Sie sind auf Seite 1von 14

TRIP-100091; No of Pages 14

Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

Contents lists available at ScienceDirect

Transportation Research Interdisciplinary Perspectives


journal homepage: https://www.journals.elsevier.com/transportation-research-
interdisciplinary-perspectives

A systematic overview of transportation equity in terms of accessibility,


traffic emissions, and safety outcomes: From conventional to
emerging technologies
Yujie Guo a, Zhiwei Chen a, Amy Stuart a,b, Xiaopeng Li a, Yu Zhang a,

a
Civil and Environmental Engineering, University of South Florida, United States of America
b
College of Public Health, University of South Florida, United States of America

A R T I C L E I N F O A B S T R A C T

Article history: Emerging transportation technologies (e.g., electric vehicles) and services (e.g., shared mobility) provide efficient, sus-
Received 15 September 2019 tainable and cost-effective alternatives to traditional travel modes. However, whether these innovative technologies
Received in revised form 25 December 2019 bring benefits to different population groups in an equal and reasonable manner is still an open question. The various
Accepted 20 January 2020
methods applied to evaluate the equity performance of these systems are also not clearly comparable. In this paper, we
Available online xxxx
comprehensively review methods from the existing literature for assessing the equity of a few important system out-
Keywords:
comes: accessibility, traffic emissions, and safety. We also identify the existing challenges of analyzing equity for
Equity assessment framework emerging transportation technologies. We unify the existing methodologies into a three-step framework that includes
Emerging transportation technologies population measurement, cost/benefit measurement and equity assessment, and we summarize the applicable mea-
Accessibility surements for each step, in detail. A handful of literature focusing on emerging transportation technologies, such as
Traffic emissions shared mobility and autonomous vehicles, were also identified and surveyed; the methodologies used were found to
Environment fit with the three-step framework. We summarize the major findings and discuss promising directions for developing
Safety more sophisticated equity assessment methodologies for emerging transportation technologies. Overall, based on a
comprehensive review, this paper contributes a framework for assessing the equity of transportation systems that in-
tegrates accessibility, traffic emissions, and safety outcomes. The summarized framework can be an overview resource
to assist researchers and transportation planners who require equity analysis methods. The research gaps identified
also provide directions for equity research on emerging transportation technologies.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
2. Selection of relevant literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
3. Framework for transportation equity analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
4. Population measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
4.1. Population measurement for horizontal equity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
4.2. Population measurement for vertical equity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
4.3. Spatial scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
5. Cost/benefit measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
5.1. Accessibility-related cost/benefit measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
5.2. Cost/benefit measures related to traffic emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
5.2.1. Measurement of traffic-related air pollution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
5.2.2. Measurement of exposure to traffic pollution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
5.3. Cost/benefit measures related to traffic safety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
6. Equity assessment approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
6.1. Mismatch analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
6.2. Statistical approaches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

⁎ Corresponding author.
E-mail address: yuzhang@usf.edu. (Y. Zhang).

http://dx.doi.org/10.1016/j.trip.2020.100091
2590-1982/© 2020 The Authors. Published by Elsevier Ltd. This is an open access
article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
Y. Guo et al. Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

6.3. Inequality indicator-based approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0


7. Equity results and policy implications for traditional transportation system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
7.1. Transportation equity issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
7.2. Policy implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
8. Applications of the framework to emerging transportation technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
8.1. Equity assessment framework for emerging transportation technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
8.1.1. Accessibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
8.1.2. Traffic emissions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
8.1.3. Safety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
8.2. Equity outcomes and policy implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
8.3. Research gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
9. Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

