+!%#1%&&$'(!)&% (Yáñez et al. 2010) Transportation (mode) South Yorkshire, UK Supertram (LRT)  '*(&))4)*!&%#-!* &$'(!)&% (Lee & Senior 2013) Transportation (car ownership, mode) 10Seattle, US Sound Transit Link (LRT)  &%!*+!%#1%&&$'(!)&% (Huang et al. 2017) Behavioural (physical activity) Singapore, SG MRT (RRT)  '*(&))4)*!&%#-!* &$'(!)&% (Zhu & Diao 2016) Transportation (mode), social & economic (density)  Taipei, TW Metro Taipei (RRT)  '*(&))4)*!&%#-!* &$'(!)&% (Chen & Whalley 2012; Ding et al. 2016)  Environmental (air pollution) Toronto, CA Subway (RRT)  '*(&))4)*!&%#-!* &$'(!)&% (Saxe et al. 2017); '*(&))4)*!&%#1%&&$'(!)&% <', ,$6ECDG40,$6ECDH=  Transportation (accessibility, mode), environmental (GHG emissions) Streetcar (LRT)  '*(&))4)*!&%#1%&&$'(!)&% (Richmond et al. 2014)Behavioural (injuries)  Washington DC, US Metrorail (RRT)  '*(&))4)*!&%#-!* &$'(!)&% (Shen 2013)  Social & economic (density)  West Midland, UK Metro (RRT)  '*(&))4)*!&%#-!* &$'(!)&% (Lee & Senior 2013) Transportation (car ownership, mode)    11 ;2* &)/%* )!)/+*&$Articles were grouped into 5 outcome categories:   1) Transportation outcomes which include accessibility, travel mode, vehicle ownership, vehicle miles travelled (VMT);  2) Behavioural outcomes which include BMI/obesity, physical activity, walking, cycling;  3) Environmental outcomes which include greenhouse gas (GHG) emissions and air pollution; 4) Social & Economic outcomes which include income, crime, land use, density, new jobs, retail locations, and property values; and 5) Health outcomes including mortality and morbidity   Eighteen studies included transportation outcomes, 13 included behavioural outcomes, five had environmental outcomes, and 15 had social & economic outcomes. Only one study assessed actual health outcomes and only the Werner et al (2016), Ewing and Hamidi (2014), and Spears et al (2017) studies were “quasi-experimental” and had control groups.  The other studies are subject to confounding influences of potential changes in factors that also influence specific outcomes. Each of the 5 groups of outcomes is discussed in greater detail in the following sections.   It should be noted that a quasi-experimental study is suitable for a situation where implementation of an intervention is without the control of the investigators, but an opportunity exists to evaluate its effect, hence the name, “quasi-experiment”. A famous example is a study by Friedman et al (2001) which examined the impact of the 1996 Atlanta Olympic Games on air quality.   While quasi-experimental design is common in many social science studies, one critical weakness is the lack of random assignment (Craig et al. 2017). Random assignment of study participants to a treatment group and control group is a gold standard in clinical trial to ensure the effect of a treatment is independent from more general trends or external shocks (e.g. economic recession) that might affect the study results. To account for the lack of random assignment, most quasi-experimental studies adopt various statistical methods to ensure that their results are internally and externally consistent. In this review, many studies employed common statistical techniques, such as regression adjustment or difference-in-differences, to account for the quasi-experimental nature of their studies. For readers not familiar with these statistical methods, a general description of all the relevant statistical methods is provided in Appendix 2.   ;292(%)'&(**!&%+*&$)A growing number of longitudinal studies have investigated the effects of new transit systems on various transportation-related outcomes, such as accessibility, travel mode, vehicle ownership, and vehicle miles travelled (VMT) which indirectly can impact health. The impact of a new transit infrastructure is largely a function of the relative utility (travel time, convenience, and comfort) of alternative modes such as driving.  As consumers, travelers will respond to changes in the transportation system based on the relative costs and benefits of other options. The effects will be greatest if new transit systems attract car drivers to switch to transit based on shifts in relative utility across competing modes.  However, the “elasticity of demand” or how much a particular shift in relative travel time, cost,    12 convenience, and comfort impacts someone is based on their own demographic and attitudinal characteristics.  Capturing choice riders who have options and are new to transit typically requires time saving. Dependent riders, and those that already chose to use transit before the investment, are more often captured resulting in less net impact on transport-related outcomes.   #'! !%$F5-%%*1'.!/*,!$+*++!&*.$'&' !,+&-,'%Authors; City; System Type (LRT, RRT, BRT) Study Design (A, B, C, D) Data Collection Tool; Recruitment Unit of Analysis Statistical Method Outcome; Direction of Findings (increase, decrease, no change) Brown & Werner, 2007; Salt Lake City; LRT longitudinal without comparison (C) Accelerometer & survey; Mail, residents Person Ordinary least squares regression Mode; increased transit share Engebretsen, et al., 2017; Bergen; LRT Repeat cross-sectional with comparison (B) Secondary data, travel survey Statistical unit, Person Spatial analysis, logistic regression  Mode; increased transit share Foth et al., 2014; Toronto; RRT Repeat cross-sectional without comparison (D) Secondary data, travel survey & census Traffic Analysis Zone Multiple linear regression Mode; no change overall, increased transit share to shop/work, increased transit share for socially disadvantaged  Guerra, 2014; Mexico City; RRT Repeat cross-sectional without comparison (D) Secondary data, travel survey Household trip Binomial logit, ordinary least squares regression, Poisson regression Mode; increased transit share Heinen & Ogilvie, 2017; Cambridge; BRT Longitudinal without comparison (C) Self-reported survey; Mail, employees Person Multinomial logistic regression Mode; no change Heinen & Ogilvie, 2016; Cambridge; BRT Longitudinal without comparison (C) Self-reported survey; Mail, employees Person Multinomial logistic regression Mode; increased active travel share, decreased car share  Lee & Senior, 2013; 4 UK cities; LRT Repeat cross-sectional with comparison (B) Secondary data, census  Ward Descriptive Mode; increased  LRT share, decreased bus share Saxe et al., 2015; Toronto; RRT Repeat cross-sectional with comparison (B) Secondary data, travel survey & census Passenger kilometer traveled Descriptive, correlation Mode; increased RRT share, decreased bus, car share    13 (PKT) Werner et al., 2016; Salt Lake City; LRT Longitudinal with comparison (A) Observational passenger counts Ridership  Bonferri adjustment, fixed effects repeated measures regression Mode; increased transit share, no displacement from bus share to LRT Xie, 2016; Beijing; RRT Repeat cross-sectional with comparison (B) Secondary data, travel survey Individual trip Difference in difference  Mode; increased rail, walking, bike share, decreased auto share  Yanez et al., 2010; Santiago; BRT & RRT Longitudinal without comparison (C) Interviews, 5-day travel diary; Mail, employees Person  Winner-looser analysis (descriptive) Mode; increased rail, rail-bus share, decreased bus share Zhu & Diao, 2016; Singapore; RRT Repeat cross-sectional with comparison (B) Secondary data, travel survey & land parcels Household Difference in difference  Mode; increased rail share, decreased car share   *BRT = Bus Rapid Transit; LRT = Light Rail Transit; RRT = Rail Rapid Transit  Studies of transit investment impacts on mode shift are shown in Table 3. Most studies in this category reported that new rail transit systems tend to replace existing bus ridership; the two exceptions that reported significant ridership increase linked to new LRT system was in Utah, USA (Werner et al. 2016) and Bergen, Norway (Engebresten et al, 2017).  Foth et al (2014) reported that the new transit investment in Toronto, Canada did not affect overall transit ridership but was associated with higher transit mode share to retail, service, and office jobs. They also reported that socially disadvantaged areas were positively related to transit ridership. Guerra (2014) reported that an extension of the existing metro line in Mexico was associated with higher metro ridership, but noted that most of the increase came from informal transit, not individual car use.  He observed that the new line was related to reduced transit expenditure and travel time, and increased residential density around Metro stations.   Yáñez et al. (2010) reported that a new BRT system in Santiago resulted in higher proportion of passengers using rail or rail-bus combination, and the majority of the increased ridership diverted from bus only users. The authors reasoned that this replacement effect occurred because rail is perceived as a reliable mode. Similarly, Xie (2016) reported that a rail transit expansion in Beijing, China was associated with an increase in rail, walking, and bike trips but a decrease in auto, as well as a non-significant decrease in bus trips. The author noted that the quantity of travel did not change due to the rail expansion. Zhu and Diao (2016) reported that mode share changed to a lower proportion of car trips and a higher proportion of rail trips.  As noted, Werner et al (2016) reported that transit ridership increased significantly after the light rail opened in Salt Lake City, UT where previous levels of transit service and corresponding ridership were low. Another interesting study was completed by Heinen and Ogilvie (2016) in Cambridge, UK.  Instead of focusing on a single mode, they examined travel mode variability. They reported that individuals with higher variation in mode choice at baseline were more likely to increase their active travel share but reduce their car share with an increased exposure to new busway. Heinen and Ogilvie (2017) did a    14 follow-up study of their 2016 study by including all samples from the four waves. The newer analysis showed that there was a large heterogeneity in travel behaviour patterns over time, and the results did not show any statistically meaningful patterns. Lastly, Engebresten et al (2017) have shown that a new LRT line in Bergen, Norway significantly increased public transit use both in volume and market share. The Norway study, while optimistic, did not differentiate between rail and bus ridership; therefore, it is not possible to determine whether there was a substitution effect between rail and bus. Furthermore, no control group or demographic covariates were included in the analysis and the ability to distinguish between “choice” and “dependent” transit markets is limited.   ( #$"$G5-%%*1'.!/*,!$+*++!& !$/&*+ !(+&-,'%Authors; City; System Type (LRT, RRT, BRT) Study Design (A, B, C, D) Data Collection Tool; Recruitment Unit of Analysis Statistical Method Outcome; Direction of Findings (increase, decrease, no change) Combs & Rodriguez, 2014; Bogota; BRT Repeat cross-sectional with comparison (B) Secondary data, travel survey Household Difference in difference, logistic regression Vehicle ownership; decrease for high income, increase for low income Lee & Senior, 2013; 4 UK cities; LRT Repeat cross-sectional with comparison (B) Secondary data, census  Ward Descriptive Vehicle ownership; increase  Two of the studies (Table 4) assessed vehicle ownership after the new lines opened and found differing results (Combs & Rodriguez, 2014; Lee & Senior, 2013). Combs & Rodriguez (2014) in Bogota found that the effect of TransMilenio on car ownership was different for high and low income groups. Those who are high income reduced vehicle ownership, while those who are low income increased vehicle ownership. This may indicate a suburbanizing effect on lower income groups due to the rising property values around rapid transit stations, since those who were low income and lived near transit also reduced vehicle ownership like the higher income group. On the other hand, Lee & Senior (2013) found that the participants in their study across four cities in the UK actually increased their car ownership in three of the four locations. They suggest this is because the mode shift to using light rail mostly came from bus users rather than those who drive personal motor vehicles.  $#'$H5-%%*1'.!/*,!$+*++!& !$!$+*.$+&-,'%Authors; City; System Type (LRT, RRT, BRT) Study Design (A, B, C, D) Data Collection Tool; Recruitment Unit of Analysis Statistical Method Outcome; Direction of Findings (increase, decrease, no change) Ewing and Hamidi, 2014; Portland; LRT Repeat cross-sectional with comparison (B) Secondary data, travel survey Household Difference of means test, 2-stage models Vehicle miles traveled; decrease Spears et al., Longitudinal Self-reported Household Descriptive, chi- Vehicle miles traveled;    15 2016; Los Angeles; RRT with comparison (A) 7-day travel diary; Mail, resident squared test, t-test, differences in difference decrease  Two studies (Table 5) assessed the impacts of rail on vehicles miles of travel (VMT) and reported that new light rail transit was generally associated with reductions in vehicle miles travelled. In Portland, Ewing and Hamidi (2014) have shown that there are both direct and indirect effects of transit investment that reduce VMT. They also argued that the indirect effect of transit investment through increased walking and density would be greater than the direct effect through transit ridership alone. Spears et al (2017) observed that households living within 1 km of new light rail transit drove less than control households living farther away. A decrease in average car trip length was observed after the light rail opened in the participants’ neighbourhood and explained the reduction in VMT.   $$%*$I5-%%*1'.!/*,!$+*++!&++!!