1. Introduction policies are vertically equitable if they favor disadvantaged groups (Di
Ciommo and Shiftan, 2017).
In recent years, various innovative transportation technologies Based on these definitions, extensive studies have been conducted to
(e.g., autonomous, electric, and connected vehicles) and services assess the distribution of benefits for various traditional transportation sys-
(e.g., bike-sharing, car-sharing) have emerged as alternatives to tradi- tems, ranging from unimodal systems (e.g., automobile, bus) to multimodal
tional travel modes and are becoming increasingly popular all over the systems (e.g., transit network consisting of buses, trains, trams) with a focus
world. These innovative services have been shown to bring substantial on accessibility, environmental impact, and safety. Despite these studies, to
economic, environmental, and health benefits to society, including im- date there is no established standard to assess the equity performance of tra-
proved transportation services (Li et al., 2016), reduced emissions of ditional transportation systems, including those involving emerging trans-
pollutants (Pal and Zhang, 2017; Yu and Stuart, 2017), and induced portation technologies and services. Further, a better understanding of
physical activities (Woodcock et al., 2014). For example, 1 car-sharing past attempts in traditional transportation systems can lay a solid founda-
vehicle removed 9–13 private vehicles in North America in 2008 tion for proposing an appropriate equity assessment methodology for
(Martin et al., 2010). Also, people receive benefits from bike sharing emerging transportation systems.
programs with increased physical activity, improved accessibility, and To address these gaps, this paper reviews the state-of-art in assessment
reduced greenhouse gas (GHG) emissions (Buck, 2013). However, of the equity of three important outcomes of traditional transportation sys-
these benefits may not be distributed among population groups in a tems and identifies the existing challenges in applying existing methods to
fair or reasonable manner, leading to what are known as equity issues emerging transportation technologies. A synthesis of the transportation re-
in urban transportation planning. search literature on the equity of accessibility, traffic emissions, and safety
Conceptually, ‘equity’ refers to the fairness of the distribution of impacts is presented. We also unify the existing methods into a framework for eq-
among populations (Litman, 2002) and is a multi-disciplinary term widely uity assessment consisting of three steps, namely population measurement,
used in various fields such as politics, economics, and environmental sci- cost/benefit measurement, and inequity measurement. This review is dif-
ence; thus, its specific definition varies among different contexts. We ferent from others in the following ways:
focus here specifically on a definition that concerns the distribution of out-
• First, it offers an integrated perspective that considers the multiple out-
comes, and not the broader concept also concerned with participation of
comes of accessibility, traffic emissions, and safety. Existing works have
marginalized groups in policy processes (Department of Trannsportation
usually studied equity performance for one type of outcome, but very
of United States, 1970; Federal Highway Administration of the United
few have offered an integrated view.
States [FHWA], 2012; FHWA and Federal Transit Administration [FTA],
• Second, it unifies existing equity assessment methods into a taxonomy
1999; FTA, 2012a; FTA, 2012b). ‘Equality’, a similar term that is also fre-
based on a three-step process. This framework summarizes the measure-
quently used in the literature, instead describes a homogeneous distribu-
ments that can be applied in each step of equity evaluation, allowing
tion of impacts among population groups (García-Valiñas et al., 2005).
transportation planning agencies to find the measurements they require
Inequality can be reduced by regulation and management, e.g., by requir-
more efficiently.
ing technologies that reduce air pollution where it is higher than in other
• Third, it discusses the adaption of available methodologies for assessment
regions. Inequity, however, is largely the subset of inequality that is deemed
of emerging transportation technologies and services and identifies existing
unfair or unjust by social norms (Macinko and Starfield, 2002). For exam-
challenges to such adaptation.
ple, health inequalities that stem from genetic differences are usually not
categorized as inequitable; while disparities in health resulting from expo- Note that this review paper does not cover all aspects of equity in trans-
sure to unhealthy levels of traffic emissions are considered inequitable portation, rather it reviews how previous transportation research has
(Levy et al., 2006). framed and measured the equity of transportation outcomes, including ac-
In transportation, equity issues first emerged with the Civil Rights Act of cessibility, traffic emissions, and safety. Furthermore, the review is not con-
1964, which requires federal agencies to distribute federal resources in the cerned with equity in the policy process or system financing, neither is it
fairest and least discriminatory manner (Welch and Mishra, 2013). Nowa- about philosophy or theory of equity.
days, equity is considered in most transportation planning documents The remainder of this paper is organized as follows: Section 2 presents
(Karner et al., 2016). Generally speaking, equity can be considered from the search terms and paper selection criteria used to identify and filter pa-
the horizontal and vertical perspectives (Bhuyan et al., 2019). Horizontal pers for this review. Section 3 synthesizes the approaches used for the anal-
equity requires each similar individual or group to have the same distribu- ysis of the equity performance of transportation into a general framework
tion of costs or benefits (Litman, 2002). A transportation program is hori- and procedure. Section 4 reviews population measurement approaches.
zontally equitable if related resources are distributed evenly among the Section 5 reviews the measurement of costs/benefits related to accessibil-
population (Bhuyan et al., 2019). Vertical equity, also known as ‘social eq- ity, traffic emissions, and safety. Section 6 reviews approaches for the mea-
uity’, aims to provide services to those who need them most. Transportation surement of equity of those outcomes. Section 7 summarizes the findings