$!,1+&-,'%Authors; City; System Type (LRT, RRT, BRT) Study Design (A, B, C, D) Data Collection Tool; Recruitment Unit of Analysis Statistical Method Outcome; Direction of Findings (increase, decrease, no change) Bocarejo et al., 2016; Bogota; BRT Repeat cross-sectional with comparison (B) Secondary data, travel survey & census Planning Zone Spatial change analysis, difference in difference, multiple linear regression Accessibility; increase for low income, no change for high income Combs, 2017; Bogota; BRT Repeat cross-sectional with comparison (B) Secondary data, travel survey Household Difference in difference, Poisson regression, partial proportional odds parameterization Accessibility; no change Delmelle & Casas,  2012; Cali; BRT Repeat cross-sectional without comparison (D) Secondary data, census Neighbourhood Spatial analysis Accessibility; increase to hospital, higher for middle income Fan et al., 2012; Minneapolis; LRT Repeat cross-sectional without comparison (D) Secondary data, census Census Block Spatial analysis, ordinary least squares regression Accessibility; increase to jobs  Several studies (Table 6) have looked at the impact of new transit investment on accessibility. The definition of accessibility varied from one study to another. In general, studies have used three measures: Hansen’s gravity measure, Hirschman-Herfindhal (HH) index, and an Entropy index. Two studies used the Hansen’s gravity measure, which quantifies the number of activities that can be reached    16 within a given range of travel costs (Hansen 1959); one study used the Hirschman-Herfindhal (HH) index and an Entropy index to measure social fragmentation; and one study used tour frequency and diversity of trip purposes as a measure of fulfillment of household travel needs.  Several studies (Bocarejo 2016; Delmelle et al 2012; Fan et al 2012) reported impacts of new transit varied across income strata. Bocarejo et al. (2016) reported that lower income groups had more access to destinations than higher income groups after the BRT started operation in Bogota, CO. Similarly, the introduction of new BRT system in Cali, Colombia, Delmelle et al (2012) resulted in greater walking access to destinations for middle-income groups than for low and high-income groups. They reported access to hospitals slightly increased after the introduction of the BRT. Fan et al (2012) reported that new LRT in Twin Cities, MN, led to a significant increase in job accessibility for all workers. The accessibility benefits were larger for low-wage workers in downtown and north of LRT stations but the benefits were less for those in other station areas. However, Combs (2017) also studied the Bogota’s BRT system and reported no significant association between BRT access and mobility measures (travel purpose diversity and tour frequency).  Methodologically, many of the studies employed a difference-in-differences (DID) approach as noted in the Table 5. Most DID studies incorporated a number of household and individual characteristics as controls. However, some studies did not incorporate any control variables, possibly causing misleading interpretation and mis-specification of model outcomes. Three studies reviewed here did not use any modeling approach; instead they just compared a percentage change before and after the intervention. Direct comparison may not be appropriate because other factors – including secular trends or economic conditions - also impact the outcomes being considered. Of the studies reviewed here, Werner et al (2016) and Spears et al (2017) are probably the most rigorous studies because they employed a treatment and control design with primary data collection before and after the intervention on the same cohort. The repeated measure mixed effects model employed by Werner et al (2016) is also an advancement over the traditional DID approach because it addresses spatial clustering and shared built environment inputs for the individuals who are reporting their travel behaviour over time (within-group difference).    ;2:2%*($!*%	#*  ,!&+(+*&$) + + *$%'%*$J5-%%*1'.!/*,!$+*++!&$#!&3!1$!&& 1+!$,!.!,1+&-,'%Authors; City; System Type (LRT, RRT, BRT) Study Design (A, B, C, D) Data Collection Tool; Recruitment Unit of Analysis Statistical Method Outcome; Direction of Findings (increase, decrease, no change) Brown & Werner, 2007; Salt Lake City; LRT Longitudinal without comparison (C) Accelerometer & survey; Mail, residents Person Ordinary least squares regression Physical activity; increase Brown et al., 2015; Salt Lake City; Longitudinal without comparison (C) Accelerometer, GPS & survey; Door to door Person One-way analysis of variance, Physical activity; increase for new riders, decrease for former riders    17 LRT ordinary least squares regression Chang, et al., 2017; Mexico City; BRT Repeat cross-sectional without comparison (D) Self-reported survey Person Propensity score matching, 2-step cluster analysis, k-means cluster analysis  Physical activity; increase for walking, bicycling, greatest for females with low education Durand et al., 2015; Houston; LRT Longitudinal without comparison (C) Accelerometer, survey, travel diary & neighbourhood audits; Mail, residents Person Latent Growth Curve model, piecewise model, full information maximum likelihood estimation, propensity scores Physical activity; no results (protocol paper) Hong et al., 2016; Los Angeles; LRT Longitudinal with comparison (A) Accelerometer, GPS, survey & 7-day travel diary; Mail, residents Person T-test, Fisher test, ANCOVA framework Physical activity; increase Huang et al., 2017; Seattle; LRT Longitudinal without comparison (C) Accelerometer, GPS, survey & travel diary; Phone, residents Person Negative binomial regression Walking; decrease overall, increase within 0.4 km MacDonald et al., 2010; Charlotte; LRT  Longitudinal without comparison (post-hoc control with propensity matching) (C) Self-reported survey; Phone, residents Person Logistic regression, propensity score weighting Physical activity; increase Miller et al., 2015; Salt Lake City; LRT Longitudinal without comparison (C) Accelerometer, GPS & survey; Door to door Person Descriptive, paired t-tests, spatial analysis Physical activity; increase for new riders, decrease for former riders Panter et al., 2016; Cambridge; BRT Longitudinal without comparison (C) Self-reported survey; Mail, employees Person T-test, chi-squared test, signed rank test, multinomial logistic regression  Physical activity; increase for bicycling, transport  Rail or BRT has the potential to increase levels of walking, cycling, and overall physical activity. Introducing a new fixed guide way corridor is often paired with additional street improvements for pedestrians and cyclists, which also may support physical activity. Eventual changes in land use including densification and mixed use also supports active travel.    18  Seven interventions – five light rail systems and two bus rapid transit systems – studied physical activity in connection with increased walking and biking (Table 7). Most studies confirm a statistically significant increase in walking or biking and associated utilitarian physical activity in response to the building of a light rail or bus rapid transit line.  Populations that have lower baseline activity (Hong et al 2016) may be most likely to change their behaviour and showed an increase in active commuting and physical activity; Brown et al (2015) also showed that switching modes (to or from being a transit rider) resulted in a larger change in physical activity. In contrast, Huang et al (2017) in Seattle found a decrease in physical activity after a transit intervention, where daily minutes of overall walking decreased from 36.2 to 25.1 minutes after the intervention. In the Cambridge UK study, Panter et al (2016) reported no statistically significant changes in walking, but significant changes for bicycling and active commuting (minutes of walking and bicycling combined).  Evidence on the effect of the transit investment in vulnerable populations is mixed. Chang et al (2017) showed that vulnerable subpopulations were most likely to benefit from the Mexico City BRT whereas Huang et al (2017) in Seattle showed prior LRT users and educated men were the most likely to walk around the immediate area of the new LRT stations.   Some researchers note that utilitarian active travel to and from light rail and the destinations around it may be a shift from other recreational physical activity – or a substitution effect.  Evidence remains unclear on this matter. The study of a new line in Salt Lake City, Utah (Miller et al (2015)) showed strong evidence that uptake of transit – and light rail in particular - had a positive effect on MVPA, a negative effect on sedentary activity, and no substitution effect. Similarly, in Mexico City (Chang, 2017), walking for transport and recreation increased after a BRT, suggesting utilitarian physical activity did not entirely substitute recreational physical activity.   Panter et al’s (2016) evaluation of a BRT in Cambridge UK did not support or refute substitution. They reported a statistically significant increase in active transportation and biking for utilitarian purposes but recreational PA did not decrease.  However, there was no overall increase in physical activity. Huang et al (2017) shows an overall decrease in all types of walking over time. Interestingly, this analysis demonstrated locational shifting of walking within the 1-mile buffer of the LRT station area; an increase of approximately 2.26 more minutes occurred near the station area (within 0.25 miles) when compared to the 0.75-1 mile area.   Several sampling and analysis methodology parameters or challenges appear in the studies that investigate walking, biking, and physical activity. Measuring physical activity can occur through objective measures with GPS and accelerometry (Brown et al 2007; Brown et al 2015; Hong et al 2016; Huang et al 2017) and subjective self-report measures from a questionnaire (Chang et al 2017; MacDonald et al 2010; Panter et al 2016). Objective measurement is the gold standard; self-report tends to overestimate and round physical activity bouts. Another challenge is maintaining sample size for follow up data collection after transit investments are made. Huang et al (2017) were only able to retain 29 percent of their original sample with GPS and Accelerometry in Seattle; Hong et al (2016) was able maintain about 50 percent of their GPS/accelerometer sample. Most of the studies above were true longitudinal cohorts; the exception is Chang et al (2017) in Mexico City which used a matching    19 procedure in a repeated cross-sectional design.  Cases and controls were generally defined by being within or outside the LRT or BRT corridor.    Buffer sizes for defining “cases” vary across studies: ½ km or 1/3 mile (Chang et al 2017); 1-mile (McDonald et al (2010), Huang et al (2017)); 2 km or 1.2 miles (Brown et al 2015); and 3 miles (Durrand et al 2015). This variation in buffer sizes may explain some of the differences in results.  The decay in walking at 0.75 miles (1.2 km) from the station (Huang et al 2017) suggests buffer sizes from 1-1.2 km likely capture the full walking effect.   Additional urban design and demographic characteristics were used as covariates in statistical analyses. Most of the interventions included complete street improvements with several authors discussing pedestrian (Chang et al 2017) and bike facility (Panter et al 2016) improvements.  Teasing apart these effects could be accomplished through pre-planned spatial and temporal sampling of areas with differing levels of pedestrian and bike facilities.  Further, land use impacts of rail improvements or TOD takes time to implement (Brown et al 2015) and mature into a key facilitator of increased physical activity.    Most pre-post studies are in the U.S. and were funded by the National Institutes of Health on 5-year timelines. This is a minimal amount of time needed to get a pre-post sample and test changes in travel and activity patterns resulting from the transit investment.  However, resulting land use changes from the rail investment also impact behaviour and work synergistically with increased accessibility and take more time to transpire.  Most researchers have difficulty completing analysis of resulting data within the 5 year time period.  Longer study periods are potentially more costly, more time will allow land use impacts to be assessed.   Pedestrian and cycling infrastructure improves local accessibility and work interactively with transit investments.  However, there was no published study found that attempted to assess the separate or combined effect of rail and complete street improvements.  There is at least one study underway nearing completion that has focused extensively on this topic.   !*$$ )+'#(%+ $%*$K5-%%*1'.!/*,!$+*++!&'1++&03.*/! ,3&+!,1+&-,'%Authors; City; System Type (LRT, RRT, BRT) Study Design (A, B, C, D) Data Collection Tool; Recruitment Unit of Analysis Statistical Method Outcome; Direction of Findings (increase, decrease, no change) Brown & Werner, 2009; Salt Lake City; LRT Longitudinal without comparison (C) Accelerometer, interviews &  survey; Mail, residents Person Regression Obesity; decrease for new riders Brown et al., 2015; Salt Lake City; LRT Longitudinal without comparison (C) Accelerometer, GPS & survey; Door to door Person One-way analysis of variance, ordinary least squares BMI; decrease for new riders, increase for former riders    20 regression Brown et al., 2017; Salt Lake City; LRT Longitudinal without comparison (C) Accelerometer, GPS & survey; Door to door Person Regression BMI; decrease for new riders, increase for former riders MacDonald et al., 2010; Charlotte; LRT  Longitudinal without comparison (post-hoc control with propensity matching) (C) Self-reported survey; Phone, residents Person Logistic regression, propensity score weighting BMI; decrease for LRT users  Increased physical activity from active modes can decrease body mass index (BMI) and obesity and overweight populations.  The literature is robust in showing that increased physical activity in general and walking in particular are helpful in decreasing BMI.  However, the longitudinal literature documenting BMI changes in fixed guideway transit intervention studies is limited.  Three studies –a station in Salt Lake City, a new LRT in Salt Lake City, and a new LRT in Charlotte – include BMI and/or obesity as an outcome (Table 8).  In a small sample around a new light rail station, Brown et al (2008) showed statistically higher obesity rates among nonriders (65%), new riders (26%) and continuing riders (15%), but did not have significant power to show differences over time.  MacDonald et al (2010) showed a significant BMI effect; Charlotte LRT users lost, on average, 6.42 lbs and were 81 percent less likely to become obese (OR = 19%, CI=0.04,0.92). Brown et al (2015) also showed statistically significant loss of 0.5 BMI points for new riders but also showed a gain of 0.66 BMI points for former riders.  Conceptually, obesity and BMI outcomes are is seen as an extension of the physical activity pathway; see figure 1.  Detecting changes in BMI may require a larger sample size (i.e. n>800) and duration of exposure (3-5 years) for effects to be realized and to detect underlying relationships.  It is also notable that some studies use BMI as a covariate for physical activity (Huang 2017; Panter 2016), which captures the effect of body weight on engagement in physical activity; physical activity and BMI share a reciprocal relationships.  There are several studies underway that have measures of body weight as an outcome which should result in more evidence.    #%*+ &#$+ %%$$L5-%%*1'.!/*,!$+*++!&*!,13&"-*!+3&,$!,!++&-,'%Authors; City; System Type (LRT, RRT, BRT) Study Design (A, B, C, D) Data Collection Tool; Recruitment Unit of Analysis Statistical Method Outcome; Direction of Findings (increase, decrease, no change) Bocarejo et al., 2012; Bogota; BRT Repeat cross-sectional without comparison (D)  Secondary data, police report Road Segment Hot spot analysis Injury or fatal collisions; decrease Richmond et Repeat cross- Secondary Road Spatial analysis, Pedestrian motor vehicle    21 al., 2014; Toronto; LRT (streetcar) sectional without comparison (D)  data, police report Segment zero inflated Poisson regression collisions; decrease  Building a rail line may result in mode shifts that can impact exposure to traffic and thus accident, injury, and even fatality.  Along with transit, a pedestrian scale improvement in a given corridor likely improves conditions for pedestrians and cyclists and could decrease accidents and serious injuries for these modes.  Increased walking and biking, however, likely increases the exposure (distance traveled) for active modes and thus could result in more overall accidents.  However, other research not reviewed in this section shows that as a proportion of trips per capita, overall injury rates often decline.   Few pre-post rail studies have looked at traffic safety and resulting injuries and fatalities (Table 9).  Bocarejo et al (2012) analyzed serious traffic accidents before and after two major BRTs were implemented in Bogata, Columbia. They found that serious traffic accidents steadily declined in the years post intervention. The number of personal injury accidents declined 48 and 60 percent over a 10-year time along the 2 BRT routes compared to 39 percent for all of Bogata.  A hotspot analysis showed a concentration in serious accidents on routes to the busiest stations and areas where vehicular speed increased; the authors attribute this to lack of “complete street” retrofitting for pedestrians.  A second study - Richmond et al (2014) - examined pedestrian-vehicular accidents before and after the conversion of a streetcar in Toronto from mixed traffic to a right-of-way (ROW) alignment.  This study showed a 48% decrease in the overall rate of collisions (IRR=0.52, 95% CI: 0.37-0.74) including rate of collisions for adults (IRR=0.61, 95% CI: 0.38-0.97) and children (IRR=0.13, 95% CI: 0.04-0.44).  The Toronto study also showed a 46% decrease in minor injuries (IRR=0.56, 95% CI: 0.40-0.80), but did not show a statistically significant difference in major injuries and had insufficient data for fatalities.   Methodologically, both studies relied heavily upon existing data.  Accident reports were geocoded and summarized. Bocarejo et al (2012) summarized overall serious accidents by year but did not perform any statistical tests.  Similarly, the hotspot analysis relies upon kernel GIS methods to create density surfaces for analysis by spatial visualization.  Richmond et al (2014) took a more rigorous approach, using a zero inflated Poisson regression analysis and spatial point pattern analysis to test for differences. Some scholars are concerned that increases in transit and active modes results in increased exposure (miles travel) of active travelers and thus disproportionate accidents, injuries, and fatalities.   Accounting for this phenomenon would require adequate understanding of the denominator (miles traveled) by each mode. Traffic accidents for pedestrian and bicyclists are also underreported, particularly for minor accidents that do not result in a major injury or fatality. Emphasis on accident counts instead of health impacts (injuries and fatalities) is common but may arguably miss key aspects of the human health impact.  This is likely due to reliance on accident (police) reports as a data source instead of health data that would be required to ascertain the extent of the injury. Furthermore, understanding traffic injuries and fatalities requires careful design to tease apart the effect from the rail line itself versus complete street type facilities within the corridor.           22 ;2;2%,!(&%$%*#+*&$)New light rail transit will likely attract more people to ride transit and likely change their commute mode choice. Mode shift from private vehicles (and buses) may reduce overall vehicle emissions and potentially lead to reductions in air pollutant concentrations and population exposure. Rail can also concentrate development over time and foster increased activities and vehicle congestion and related emissions within a given area.  There is a limited body of longitudinal studies looking at the air quality and noise impact of light rail transit. One study explored the impact of LRT on a measure of traffic-related air pollution and its impacts on stroke mortality, with several other papers focusing on air quality and greenhouse gas emissions (GHG).   #!&%! -	$$! $$DC5-%%*1'.!/*,!$+*++!&!*'$$-,!'&@	%!++!'&++&-,'%Authors; City; System Type (LRT, RRT, BRT) Study Design (A, B, C, D) Data Collection Tool; Recruitment Unit of Analysis Statistical Method Outcome; Direction of Findings (increase, decrease, no change) Boarnet et al., 2017; Los Angeles; LRT Longitudinal  with comparison (A) Self-reported survey; Mail, residents Household Difference in difference GHG emissions; decrease Chen & Whalley, 2012; Taipei; RRT Repeat cross-sectional with comparison (B) Secondary data, air monitoring  City Ordinary least squares regression Air pollution; decrease Ding et al, 2016; Taipei; RRT Repeat cross-sectional with comparison (B) Secondary data, air monitoring  City Spearman correlation, Kruskal-Wallis test, Wilcoxon rank sum test, Friedman test, logistic regression  Air pollution; decrease Goel & Gupta, 2015; Dehli; RRT Repeat cross-sectional without comparison (D) Secondary data, air monitoring  Line extension Ordinary least squares regression Air pollution; decrease Park & Sener, 2017; Houston; LRT Longitudinal  with comparison (A) Secondary data, air monitoring  Catchment (16 km) Spatial analysis, zero inflated Poisson regression Air pollution; decrease Saxe et al., 2015; Toronto; RRT Repeat cross-sectional with comparison (B) Secondary data, travel survey & census Dissemination Area Descriptive, correlation GHG Emissions; decrease Saxe et al., 2017; Toronto; RRT Repeat cross-sectional with comparison (B) Secondary data, travel survey & census Station Catchment (3.2 km) GHG savings analysis GHG emissions; decrease    23 Seven studies were included in the category of studies of air pollution and GHG emissions (Table 10). Park and Sener (2017) reported that the new LRT line in Houston, Texas led to a 13% reduction (95% CI: -0.23 to -0.02) in concentrations of acetyl, a volatile organic compound  that is emitted mainly by automobile exhaust with levels that are proportional to the magnitude of automobile emissions. Park and Senor (2017) used a two-year window – 2 years before and after the LRT opening. The emission reduction was considerably smaller for control sites (those not influenced by road proximity) compared to the exposure sites (those close to the LRT stations). Using the similar two-year window, Chen and Whalley (2012) reported that the new rail transit system in Taiwan led to 5-15% reduction in CO and NOx but had little effect on ground level ozone. A simple calculation of traffic diversion from automobile to rail indicated that the percentage change of mode shift was just over 10%, similar to the estimated CO effects.   Similarly, Ding et al. (2016) found that the opening of the Taipei rail transit system was associated with a downward trend in the number of days with high levels of CO and NO2 but an upward trend in the days with high O3 levels. However, given the general downward trend in CO and NO2 in Taiwan, we suspect that the ozone results are more likely to be driven by other factors than the new rail intervention. Ding et al examined a time trend without any time window. Goel and Gupta (2015) reported that an extension of the rail transit system in Delhi, India was associated with a 34% reduction (95% CI: -54.38 to -13.62) in localized CO and a declining trend in NO2. They examined 5-week and 9-week window for the pre/post comparison. They also compared the growth rates of registered vehicles in Delhi against Mumbai, and found a decrease in the growth rate of automobiles in Delhi after the rail was opened, suggesting that the improvement in CO levels may be attributed to the traffic diversion effect from automobiles to rail transit.   Saxe et al. (2015) reported that a new extension of the subway system in Toronto, Canada led to a net reduction in estimated greenhouse gas emissions (66.4 kilotons of CO2e). Emissions reductions were modelled based on surveys of travel patterns before and after the extension opened. They reasoned that the reduction mainly came from GHG savings shifted from cars, and noted that there was an increase in GHG emissions associated with the passenger kilometer travelled shifted from buses. In a companion study, Saxe et al (2017) expanded the initial analysis to include the GHG impacts of construction, operation, ridership patterns, and land use changes associated with the new subway line. Their results suggest that the combined effect of GHG reduction from the new subway line depends on how much GHG savings are attributed to buses or automobiles. GHG savings were substantially larger for passenger kilometer traveled shifted from automobiles than from buses. In both studies, Saxe et al (2015, 2017) did not employ any time window; instead they examined a time trend after the subway line was opened in 2002.  Boarnet et al (2017) reported that after the opening of the new LRT system in Los Angeles, CA, there was a significant reduction in CO2 emissions (27%) for the households living close to the new LRT station. They incorporated life-cycle CO2 emissions due to system building and maintenance, and concluded that the reduction in private vehicle emissions comes very close to the changes in total transportation emissions as a result of the opening of the new LRT line. Boarnet et al (2017) employed a 6-8 months window before and after the LRT opening.     24 Methodologically, most studies employed statistical methods commonly used for program evaluation. Park and Sener (2017) used a disrupted time series including a comparison between exposure and control sites to model the effects of new LRT line on VOC levels and stroke mortality. Saxe et al (2015; 2017) calculated GHG savings from passenger kilometer travelled and emission factors. Among the five studies reviewed, Park and Sener (2017) used the most rigorous research method as they compared the LRT impact of the treatment group against the control group over time.  Boarnet et al (2017) also used a rigorous method of comparing households living close and farther away from new transit, and employed a difference-in-differences approach to estimate changes in GHG emission. Chen and Whalley (2012) and Goel and Gupta (2015) used a regression discontinuity design, which assigns a control group for observations before the rail investment and a treatment group after the investment. No detailed specification of the models nor regression results were reported by Ding et al (2016), making it difficult to evaluate their methods and findings.   Except for one study (Boarnet et al, 2017) that estimated GHG emissions from a travel diary, none of the studies measured person-level exposure. Most studies employed ambient concentration data from regional monitoring stations. This is not necessarily a weakness of the existing studies because it can be argued that transit investments may lead to a region-wide reduction in air pollution. The studies reviewed here mostly had a longer time window (2 years), so the use of regional monitoring data would be appropriate. However, to get more immediate health impacts of rail investment with a shorter time-window, more precise measurement, preferably personal exposure assessment, would be recommended.     ;2<2&!#%&%&$!+*&$) %	% #$DD5-%%*1'.!/*,!$+*++!&&,$	$, &*!%+&-,'%Authors; City; System Type (LRT, RRT, BRT) Study Design (A, B, C, D) Data Collection Tool; Recruitment Unit of Analysis Statistical Method Outcome; Direction of Findings (increase, decrease, no change) Billings et al., 2011; Charlotte; LRT Repeat cross-sectional with comparison (B) Secondary data, police report Station Catchment (0.4 km) Difference in difference Crime; decrease Brown & Werner, 2011; Salt Lake City; LRT Longitudinal  without comparison (C) Self-reported survey; Mail, residents Person  T-test, general linear regression Residential perspectives; increase in safety, decrease in crime perceptions Mansoor et al., 2016; Lahore; BRT Repeat cross-sectional without comparison (D) Self-reported survey; intercept Person  Descriptive  Residential perspectives; increase positive perceptions  Ridgeway & MacDonald, 2017; Los Angeles; RRT Repeat cross-sectional without comparison (post-hoc control with propensity Secondary data, police report Police Reporting District  Stepped wedge design, Poisson regression, permutation tests Crime; decrease    25 matching) (D)  Transit investments serve as a redevelopment strategy and likely influence perceptions of well-being.  Four studies were included in this category (Table 11). Brown and Werner (2011) reported that there was a statistically significant increase in perception of child and pedestrian safety after the opening of the LRT in Salt Lake City, Utah. Residents’ sense of community and neighbourhood also improved after the intervention. Mansoor et al (2016) also reported that residents found new metro bus transit in Pakistan to be a more convenient and affordable transportation option with widespread benefits. In terms of crime, Billings et al (2011) found that the announcement of opening of the LRT in Charlotte, NC was associated with a decreased rate of crime. They noted that the actual opening of the LRT had no effect on crime.   ! !'!" %+"!* %+ #$!  !$DE5-%%*1'.!/*,!$+*++!&'&'%!.$'(%&,3%($'1%&,3&*+'&$&'%+&-,'%Authors; City; System Type (LRT, RRT, BRT) Study Design (A, B, C, D) Data Collection Tool; Recruitment Unit of Analysis Statistical Method Outcome; Direction of Findings (increase, decrease, no change) Bardaka et al., 2016; Denver; LRT  Repeat cross-sectional without comparison (D) Secondary data, census Census Block Group Spatial analysis, global & local Moran's I, ordinary least squares regression, LaGrange multiplier test Per capita income; increase    Fan et al., 2012; Minneapolis; LRT Repeat cross-sectional without comparison (D) Secondary data, census Census Block Spatial analysis, ordinary least squares regression Job accessibility; increase Heres et al., 2014; Bogota; BRT Repeat cross-sectional with comparison (B) Secondary data, census Household Weighted least squares regression, probit models Income; increase, highest for low and middle income strata Mansoor et al., 2016; Lahore; BRT Repeat cross-sectional without comparison (D) Self-reported survey; intercept Person  Descriptive  Residential perspectives; increased jobs, decreased gender discrimination  Nelson et al., 2012; Eugene; BRT Repeat cross-sectional without comparison (D) Secondary data, InfoGroup Station Catchment (0.4 km) Descriptive   New Jobs; increase  Rotger & Nielsen, 2015; Copenhagen; Repeat cross-sectional with comparison (B) Secondary data, public register Household Differences in difference regression Earnings; increase    26 RRT Yarbough, 2014; Dallas; RRT Repeat cross-sectional with comparison (B) Interviews & secondary data, census; Phone, businesses Station Catchment (0.8 km) Difference in difference  Business locations; increase  Transportation investments in general and transit in particular are an integral part of supporting a robust urban economy.  High quality, fixed transit is an efficient way to link individuals to employment and bring customers to business.  There is, however, some concern that transit can make an area “too” desirable, initiating or exacerbating gentrification and displacement for pre-existing low and mid-income communities.2   In this review, we found seven recent intervention studies that examined economic development impacts of rapid transit (Table 12). However, it is important to note that this is an area of research with a long tradition in the U.S. context.  Rail investments have historically needed to be justified in part based on their economic development and land use impacts.  It is notable that 2 of the 3 BRT projects are in developing countries with the third in a small North American city (Eugene, Oregon) where the density and types of employment as well as the spatial distribution of residential income may not be comparable to Vancouver.       Rapid transit investments increase accessibility in a concentrated area and are correlated with net increases in development around station areas.  For example, Bhattacharie et al (2016) showed that commercial density increased within a 0.5 mile buffer of rail stations in Denver. Yarbough (2014) documented net retail growth and some spatial clustering around some stations in Denton, Texas with differentiated retailers (restaurants, clothing, novelty shops) more likely to be attracted to the area than non-differentiated retailers such as furniture or automotive shops. Nelson et al (2012) addressed not just business location, but employment location; this study documents that 42% of new jobs were within ¼ mile of a BRT station with health, social, and administrative services highly attracted and tourism, arts, and entertainment weakly attracted to station areas in Eugene, Oregon.   Another economic development aspect that has been studied is access to low-wage employment opportunities.  Fan et al (2012) showed accessibility to low-income jobs improves after investment in LRT in the Twin Cities.  Rotger & Neilsen (2015) in Copenhagen showed that LRT can be a successful strategy in connecting isolated neighbourhoods, even if that means that some individuals are more likely to use the increased travel speeds to increase their travel to work from short (<5km) to more moderate (5-10km) trips.                                                   2 Gentrification has become an increasingly important concern with rail investment and transit-oriented development; however, very few empirical studies exist documenting the phenomena and most show mixed results. No studies surfaced using this project’s search methodology. For more information on current methodology for understanding the gentrification pressures of rail investment, please refer to these studies (Zuk et al. 2017; Dominie 2012; Pollack et al. 2010; Kahn 2007)    27 Using transit to support economic development is also seen as a strategy to increase employment and household/personal income.  On this front, evidence is consistent in the three studies that investigate this aspect.  Heres et al (2014) showed an increase in personal income, particularly for low and moderate income households, after installation of Bogata’s BRT. Rotger & Nielsen (2015) also showed an increase in income in Copenhagen after installation of a rail line.  Finally, Bardaka et al (2016) documented increase in household incomes within 2 miles of Denver’s LRT.  However, Bardaka et al (2016) showed that areas further from the LRT have income decreases, suggesting a locational shift in where low-income households are living instead of LRT investment as a way to supply more living wage jobs.  This suggests that studies need to more carefully track who moves in and out of investment areas to understand if low-income households are benefiting as much as moderate and high-income households.      In terms of study design, the studies investigating community development and personal income rely heavily on secondary data collection including: Census data (American Community Survey and Longitudinal Employer and Housing Dynamics (LEHD)); business and labor market data – some public and some proprietary; and land parcel data.  The reliance on secondary data meant that most studies use a spatial unit (i.e. census block) as the unit of analysis; exceptions were Heres et al in Bogata; Mansoor et al in Lahore; and Rotger & Nieslen in Copenhagen which used the household or person as the unit of analysis.  The studies used a range of statistical analyses and methods.  Regression models of different types and geospatial analyses were most likely to be implemented.  The primary methodological challenge in determining if transit is an effective instigator of economic development is the ability to tease out natural business and individual mobility patterns from those that can be attributed to the transit investment.  Some studies try and control for this by comparing the change in the study area near the transit investment with some type of control.  The long-term planning nature of fixed transit lines also contributes to this challenge as even in a pre-post setup, it is difficult to know if (1) businesses and/or households are anticipating the investment and (2) if the investment is merely augmenting a trend of increased density of businesses and households that suggested fixed rail could be viable.  Finally, studies that show increased personal income struggle to address the tensions surrounding gentrification.  Demonstrating increases in income could just be a demographic shift – particularly when relying on secondary, non-household-level data.  Understanding the likelihood of displacement of low-income households likely requires primary data collection inside the transit area and in “matched” non-transit communities to compare rates of household movement.    $%*, &$DF5-%%*1'.!/*,!$+*++!&'(-$,!'&&+!,1&&$-+-,'%+Authors; City; System Type (LRT, RRT, BRT) Study Design (A, B, C, D) Data Collection Tool; Recruitment Unit of Analysis Statistical Method Outcome; Direction of Findings (increase, decrease, no change) Bhattacharje et al., 2016; Denver; LRT & RRT Repeat cross-sectional with comparison (B) Secondary data, census & land parcels Census Tract Descriptive, inferential, Wilcoxon/Kruskal-Wallis test Population Density; increase    28 Bocarejo et al., 2013; Bogota; BRT Repeat cross-sectional with comparison (B) Secondary data, land parcels Planning Zone Hot spot analysis, difference in difference Population density; increase Calvo et al., 2013; Madrid; RRT Repeat cross-sectional with comparison (B) Secondary data, census Station Catchment (0.6 km) Descriptive Population density; increase Guerra, 2014; Mexico City; RRT Repeat cross-sectional without comparison (D) Secondary data, travel survey Household Binomial logit, ordinary least squares regression, Poisson regression Population density, increase Hurst, & West, 2014; Minneapolis,US; LRT Repeat cross-sectional with comparison (B) Secondary data, land parcels Land Parcel Difference in difference, propensity score matching, ordinary least squares regression  Land use; increase for multi-family, decrease for industrial Rodrigues, et al., 2016; Bogota & Quito; BRT Repeat cross-sectional with comparison (B) Secondary data, land parcels Station Catchment (1 km) Linear regression Land use; increase for residential, commercial Shen, 2013; 4 US Cities; LRT & RRT Repeat cross-sectional with comparison (B) Secondary data, census Census Block Group Spatial analysis, hotspot analysis, difference in difference, regression Population density; increase  Zhu & Diao, 2016; Singapore; RRT Repeat cross-sectional with comparison (B) Secondary data, travel survey & land parcels Household Difference in difference  Population density; increase   Bocarejo et al., 2013; Bogota; BRT Repeat cross-sectional with comparison (B) Secondary data, land parcels Planning Zone Hot spot analysis, difference in difference Land value; increase Brown & Werner, 2011; Salt Lake City; LRT Longitudinal  without comparison (C) Self-reported survey; Mail, residents Person T-test, general linear regression Land value; Self Report  Many of the economic benefits that accrue from rail investment are expressed through increased demand to be near transit resulting in increased population density.  Ten studies were included in this category (Table 13). Of these studies, two studies addressed this explicitly, both showing a positive association between rail and density. Shen’s (2013) multi-site study of medium to large metro areas in the USA indicated that density is facilitated by certain conditions: medium income neighborhoods; pre-existing compact land use; and good connectivity to the larger network. Calvo et al (2013) documented how a    29 rail line in Madrid with the explicit goal of encouraging more development was able to increase population by 3 times that of traditional rail areas and 4 times that of areas without rail. These studies employ secondary data. The buffer definitions are 500m (Bocarejo et al, 2013), 600m (Calvo et al, 2013) and 0.5 miles (Shen, 2013).  Analysis employs summary statistics and – in the case of Shen and Bocarejo – spatial analysis of hotspots as well as difference in differences regressions.   Transit lines, as a source of major public investment and as a sought after amenity, are thought to increase land values.  There is a large and robust hedonic (spatially focused) pricing literature that compares housing values near stations with housing values away from stations with and without longitudinal components (See Stokenberga (2014) for a comprehensive review).  Two studies included in this review used more rigorous research design (repeat cross-sectional and longitudinal) than what is more common in the hedonic pricing literature that relied on cross-sectional design. Bocarejo et al (2013) showed an increase in middle-income properties in the corridor with a premium value in some areas. Brown & Werner’s (2011) study of resident perceptions showed before and after a new light rail station showed that residents expected and then later perceived that the addition would increase land value and taxes would go up as a result.     ;2=2	#* +*&$)$DG5-%%*1'.!/*,!$+*++!&	$, -,'%+Authors; City; System Type (LRT, RRT, BRT) Study Design (A, B, C, D) Data Collection Tool; Recruitment Unit of Analysis Statistical Method Outcome; Direction of Findings (increase, decrease, no change) Park & Sener, 2017; Houston; LRT Longitudinal  with comparison (A) Secondary data, air monitoring  Catchment (16 km) Spatial analysis, zero inflated Poisson regression Stroke mortality; decrease  While there are numerous studies examining health-related impacts of public transportation (such as physical activity or air pollution reviewed in sections 3.1-3.3) we found only one pre-post LRT study on an actual health outcome (Table 14). This study, conducted by Park and Sener (2017) in Houston, Texas, reported that the new LRT line led to a 30% reduction (95% CI: -44.6 to -12.7) in stroke mortality. Different buffers were used to compare the treatment group (residents within 3, 5, and 10-mile buffers surrounding the LRT) and the control group (residents outside the treatment group buffer). The reduction in mortality was the largest (34%, 95% CI: -52.4 to -8.8) for the population within 5 miles of the new LRT line.   <2* &&#&!#&%(%) <29'*+(!%&#%!&%#$'*)Transit changes impact both residents living nearby where investments are made and those living elsewhere long a corridor or within a given region.  Relative to bus service, fixed-rail transit has the potential to change behaviour of populations living even further from the stations, leading to changes in    30 health behaviours in users of the wider system. Further, any changes in travel behaviour (e.g., VKT reductions or engagement in active transportation) in users of the system may have additional impacts on environmental outcomes (e.g., emissions) and health outcomes (physical activity benefits and/or traffic-related injuries).  These health and environmental impacts can be localized or be population wide.  Further, the direction and magnitude of some exposure-based impacts may differ in the corridor versus the region.  For example, increased density along the rail corridor may increase localized emissions but reduce overall VKT and improve regional air quality.  The geographic sampling of a given study design may reflect a local or regional orientation. Sampling approaches vary in the research designs we have summarized. Some pre-post studies use aggregate data over large areas primarily administrative areas such as census tracts or traffic analysis zones (Calvo, 2013 or Foth, 2014) across an entire city. Some studies compare entire corridors.  For Example Ewing at al compared a BRT with a Highway corridor). Often studies compare participants in areas adjacent to stations (cases) to participants located in areas further from transit (controls); but the scales at which these are defined vary considerably across studies.   Spatial method and scale to measure environments in which participants are located vary considerably.  Some studies use a road network buffer approach that only includes areas that can be accessed on the road network to more accurately capture the participants’ experience.  Some studies use simpler radial crow fly buffers that include areas and destinations that cannot be as readily accessed while others rely in census geography.   Size of buffers or scale of census geographies also vary widely.  They range from as small as 400 meter network buffers to several kilometers in size.  The study areas also varied: some studies focus on the line itself, while others include regional connections (e.g., Bogota); some look at the TransMilenio line itself (Coombs, 2014 and 2017) whereas others include the feeder routes as well (Bocarejo, 2013 and 2016). Other studies gather individual data and focus on a sample of people who live or work near the stations. In the studies we summarized, recruitment of individual populations may have been accomplished through door-to-door, telephone recruitment, mail-out, or workplace recruitment.    Methodological challenges for inclusion in pre-post study  A methodological decision point is how to capture local and/or regional impacts.  Different methods are required in terms of spatial sampling of participants and the primary (new) versus secondary (existing) data sources that are required. For example, longitudinal studies tracking individuals over time (e.g., using GPS or accelerometry) are costly.  Targeted sampling is required to recruit individuals living nearby the station who would be more likely to be influenced by changes; and to find participants of similar demographic characteristics who are exposed to similar land use and transportation conditions at baseline.   Studies with a focus on population-level impacts may use regional data on from air pollution monitors (Park, 2017), police reported collisions (Richmond, 2014) or travel survey data (Foth 2014, Ewing 2014).  Environmental outcomes studied to date have tended to be regional in nature relying on air pollution monitors.  Other study designs are possible that also measure localized impacts. Studies varied based on the type of data collection that was employed on participants ranging from objective monitoring, transportation related activity diaries, and longer term memory recall more commonly found in surveillance health surveys.     31  <2:2)&#*!%* $'*)&(%)!*%,)*$%*A challenge in all observational research is to disentangle the impacts due to a specific change (e.g., BRT investment) from broader changes that are occurring concurrently in the study area. Changes may relate to transportation and land use, such as investment in cycling infrastructure or re-development, or they may be demographic shifts such as gentrification, or economic factors such as housing prices, gas prices, and local economic shifts. Changes may also come in the form of new transportation programs and options targeted and shifting travel patterns such as car and bike sharing, Uber or Lyft, or road pricing being introduced while a study is underway.  Presence of a control group with similar demographic characteristics that is also exposed to the any changes underway within a given area but not exposed to the transit investment allows researchers to isolate the effect of the transit investment from other factors.  The research studies we present use different methods in their design and analysis to try to account for general trends and isolate the impact of transit investments through regression-based models. A common approach is to include variables in the regression model for factors that may be changing, for example, pulling administrative data on demographic changes from the census (Bardaka, 2016), on urban form or land use data (Hurst, 2014), on crime (Billings, 2011), weather (Goel, 2017), or jobs (Bocarejo, 2013). Additionally, the studies we summarized use different statistical analyses (differences in differences models, propensity scores) which may serve to better isolate impacts of specific changes.   However, in these observational studies there are often data gaps that may result residual confounding within a study. For example, in the studies with physical activity or traffic safety, there was no attempt to control for any complete street improvements that may have been undertaken. Spatially and temporally collected data on traffic volumes (pedestrian, bicyclist, or motor vehicles) is a critical gap. Issues of bias and unobservable characteristics are mentioned in the discussion section of most manuscripts.   <2;2'*+(!%!$'*)))&!*-!* ,+#%(#'&'+#*!&%)The impacts of urban form changes are not felt equally by all segments of the populations. Decision-makers may be particularly interested in impacts on more vulnerable populations, such as older adults, children, people with low incomes, or those who are currently transportation disadvantaged.   The vast majority of the studies we summarized focus on general population impacts. Overall there was a tendency to focus on commuter populations.  Several studies recruited only working age populations (e.g., Yanez 2010, Rotger 2015); no studies specifically recruited older adults or children. There are a handful of studies that do have a vulnerable populations lens, such as those with social outcomes. For example, Bardaka et al (2016) were interested in gentrification and displacement in Denver, and used census data to assess how income levels changed over time in areas nearby the BRT. This kind of ecological study design cannot draw conclusions at an individual level, for example, if the BRT was used by higher or lower income groups. Likewise, Bocajero et al (2016) assessed social fragmentation over time near TransMilenio and feeder routes in Bogota.        32 Methodological challenges for inclusion in pre-post study Indeed, any researchers interested in impacts in vulnerable populations needs to ensure these harder to reach populations are in fact captured part of the study sample. Some studies employ targeted qualitative methods such as focus groups, photovoice, and walk-along interviews to assess impacts in these groups. These are qualitative methods and as expected, we didn’t find any pre-post studies that reported using these approaches.  Some studies have employed geographically targeted sampling within lower income areas to increase representation of these communities. Second, population segments also could be captured through observational counts. Werner (2016) did conduct observational counts to assess changes in total ridership. However, no studies we reviewed considering the age and gender mix of users, and how that might shift over time or across sites. Third, using aggregated, administrative data makes it impossible to do stratified analysis by age, gender, or income. It may be possible if data can be spatially and demographically disaggregated. Further, studies where the unit of analysis is a unit of census geography (e.g. tract) or a traffic analysis zone with neighbourhood-level demographic data are at risk of ecological fallacy – where associations seen at in the aggregate data do not confer associations at the individual level. Further, children, youth, and people who are homeless or in institutional care may not be included in many administrative datasets.   <2<2))))!% ,!&+(# %%%,!(&%$%*#.'&)+(Of note, a BRT or rail investments do not occur in isolation; but are typically justified due to increases in travel demand from urban form changes. In addition, transit investments are typically accompanied with investments in the pedestrian environment.  For example, in Brown et al (2017) study of the track extension they explicitly acknowledge a complete streets intervention on the same corridor, and in Heinen and Olgivie (2015) describe their intervention as a guided busway and shared path. Yanez et al describe the very complicated implementation of the TransSantiago BRT, which included the simultaneous launch of the new BRT, a new way of running existing buses, and a new payment system. Therefore, multiple components are needed to launch a new system; but it makes it hard to isolate cause effect relationships.   <2=2&*!&%4)'!!&"*!,*&(.'&)+(% ,!&+(With technology developments in mobile sensing there are increasing possibilities for tracking participant movement. This is an important methodological development as it can provide objective data of use or exposure to urban form, as compared to self-report. For example, in studies interested in walking or cycling, GPS and accelerometry data from participants can be mapped in GIS to understand whether people use particular infrastructure (exposure), and how active they are (behaviour). A framework for such studies is described by Jankowsa et al (2015). A handful of the pre-post studies we summarized use these technologies (Brown 2015, Hong 2016, Huang 2017, Miller 2015). While costly, these methods provide more accurate and continuous measurement of human exposure and behaviour, as opposed to more traditional survey-based measures. National-level travel surveys (such as 2010-2012 California Household Travel Survey) are increasingly adopting these sensor-based measures, and with the declining cost of these sensors, some combination of traditional surveys and sensor-based measurements would be quite feasible and preferable for future intervention studies.   <2>2$'#)!0%'(*!!'%*(*%*!&%Maintaining an adequate sample size for longitudinal studies is critical. Small sample size increases the risk of type II errors (false negative) because it is hard to detect an effect with reduced sample size.    33 Although larger sample size is preferred over a small one, obtaining larger numbers of observation to increase statistical power can result in an unnecessarily expensive study. Therefore, a careful sample size calculation would be needed. No studies reviewed here offer methods to calculate appropriate sample size; however, some general guidelines are available to determine appropriate sample size (Lenth 2001; Motrenko et al. 2014; Kreidler et al. 2013).   Related to the problem of sample size, participant retention is also critical for a longitudinal study. Especially, a longitudinal study that follows the same person over time faces significant challenges in retaining the original sample size because participants often get dropped out of the survey or lose interests in participating in follow-up surveys. Typically, participants are given a financial compensation to keep engaged in the study for an extended period of time. For example, one study (Hong et al. 2016) offered participants a USD $30 gift card for the baseline survey and a $75 gift card for the follow-up survey. Another study (Durand et al. 2016) offered a USD $25 gift card for the basic survey and a $50 gift card for a more in-depth survey. Thinking about the financial compensation mechanism and effective sample retention strategy would be critical for conducting a multi-year longitudinal study that often suffers from significant attrition, in other words, loss of individual observations in the study over time.  2 92)*(*!)!%(4&)*!#*+!)A limited number of the studies incorporated rigorous quasi-experimental study designs that allow for control of other factors, which may have coincided with the transit intervention. Given this limitation, Table 15 shows a list of studies that can be considered as the most rigorous studies based on their study design.  $DH5,-!+, ,$$-&*, ,1(,'*1Authors; City; System Type (LRT, RRT, BRT) Study Design (A, B, C, D) Data Collection Tool; Recruitment Unit of Analysis Statistical Method Outcome; Direction of Findings (increase, decrease, no change) Park & Sener, 2017; Houston; LRT Longitudinal  with comparison (A) Secondary data, air monitoring  Catchment (16 km) Spatial analysis, zero inflated Poisson regression Stroke mortality; decrease Boarnet et al., 2017; Los Angeles; LRT Longitudinal  with comparison (A) Self-reported survey; Mail, residents Household Difference in difference GHG emissions; decrease Hong et al., 2016; Los Angeles; LRT Longitudinal  with comparison (A) Accelerometer, GPS, survey & 7-day travel diary; Mail, residents Person T-test, Fisher test, ANCOVA framework Physical activity; increase Spears et al., Longitudinal  Self-reported Household Descriptive, chi- Vehicle miles traveled;    34 2016; Los Angeles; RRT with comparison (A) 7-day travel diary; Mail, resident squared test, t-test, differences in difference decrease Werner et al., 2016; Salt Lake City; LRT Longitudinal  with comparison (A) Observational passenger counts Ridership  Bonferri adjustment, fixed effects repeated measures regression Mode; increased transit share, no displacement from bus share to LRT  Park and Sener (2017) used a rigorous research method as they compared the LRT impact of the treatment group against the control group over time and included measurement of an intermediate, air pollutant concentrations, with measures of health. Boarnet et al (2017) also used a rigorous method of comparing households living close and farther away from new transit, and employed a difference-in-differences approach to estimate changes in GHG emissions. Hong et al (2016) used a true longitudinal design to follow individuals’ walking and physical activity behavior over time. Furthermore, Hong et al (2016) used multiple methods to corroborate their findings using measurements from surveys, GPS, and accelerometer. Werner et al (2016) and Spears et al (2017) also employed a rigorous treatment and control design with primary data collection before and after the intervention on the same cohort. The repeated measure mixed effects model employed by Werner et al (2016) is also an advancement over a traditional difference-in-difference (DID) approach because it addresses spatial clustering and shared built environment inputs for the individuals who are reporting their travel behaviour over time.   Based on the synthesis of the papers and their methods reviewed, future studies around the MLBE should consider evaluating both behavioural and exposure based impacts on health outcomes resulting from transit investment. It is also important to assess the health impacts of both transit and related non-motorized infrastructure investments, and the impacts on both dependent and choice riders. Future studies can also leverage and integrate existing data including administrative and other newly collected health survey data with newly collected data into study designs. Given the high cost of living and issues around housing affordability in Metro Vancouver, it would be highly desirable to evaluate any displacement that is observed and the relative health impacts of changes in residential environments due to displacement.    :2)( ')%''&(*+%!*!)There is a wide variation in the methods used to study the community-level health impacts of fixed guideway transit investment.  This makes it difficult to compare studies. Some of the key methodological gaps and opportunities have been identified.  First, transit intervention studies that explicitly focus on health outcomes are very limited. Published literature links physical activity with decreased body mass index, chronic disease and premature death. While several studies are underway that make these linkages with data collection nearing completion; published evidence showing the linkage between transit investments and direct health outcomes is limited to only one study. This presents opportunities for scientific research around the MLBE.      35 Second, study design, sample size, control groups, and other methodological choices differ depending on outcomes, study population, and geographic scale from neighbourhood, corridor, to regional impacts.  Notably, there were inherent differences in types of transit investments and stages of system development. Some studies reviewed were of BRT and others of rail and some were of a region’s initial rail investment versus others connecting a corridor to an already developed rail network offering considerable travel time savings to major destinations. Differences in methodological approaches to study designs also present challenges for comparing one study to another. Studies used different spatial scales and approaches to geographic buffering of a sample population around transit corridor, ranging from buffering around rail stations to buffering of entire lines and feeder routes.   There were also considerable variations in statistical methods employed, and outcome measurements (e.g. self-reported versus objective measurement). In addition, different impacts for dependent versus choice riders were most often not directly assessed. Impact of new transit investment may differ between people relying on public transit (dependent riders) versus people that own a car and thus not fully dependent on public transit (choice riders). A few studies examined the impact of car ownership on travel behavior while the majority did not, making it difficult to compare the results. Methodological considerations for further studies include: • Isolating and/or assessing synergistic impacts of rail and related land use changes, pedestrian infrastructure investments, transportation programs (e.g. car and bike sharing) requires explicit methodological consideration.  • Existing data may not address attitudinal considerations, methodological concerns, and provide the ability to control for other factors. • Objective data on physical activity and travel patterns are key features in most state of the art rail interventions studies underway. • Objective real time “on-person” spatial assessment of exposure to air quality has not yet been conducted in any rail intervention study.  Third, few studies of health-related outcomes have isolated the impact of transit from other factors. Longitudinal studies with treatment and control groups typically employ statistical approaches to assess the “causal” impact of an intervention on a defined population. Such studies emulate a randomized controlled trial with a “control” population that is comparable to the “treatment” or intervention population except for not being exposed to a rail investment.  Repeated one time cross-sectional surveys on different samples and even longitudinal designs on the same population without a control are subject to bias and/or confounding effects of other factors. An ideal study design would be a true longitudinal design that follows the same individuals over time with a sufficient time lag before and after a transit intervention. Alternatively, nixing several types of study designs is possible and can leverage existing data on larger samples while conducting targeted primary data collection.  This has been done in a few locations (e.g. Salt Lake, Utah; Los Angeles, California; and Cambridge, UK) to study health impacts of rail investment and can be a cost effective and informative approach.   Fourth, as illustrated by the pathway diagram (Figure 1), the relationship between a transit investment and measurable health outcomes is complex and includes multiple steps. As one moves down the causal pathway from the transportation investment to a measurable impact on population or personal health the signal to noise ratio decreases as other factors (including those related to intermediates) become increasingly important. As such, for each step along the causal pathway there is a need to control for    36 additional potentially confounding factors. Studies that attempt to measure all steps along the causal pathway at the population level are inherently complex and it may be preferable to use quasi-experimental designs (for example those using population-based administrative data) that isolate the intervention and the outcome only without assessment of any intermediate factors. For example, comparing with pre-post measurement of a specific health measure in a population affected by a transit investment with an (ideally) identical geographically proximal population not affected by the investment.  This can be a rigorous approach to assess the direct impact of the investment but it provides no information on the mechanism or how the transit “exposure” may be mediated by intermediate factors. Alternatively it may be preferable to isolate specific links within the causal chain and to study those independently (e.g. relationship between transit investment and travel behaviour, between travel behaviour and specific exposures, between exposures and biological responses, or between biological responses and health outcomes). While the latter approach may be more straightforward and allows for more detailed understanding of complexities within the casual pathway, it is usually only possible to indirectly assess the overall link between the upstream intervention and the health outcome.   As another approach, longitudinal studies of individuals in which there is a gradient in “exposure” to the transit investment and in which the full suite of intermediate factors can be measured may be feasible, but only for small study populations. Such studies would likely be more costly as they would require detailed measurement and interactions with participants. Participant drop-out during follow-up, which may be different by age, socioeconomic status, etc. may also be a risk in such prospective study designs which require intense participation.   Fifth, studies that capture potential benefits from physical activity could also include potential risks from exposure to air pollution, noise, and traffic injuries.  Capturing both behavioral and exposure-based impacts can help understand the separate and combined effect of these factors on several health outcomes, such as cardiovascular disease and diabetes.  Failure to consider both behaviours and exposures may result in the underestimation of potential health benefits which are offset from related impacts of air pollution, noise, and injury risk stemming from densification around transit stations. For the evaluation of potential risks, it would be critical to capture vulnerable underserved populations potentially displaced by transit, and capturing the displacement effect would require targeted geographic sampling.  This population is hard to reach and to retain over time, which is a requirement of a longitudinal study design.  Sixth, no studies reviewed here documented health impacts related to social aspects of health, such as social cohesion and social capital. Although topics related to the built environment and social capital have been explored extensively in the literature, limited studies exist to date that examined community-level impact of public transit (Utsunomiya 2016; Kamruzzaman et al. 2014). In further studies, it would be useful to investigate the role of public transit as “social equalizer”, i.e. the role of public transit on enhancing social capital and community resilience, as they relate to health and wellbeing.   Lastly, there are opportunities for analysis of potential health care cost savings benefits in relation to transit investment.  This topic is the focus of several on-going studies in the U.S. One example includes a study in Portland, Oregon being co-led by Kaiser Permanente and the Principal Investigator of this review. Other related studies are underway in Seattle, Washington, and Austin, Texas, but no such study of this nature is underway in Canada.  Employing some similar research methods would allow    37 comparative work to be done between these similarly sized regions with contrasting political structures and health care systems.   ;2!$!**!&%)This systematic review relied on search results from more than 8 journal database systems in conjunction with specific search terms to select the final studies to be included in the review. Although the search terms used to filter the studies were carefully chosen and modified through extensive discussions among the review authors, it is possible that the final selection may not include some papers that use ambiguous keywords or terms outside the disciplines that the authors are familiar with. In addition, some of the newer articles just published or forthcoming may not have made it to the large database systems during our formal search process. The authors of this review were aware of this issue regarding the time delay with newer publications and carefully included newer articles that met our criteria but did not come through during the formal search.    2 We reviewed 52 papers on pre- and post- test impacts of new rapid transit interventions. Few studies have been conducted to date and those that have been done use a wide array of study designs with varying degrees of rigor.  The few studies that have used a control group and retained the same participants before and after the transit improvement happened have found significant impacts on physical activity, GHG emissions, and pollutant concentrations. Economic impacts of transit on land value and wages are somewhat clearer, but the question for whom these benefits accrue remains unclear.  The potential for gentrification is a critical issue in Vancouver and the potential of displacement and increased hardship for vulnerable populations is considerable. Going forward, a carefully designed longitudinal study around the MLBE project addressing some, if not all, methodological issues raised in this review will provide credible evidence on health-related impacts of transit investment and to help inform future decisions about transit investments in Metro Vancouver and beyond.  3895	!-*F6* (*'++'*!&,!1!&+,-!++'&*+* +!&''$& *++-+5D6 <?*(!,*&+!,?7-+*(!,*&+!,87(-$!,*&+!,8?$! ,*!$?7%,*'*!$8+-/1=6 < $, 7( 1+!$,!.!,187'1%++!&087 *'&!!++87!*('$$-,!'&8&'!+7,*!+,187*'+,18*!%!&"-*>%'!$!,1++!!$!,17$&-+8=6 <$'&!,-!&$*,*'+(,!.(*'+(,!.7&,-*$0(*!%&,80(*!%&,$!&,*.&,!'&0(&+!'&=6 <(+,*!&>%($'1>*+!&,>'/&*>=   39 :5	4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	,*,(*'*%'& !$%'*,$!,1!&, &!,,,+:,+''&!,!'&$+ ,*&+*+'&*,+''.*/! ,9'+!,1!&0!'&+,*-%&,$.*!$+.*!$++'!,/!, 0('+-*,', !&,*.&,!'&3-,&',/!, ', *,'*+++'!,/!, , '-,'%'!&,*+,3-+,'%'$, ,', !&,*.&,!'&:&!&+,*-%&,$.*!$, ,+,!+!+, +++-%(,!'&++ '-$(*'.!&-&'&'-&+,!%,', ,', !&,*.&,!'&:- .*!$+***3&&',$$', ++-%(,!'&+&,+,!*,$1:,'''+,%(+'&''!&+-*!,1:,''%%-&!,1+$'&+'&+'!$(*,!!(,!'&&+$:*, $, %'&'$*('($!&(&Note: This table is adapted with permission from Craig et al. (2017)    41 Author, year; location Type of Rapid Transit Study Design* Prospective (P) or Retrospective (R) Spatial Scale (km) Specific Population Data Collection Tool Recruitment Unit of Analysis Sample Size (n in analysis OR n T1/T2) Study Duration (years) Statistical Methods Outcome(s); Direction of Findings (increase, decrease, no change) Bardaka, et al., 2016; US LRT  D R  3.2   Low- income Secondary data, census - Census Block Group NA  9 Spatial analysis, global & local Moran's I, ordinary least squares regression, LaGrange multiplier test Per capita income; increase Bhattacharjee, & Goetz, 2016; US LRT & RRT B R  0.8   - Secondary data, census & land parcels - Census Tract  144  10 Descriptive, inferential, Wilcoxon/Kruskal-Wallis test Land use; increase for retail density Billings, et al., 2011; US LRT  B R  0.8   - Secondary data, police report - Station catchment (0.4 km) 41  11 Difference in difference Crime; decrease Boarnet, et al., 2017; US RRT  A P  0.8   - Self-reported survey Mail, residents Household  285/208  1 Difference in difference GHG emissions Bocarejo, et al. 2016; CO BRT   B R  0.5    Disadvantaged Secondary data, travel survey & census - Planning Zone 112  10 Spatial change analysis, difference in difference, multiple linear regression Accessibility; increase for low income, no change for high income Bocarejo, et al., 2013; CO BRT  B R  0.5   - Secondary data, land parcels - Planning Zone 112  7 Hot spot analysis, difference in difference Population density; increase; Land value; increase Bocarejo, et al., 2012; CO BRT   D R  n/a  - Secondary data, police report - Road Segment 10  10 Hot spot analysis Injury or fatal collisions; decrease Brown, et al., 2017; US LRT  C P  0.4   - Accelerometer,  GPS, & survey Door to door Person  939/536/357 1 Regression  BMI; decrease for new riders, increase for former riders Brown, & Werner, 2007; US LRT  C P  0.8   Low- income, mixed Accelerometer & survey Mail, residents Person  102/51  1 Ordinary least squares regression Physical activity; increase; Mode; increased transit    42 ethnicity share Brown, & Werner,  2009; US LRT C P 0.8  Low-income, mixed ethnicity Accelerometer, interviews &  survey Mail, residents Person 102/51 1 Regression Obesity; decrease for new riders Brown, & Werner, 2011; US LRT C P 0.8  Low-income, mixed ethnicity Self-reported survey Mail, residents Person 102/51 1 T-test, general linear regression Residential perspectives; increase in safety, decrease in crime perceptions Brown, et al., 2015; US LRT C P 0.4  - Accelerometer, GPS & survey Door to door Person 939/536 1 One-way analysis of variance, ordinary least squares regression Physical activity; increased for new riders, decreased for former riders; BMI; decrease for new riders, increase for former riders Calvo, et al., 2013; ES RRT B R 0.6  - Secondary data, census - Station Catchment (0.6 km) 29 11 Descriptive Population density; increase Chang, et al., 2017; MX BRT D P 0.5  Subgroups by gender, employment, and education Self-reported survey Phone, random residents Person 1067/1420 3 Propensity score matching, 2-step cluster analysis, k-means cluster analysis  Physical activity; increase in walking, bicycling, greatest for females with low education Chen, & Whalley, 2012; TW RRT B R n/a  - Secondary data, air monitoring  - City 3 2 Ordinary least squares regression Air pollution; decrease Combs, 2017; CO BRT B R 0.8  Lower wealth households Secondary data, travel survey - Household 10090/ 10222  10 Difference in difference, Poisson regression, partial proportional odds parameterization Accessibility; no change  Combs, & Rodriguez, 2014; CO BRT B R 0.4 & 0.8  Lower-mid wealth households Secondary data, travel survey - Household 9107/9330 10 Difference in difference, logistic regression Vehicle ownership; decrease for high income, increase for low income Delmelle, & Casas, 2012; CO BRT D R 0.4, 0.8, 1.2, & 1.6  Low income strata Secondary data, census - Neighbourhood n not reported 1 Spatial analysis Accessibility; increased to hospital, higher for middle income Ding, et al., 2016; TW RRT A R n/a  - Secondary data, air - City 4 19 Spearman correlation, Air pollution; decrease    43 monitoring  Kruskal-Wallis test, Wilcoxon rank sum test, Friedman test, logistic regression  Durand, et al., 2016; US LRT C P 4.8  African-American & Hispanic Accelerometer, survey, travel diary & neighbourhood audits Mail, residents Person 750 4 Latent Growth Curve model, piecewise model, full information maximum likelihood estimation, propensity scores Physical activity; no results; Mode; no results (protocol paper) Engebretsen, et al., 2017; NO LRT B R 1  - Secondary data, travel survey - Person 9683/902/3000/10570/4205 6 Spatial analysis, logistic regression  Mode; increased transit share Ewing, & Hamidi, 2014; US LRT B R 2  - Secondary data, travel survey - Household 302/605 17 Difference of means test, 2-stage models Vehicle miles traveled; decrease Fan, et al., 2012; US LRT D R 0.4  Low wage workers Secondary data, census - Census Block 22588 4 Spatial analysis, ordinary least squares regression Accessibility; increase to jobs Foth, et al., 2014; CA RRT D R 1  Lowest decile of  disadvantaged Secondary data, travel survey & census - Traffic Analysis Zone 935 10 Multiple linear regression Mode; no change overall, increased transit share to shop/work, increased transit share for socially disadvantaged Goel, & Gupta, 2017; IN RRT D R n/a - Secondary data, air monitoring  - Line extension 4 2 Ordinary least squares regression Air pollution; decrease Guerra, et al., 2014; MX RRT D R 1  - Secondary data, travel survey - Household N/A 13 Binomial logit, ordinary least squares regression, Poisson regression Mode; increased transit share; Population density; increase Heinen, & Ogilvie, 2016; UK BRT C P  30  Work near intervention Self-reported survey Mail, employees Person 1143/469 3 Multinomial logistic regression Mode; increased active travel share, decreased car share Heinen, & BRT C P 30  Work near Self-reported Mail, Person 1143/469 3 Multinomial Mode; no change    44 Ogilvie, 2017; UK intervention survey  employees logistic regression Heres, et al., 2014; CO BRT B R 0.75 & 1.5  Employed individuals Secondary data, census - Household ~2000 5 Weighted least squares regression, probit models Income; increase, highest for low and middle income strata Hong, et al., 2016; US RRT A P 0.8  - Accelerometer, GPS, survey & 7-day travel diary Mail, residents Person 204 survey/ 73 GPS 1 T-test, Fisher test, ANCOVA framework Physical activity; increase Huang, et al., 2017; US LRT C P 1.6  - Accelerometer, GPS, survey & travel diary Phone, residents Person 214/198 2 Negative binomial regression Walking;  decrease overall, increase within 0.4 km Hurst, & West, 2014; US RRT B R 0.8  - Secondary data, land parcels - Land Parcel 7635 13 Difference in difference, propensity score matching, ordinary least squares regression  Land use; increase for multi-family, decrease for industrial Lee, &  Senior, 2013; UK LRT B R 0.6  Working households Secondary data, census - Ward N/A 10 Descriptive Vehicle ownership; increase; Mode; increased  LRT share, decreased bus share MacDonald, et al., 2010; US LRT C P 1.6  - Self-reported survey Phone, residents Person 839/498 2 Logistic regression, propensity score matching Physical activity; increase;  BMI; decrease for LRT users Mansoor, et al., 2016; PK BRT D P n/a  - Self-reported survey Intercept Person 250/250 1 Descriptive Residential perspectives; increased positive perceptions; increased jobs, decreased gender discrimination Miller, et al., 2015; US LRT C P 0.4 - Accelerometer, GPS & survey Door to door Person 939/536 1 Descriptive, paired t-tests, spatial analysis Physical activity; increase for new riders, decrease for former riders Nelson, et al., 2016; US BRT D R 0.4  Employees Secondary data, InfoGroup - Station Catchment (0.4 km) 2 6 Descriptive New jobs; increase Panter, et al., 2016; UK BRT  C P 30  - Self-reported survey Mail, employees Person 1143/469 3 T-test, chi-squared test, signed rank test, multinomial Physical activity; increase for bicycling, transport    45 logistic regression  Park, & Sener, 2017; US LRT A R 4.8, 8, 16  - Secondary data, air monitoring  - Catchment (16 km) 5 3 Interrupted time series analysis, Poisson regression Stroke mortality; decrease; Air pollution, decrease Richmond, et al., 2014; CA Streetcar D R 0.025  Pedestrians Secondary data, police report - Road Segment 3 11 Spatial analysis, zero inflated Poisson regression Pedestrian motor vehicle collisions; decrease Ridgeway, & MacDonald, 2017; US RRT D R 1 - Secondary data, police report - Police Reporting District  281 26 Stepped wedge design, Poisson regression, permutation tests Crime; decrease Rodrigues, et al., 2016; CO & EC BRT B R 0.5 & 1  - Secondary data, land parcels - Station Catchment (1 km) 18 10 Linear regression Land use; increase for residential, commercial Rotger, & Nielsen, 2015; DK  RRT B R 0.5, 2.7, & 6.2  - Secondary data, public register - Household 8374 11 Differences in difference regression Earnings; increase Saxe, et al., 2015; CA RRT D P 3.2 - Secondary data, travel survey & census - Dissemination Area n not reported 12 Descriptive, correlation Mode; increased RRT share, decreased bus, car share; GHG emissions; decrease Saxe, et al., 2017; CA RRT B R 3.2 - Secondary data, travel survey & census - Station Catchment (3.2 km) 5 9 GHG savings analysis GHG emissions; decrease Shen, 2013; US RRT B R 0.8 - Secondary data, census - Census Block Group 1897 10 Spatial analysis, hotspot analysis, difference in difference, regression Population density; increase Spears, et al., 2016; US RRT A P 1 & 5 - Self-reported 7-day travel diary Mail, residents Household 285/208 2 Descriptive, chi-squared test, t-test, differences in difference Vehicle miles traveled; decrease Werner, et al., 2016; US LRT A P 0.4 - Observational passenger counts - Station Catchment (0.4 km) 13 1 Bonferri adjustment; fixed effects repeated measures regression Mode; increased transit share, no displacement from bus share to LRT    46 Xie, 2016; CN  RRT  B R  n/a  - Secondary data, travel survey - Traffic Analysis Zone 71  2 Difference in difference Mode; increased rail, walking, bike share, decreased auto share Yanez, et al., 2010; CL BRT  C P  n/a  Employees  Interviews, 5-day travel diary Mail, employees Person  258  2 Winner-looser analysis(descriptive) Mode; increased rail, rail-bus share, decreased bus Yarbrough, 2014; US RRT  B R  0.4 & 0.8 Businesses  Interviews & secondary data, census Phone, businesses Station Catchment (0.8 km) 5  8 Spatial analysis, nearest neighbour ratios, Geary's C analysis, bivariate correlation, multiple regression, ANOVA Business locations; increase Zhu & Diao, 2016; SN RRT  B P  0.5 & 1   High income households Secondary data, travel survey & land parcels - Household  142/70  4 Difference in difference  Mode; increased rail share, decreased car share; Density; increase for residential  A) longitudinal with comparison; B) repeat cross-sectional with comparison; C) longitudinal with no comparison; D) repeat cross-sectional with no comparison   47  Bardaka, E. et al., 2016. Assessing the Socioeconomic Effects of Rail Transit Development: An Exploration of Displacement and Gentrification. In TRB 95th Annual Meeting. Available at: https://trid.trb.org/view.aspx?id=1393319. Beaudoin, J., Farzin, Y.H. & Lin Lawell, C.-Y.C., 2015. Public transit investment and sustainable transportation: A review of studies of transit’s impact on traffic congestion and air quality. Research in Transportation Economics, 52, pp.15–22. Available at: http://dx.doi.org/10.1016/j.retrec.2015.10.004. Besser, L. & Dannenberg, A.L., 2005. Walking to public transit: steps to help meet physical activity recommendations. American journal of preventive medicine, 29(4), pp.273–280. Bhattacharjee, S. & Goetz, A.R., 2016. The rail transit system and land use change in the Denver metro region. Journal of Transport Geography, 54, pp.440–450. Available at: http://www.sciencedirect.com/science/article/pii/S096669231600017X. Bigazzi, A.Y. & Figliozzi, M. a., 2014. Review of Urban Bicyclists’ Intake and Uptake of Traffic-Related Air Pollution. Transport Reviews, 34(2), pp.221–245. Billings, S.B., 2011. Estimating the value of a new transit option. Regional Science and Urban Economics, 41(6), pp.525–536. Bocarejo, J.P. et al., 2012. Impact of Bus Rapid Transit Systems on Road Safety. Transportation Research Record: Journal of the Transportation Research Board, (2317), pp.1–7. Available at: http://trrjournalonline.trb.org/doi/abs/10.3141/2317-01. Bocarejo, J.P., Portilla, I. & Meléndez, D., 2016. Social fragmentation as a consequence of implementing a Bus Rapid Transit system in the city of Bogotá. Urban Studies, 53(8), pp.1617–1634. Available at: http://journals.sagepub.com/doi/abs/10.1177/0042098015588739. Bocarejo, J.P., Portilla, I. & Pérez, M.A., 2013. Impact of Transmilenio on density, land use, and land value in Bogotá. Research in Transportation Economics. Brown, B.B. et al., 2017. Transit Rider Body Mass Index Before and After Completion of Street Light-Rail Line in Utah. American Journal of Public Health, 107(9), pp.1484–1486. Available at: http://ajph.aphapublications.org/doi/10.2105/AJPH.2017.303899. Brown, B.B. et al., 2015. Transit Use, Physical Activity, and Body Mass Index Changes: Objective Measures Associated With Complete Street Light-Rail Construction. American Journal of Public Health, 105(7), pp.1468–1474. Available at: http://ajph.aphapublications.org/doi/10.2105/AJPH.2015.302561. Brown, B.B. & Werner, C.M., 2007. A new rail stop: tracking moderate physical activity bouts and ridership. American journal of preventive medicine, 33(4), pp.306–9. Brown, B.B. & Werner, C.M., 2008. Before and After a New Light Rail Stop: Resident Attitudes, Travel Behavior, and Obesity. Journal of the American Planning Association, 75(1), pp.5–12. Brown, B.B. & Werner, C.M., 2011. The Residents’ Benefits and Concerns Before and After a New Rail Stop: Do Residents Get What They Expect? Environment and Behavior, 43(6), pp.789–806. Available at: http://journals.sagepub.com/doi/abs/10.1177/0013916510392030. Calvo, F.J., De Oña, J. & Arán, F., 2013. Light Rail Transit Experience in Madrid, Spain: Effects on Population Settlement and Land Use. Transportation Research Record Journal of the Transportation Research Board Transportation Research Board of the National Academies, (2353), pp.82–91.    48 Caspi, C.E. et al., 2012. The local food environment and diet: A systematic review. Health & Place, 18(5), pp.1172–1187. Available at: http://linkinghub.elsevier.com/retrieve/pii/S1353829212001037. Chang, A. et al., 2017. The effect of BRT implementation and streetscape redesign on physical activity: A case study of Mexico City. Transportation Research Part A: Policy and Practice, 100, pp.337–347. Available at: http://www.sciencedirect.com/science/article/pii/S0965856415301324. Chen, Y. & Whalley, A., 2012. Green Infrastructure: The Effects of Urban Rail Transit on Air Quality. American Economic Journal: Economic Policy, 4(1), pp.58–97. Available at: http://pubs.aeaweb.org/doi/abs/10.1257/pol.4.1.58. Combs, T.S., 2017. Examining changes in travel patterns among lower wealth households after BRT investment in Bogotá, Colombia. Journal of Transport Geography, 60, pp.11–20. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0966692317301060. Combs, T.S. & Rodriguez, D.A., 2014. Joint impacts of Bus Rapid Transit and urban form on vehicle ownership: New evidence from a quasi-longitudinal analysis in Bogot??, Colombia. Transportation Research Part A: Policy and Practice, 69, pp.272–285. Available at: http://www.sciencedirect.com/science/article/pii/S0965856414002055. Craig, P. et al., 2017. Natural Experiments: An Overview of Methods, Approaches, and Contributions to Public Health Intervention Research. Annual Review of Public Health, 38(1), pp.39–56. Available at: https://www.ncbi.nlm.nih.gov/pubmed/28125392. Delmelle, E.C. & Casas, I., 2012. Evaluating the spatial equity of bus rapid transit-based accessibility patterns in a developing country: The case of Cali, Colombia. Transport Policy, 20, pp.36–46. Ding, P.-H. et al., 2016. Urban Air Pollution in Taiwan before and after the Installation of a Mass Rapid Transit System. Journal of Environment Quality, 45(5), pp.1731–1739. Available at: https://dl.sciencesocieties.org/publications/jeq/abstracts/45/5/1731. Dominie, W., 2012. Is Just Growth Smarter Growth?: The Effects of Gentrification on Transit Ridership and Driving in Los Angeles’ Transit Station Area Neighborhoods. UCLA. Durand, C.P. et al., 2016. The Effect of Light Rail Transit on Physical Activity: Design and Methods of the Travel-Related Activity in Neighborhoods Study. Frontiers in Public Health, 4. Edwards, R.D., 2008. Public transit, obesity, and medical costs: assessing the magnitudes. Preventive medicine, 46(1), pp.14–21. Engebretsen, Ø., Christiansen, P. & Strand, A., 2017. Bergen light rail – Effects on travel behaviour. Journal of Transport Geography, 62(May 2016), pp.111–121. Available at: http://dx.doi.org/10.1016/j.jtrangeo.2017.05.013. Ewing, R. & Hamidi, S., 2014. Longitudinal Analysis of Transit’s Land Use Multiplier in Portland (OR). Journal of the American Planning Association, 80(2), pp.123–137. Available at: http://www.tandfonline.com/doi/abs/10.1080/01944363.2014.949506. Fan, Y., Guthrie, A.E. & Levinson, D.M., 2012. Impact of light rail implementation on labor market accessibility: A transportation equity perspective. Journal of Transport and Land Use, 5(3). Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1736206. Foth, N., Manaugh, K. & El-Geneidy, A., 2014. Determinants of Mode Share over Time. Transportation Research Record: Journal of the Transportation Research Board, 2417(2417), pp.67–77. Available at: http://trrjournalonline.trb.org/doi/10.3141/2417-08. Frank, L.D. et al., 2008. A hierarchy of sociodemographic and environmental correlates of walking and obesity. Preventive medicine, 47(2), pp.172–8. Available at: http://www.sciencedirect.com/science/article/pii/S0091743508002041 [Accessed July 18, 2014].    49 Frank, L.D. et al., 2010. Carbonless footprints: promoting health and climate stabilization through active transportation. Preventive medicine, 50 Suppl 1, pp.S99-105. Available at: http://www.ncbi.nlm.nih.gov/sites/entrez?Db=pubmed&DbFrom=pubmed&Cmd=Link&LinkName=pubmed_pubmed&LinkReadableName=Related Articles&IdsFromResult=19850071&ordinalpos=3&itool=EntrezSystem2.PEntrez.Pubmed.Pubmed_ResultsPanel.Pubmed_RVDocSum. Freeland, A.L. et al., 2013. Walking associated with public transit: moving toward increased physical activity in the United States. American journal of public health, 103(3), pp.536–42. Friedman, M.S. et al., 2001. Impact of Changes in Transportation and Commuting Behaviors During the 1996 Summer Olympic Games in Atlanta on Air Quality and Childhood Asthma. The Journal of the American Medical Association, 285(7), pp.897–905. Available at: http://jama.ama-assn.org/cgi/doi/10.1001/jama.285.7.897 [Accessed September 17, 2011]. Goel, D. & Gupta, S., 2015. The Effect of Metro Expansions on Air Pollution in Delhi. The World Bank Economic Review, 31(1), pp.271–294. Available at: https://academic.oup.com/wber/article/2897298. Grengs, J., 2015. Nonwork accessibility as a socila equity indicator. International Journal of Sustainable Transportation, 9(1), pp.1–14. Guerra, E., 2014. Mexico City’s suburban land use and transit connection: The effects of the Line B Metro expansion. Transport Policy, 32, pp.105–114. Available at: http://www.sciencedirect.com/science/article/pii/S0967070X14000079. Ham, W. et al., 2017. Commuter exposure to PM 2.5 , BC, and UFP in six common transport microenvironments in Sacramento, California. Atmospheric Environment, 167, pp.335–345. Available at: http://linkinghub.elsevier.com/retrieve/pii/S1352231017305290. Hansen, W.G., 1959. How Accessibility Shapes Land Use. Journal of the American Institute of Planners, 25(2), pp.73–76. Available at: http://www.tandfonline.com/doi/abs/10.1080/01944365908978307. Heinen, E. et al., 2017. Does exposure to new transport infrastructure result in modal shifts? Patterns of change in commute mode choices in a four-year quasi-experimental cohort study. Journal of Transport & Health, 6(May), pp.396–410. Available at: http://linkinghub.elsevier.com/retrieve/pii/S221414051730155X. Heinen, E. & Ogilvie, D., 2016. Variability in baseline travel behaviour as a predictor of changes in commuting by active travel, car and public transport: a natural experimental study. Journal of Transport & Health, 3(1), pp.77–85. Available at: http://www.sciencedirect.com/science/article/pii/S2214140515006933. Heres, D.R., Darby, J. & Salon, D., 2009. Do public transport investments promote urban economic development? Evidence from bus rapid transit in Bogotá, Colombia. Transportation, 41(1), pp.57–74. Hong, A., Boarnet, M.G. & Houston, D., 2016. New light rail transit and active travel: A longitudinal study. Transportation Research Part A: Policy and Practice, 92, pp.131–144. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0965856416306309. Huang, R. et al., 2017. Light rail leads to more walking around station areas. Journal of Transport & Health. Available at: http://www.sciencedirect.com/science/article/pii/S2214140516303899?via%3Dihub. Hurst, N.B. & West, S.E., 2014. Public transit and urban redevelopment: The effect of light rail transit on land use in minneapolis, minnesota. Regional Science and Urban Economics, 46, pp.57–72.    50 Jankowska, M.M., Schipperijn, J. & Kerr, J., 2015. A Framework for Using GPS Data in Physical Activity and Sedentary Behavior Studies. Exercise and Sport Sciences Reviews, 43(1), pp.48–56. Available at: http://content.wkhealth.com/linkback/openurl?sid=WKPTLP:landingpage&an=00003677-201501000-00009. Julien, D. et al., 2015. Transit use and walking as potential mediators of the association between accessibility to services and amenities and social participation among urban-dwelling older adults: Insights from the VoisiNuAge study. Journal of Transport & Health, 2(1), pp.35–43. Available at: http://linkinghub.elsevier.com/retrieve/pii/S2214140514000759. Kahn, M.E., 2007. Gentrification Trends in New Transit-Oriented Communities: Evidence from 14 Cities That Expanded and Built Rail Transit Systems. Real Estate Economics, 35(2), pp.155–182. Available at: http://doi.wiley.com/10.1111/j.1540-6229.2007.00186.x. Kamruzzaman, M. et al., 2014. Patterns of social capital associated with transit oriented development. Journal of Transport Geography, 35, pp.144–155. Kreidler, S.M. et al., 2013. GLIMMPSE: Online power computation for linear models with and without a baseline covariate. Journal of Statistical Software, 54(10), pp.1–26. Lachapelle, U. et al., 2016. Active Transportation by Transit-Dependent and Choice Riders and Potential Displacement of Leisure Physical Activity. Journal of Planning Education and Research, 36(2), pp.225–238. Lachapelle, U., Frank, L.D. & J, 2009. Transit and health: Mode of transport, employer-sponsored public transit pass programs, and physical activity. Health Policy 30 Suppl, 1, pp.S73-94. Lee, S.S. & Senior, M.L., 2013. Do light rail services discourage car ownership and use? Evidence from Census data for four English cities. Journal of Transport Geography, 29, pp.11–23. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0966692312002943. Lenth, R. V, 2001. Some Practical Guidelines for Effective Sample Size Determination. The American Statistician, 55(3), pp.187–193. MacDonald, J.M. et al., 2010. The effect of light rail transit on body mass index and physical activity. American journal of preventive medicine, 39(2), pp.105–112. Mansoor, A., Zahid, I. & Shazahd, L., 2016. Evaluation of Social and Environmental Aspects of Lahore Metro Bus Transit Through Public Opinion. Journal of Environmental Science and Management, 19(2), pp.27–37. Available at: https://www.journals.uplb.edu.ph/index.php/JESAM/article/view/1554. McAndrews, C. et al., 2017. Understanding and Improving Arterial Roads to Support Public Health and Transportation Goals. American Journal of Public Health, 107(8), pp.1278–1282. Miller, H.J. et al., 2016. Public transit generates new physical activity: Evidence from individual GPS and accelerometer data before and after light rail construction in a neighborhood of Salt Lake City, Utah, USA. Health & Place, 36, pp.8–17. Available at: http://www.sciencedirect.com/science/article/pii/S1353829215001148?via%3Dihub. Motrenko, A., Strijov, V. & Weber, G.-W., 2014. Sample size determination for logistic regression. Journal of Computational and Applied Mathematics, 255, pp.743–752. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0377042713003294. de Nazelle, A.J. et al., 2011. Improving health through policies that promote active travel: a review of evidence to support integrated health impact assessment., Nelson, A.C. et al., 2013. Bus Rapid Transit and Economic Development: Case Study of the Eugene-Springfield, Oregon BRT System. Journal of Public Transportation, 16(3), pp.41–57. Available at:    51 http://scholarcommons.usf.edu/jpt/vol16/iss3/3/. Neutens, T., 2015. Accessibility, equity and health care: Review and research directions for transport geographers. Journal of Transport Geography, 43, pp.14–27. Available at: http://dx.doi.org/10.1016/j.jtrangeo.2014.12.006. Panter, J. et al., 2016. Impact of New Transport Infrastructure on Walking, Cycling, and Physical Activity. American Journal of Preventive Medicine, 50(2), pp.e45--e53. Park, E.S. & Sener, I.N., 2017. Impact of light rail transit on traffic-related pollution and stroke mortality. International Journal of Public Health, 62(7), pp.721–728. Pollack, S., Bluestone, B. & Billingham, C., 2010. Maintaining Diversity In America’s Transit-Rich Neighborhoods: Tools for Equitable Neighborhood Change. Dukakis Center for Urban and Regional Policy, pp.1–68. Available at: http://www.northeastern.edu/dukakiscenter/wp-content/uploads/2011/12/TRN_Equity_final.pdf. Richmond, S. a et al., 2014. Exploring the impact of a dedicated streetcar right-of-way on pedestrian motor vehicle collisions: A quasi experimental design. Accident; analysis and prevention, 71C, pp.222–227. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24950129 [Accessed June 29, 2014]. Ridgeway, G. & MacDonald, J.M., 2017. Effect of Rail Transit on Crime: A Study of Los Angeles from 1988 to 2014. Journal of Quantitative Criminology, 33, pp.277–291. Available at: https://link.springer.com/article/10.1007/s10940-016-9296-7. Rodriguez, D.A., Vergel-Tovar, E. & Camargo, W.F., 2016. Land development impacts of BRT in a sample of stops in Quito and Bogota. Transport Policy, 51, pp.4–14. Available at: http://www.sciencedirect.com/science/article/pii/S0967070X15300615. Rotger, G.P. & Nielsen, T.S., 2015. Effects of job accessibility improved by public transport system: Natural experimental evidence from the copenhagen metro. European Journal of Transport and Infrastructure Research, 15(4), pp.419–441. Available at: http://pure.sfi.dk/ws/files/413417/2015_04_02.pdf. Saelens, B.E. et al., 2014. Relation Between Higher Physical Activity and Public Transit Use. American journal of public health, pp.1–6. Saxe, S., Cruickshank, H. & Miller, E., 2015. Greenhouse Gas Impact of Ridership on Sheppard Subway Line, Toronto, Canada. Transportation Research Record: Journal of the Transportation Research Board, 2502(2502), pp.62–70. Available at: http://trrjournalonline.trb.org/doi/10.3141/2502-08. Saxe, S., Miller, E. & Guthrie, P., 2017. The net greenhouse gas impact of the Sheppard Subway Line. Transportation Research Part D: Transport and Environment, 51, pp.261–275. Available at: http://www.sciencedirect.com/science/article/pii/S1361920916300621. Schweitzer, L. & Zhou, J., 2010. Neighborhood Air Quality, Respiratory Health, and Vulnerable Populations in Compact and Sprawled Regions. Journal of the American Planning Association, 76(3), pp.363–371. Shen, Q., 2013. Under What Conditions Can Urban Rail Transit Induce Higher Density? Evidence from Four Metropolitan Areas in the United States, 1990-2010. University of Michigan. Available at: http://search.proquest.com.ezproxy.cul.columbia.edu/docview/1497967985/abstract?accountid=10226. Spears, S., Boarnet, M.G. & Houston, D., 2017. Driving reduction after the introduction of light rail transit: Evidence from an experimental-control group evaluation of the Los Angeles Expo Line. Urban Studies, 54(12), pp.2780–2799. Stokenberga, A., 2014. Does Bus Rapid Transit Influence Urban Land Development and Property    52 Values: A Review of the Literature. Transport Reviews, 34(3), pp.276–296. Available at: http://www.tandfonline.com/doi/abs/10.1080/01441647.2014.902404. Tainio, M. et al., 2016. Can air pollution negate the health benefits of cycling and walking? Preventive Medicine, 87, pp.233–236. Utsunomiya, K., 2016. Social capital and local public transportation in Japan. Research in Transportation Economics, 59, pp.434–440. Werner, C.M. et al., 2016. Evaluating the attractiveness of a new light rail extension: Testing simple change and displacement change hypotheses. Transport Policy, 45, pp.15–23. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0967070X15300482. Xie, L., 2016. Automobile usage and urban rail transit expansion: evidence from a natural experiment in Beijing, China. Environment and Development Economics, 21(5), pp.557–580. Available at: https://www.scopus.com/inward/record.url?eid=2-s2.0-84958780088&partnerID=40&md5=85301a53e55e058f27b3f58694b950bc. Yáñez, M.F., Mansilla, P. & Ortúzar, J. de D., 2010. The Santiago Panel: measuring the effects of implementing Transantiago. Transportation, 37(1), pp.125–149. Yarbrough, T.S., 2014. Retail change and light rail: An exploration of business location changes accompanying commuter rail development in Denton County, Texas. Zhu, Y. & Diao, M., 2016. The impacts of urban mass rapid transit lines on the density and mobility of high-income households: A case study of Singapore. Transport Policy, 51, pp.70–80. Available at: http://www.sciencedirect.com/science/article/pii/S0967070X16301470. Zuk, M. et al., 2017. Gentrification, Displacement, and the Role of Public Investment. Journal of Planning Literature, p.88541221771643. Available at: http://journals.sagepub.com/doi/10.1177/0885412217716439.  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.52383.1-0368659/manifest

Comment

Related Items