2
Y. Guo et al. Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

and policy implications from the current literature related to traditional such as census areas, is quantified. If vertical equity is evaluated, the dis-
transportation systems. Section 8 synthesizes the current literature that tribution of the population by subgroups (e.g., based on socioeconomic
has measured equity for emerging transportation technologies, discusses status) is further quantified.
the application of an equity analysis framework to emerging transportation
Cost/benefit measurement – quantifies the costs or benefits (i.e., out-
systems, and identifies research gaps. Section 9 provides concluding
comes) of interest (e.g., accessibility) for each population group. For ex-
thoughts.
ample, a public transit accessibility index for each census track has been
2. Selection of relevant literature used to measure accessibility for the population of each corresponding
area.
To identify the relevant literature for this review, we used both Google Inequality measurement – compares the outcomes among the spatially
Scholar and the ScienceDirect database. The search fields for Google distributed population (horizontal equity) or among population sub-
Scholar were its default search engine, and for ScienceDirect we only groups (vertical equity). The equity measurement tools include mis-
used ‘Keywords’ search field. All searches looked for works that contained match analysis (e.g., using GIS mapping), calculation of inequality
both a search term related to equity and a search term related to the trans- metrics (e.g., Gini index), and regression modeling (e.g., multivariate
portation outcome, emerging technology, or service. For equity, we used regression).
the search terms ‘justice of’ or ‘equity of’. For accessibility, our search
terms were ‘transportation accessibility’ and ‘transportation connectivity’. This three-step framework synthesizes the core process traditionally
For environmental impacts, we used the terms ‘transportation emission used to assess the equity of transportation system outcomes. Results from
exposure’ and ‘traffic pollution’. For safety, we used ‘traffic accident’ and this framework can assist researchers or practitioners to better understand
‘transportation safety’. For emerging transportation concepts, the search the equity performance of transportation systems. Based on above defini-
terms were ‘shared mobility’, ‘bike sharing’, ‘car sharing’, ‘emerging trans- tions and discussions, the sections below provide detailed discussion of
portation’, ‘electric vehicle’, ‘autonomous vehicles’, and ‘ride hailing’. each step in the equity analysis framework.
Many of the papers identified through this search included brief reviews
of the existing literature, which enabled the identification of additional ar- 4. Population measurement
ticles. To select relevant papers, we carefully screened each paper found to
identify those that directly address the topics within the scope our review. Population measurement defines the subsets of the human population
The papers we selected include studies that focused on equity assessment or for which the cost/benefit measure will be determined and compared, basi-
on cost/benefit measurement. We also selected papers on the equity of pol- cally answering the question “equity for whom?” Outcomes can be mea-
icies and practices for both traditional and emerging transportation sys- sured for individuals and for population groups that are defined based on
tems. Overall, a total of 61 papers were selected for detailed review. either spatial location, if horizontal equity is to be evaluated, or based on
These include 41 papers that measured the equity performance of tradi- other characteristics (such as socioeconomic status, race/ethnicity, or
tional transportation systems, 13 papers that measured the equity perfor- mobility needs), if vertical equity is evaluated. This section will discuss dif-
mance of emerging transportation systems, and 7 papers that did not ferent approaches to population measurement for equity evaluation and
measure transportation equity but were nonetheless found to be relevant related issues.
to the scope here.
4.1. Population measurement for horizontal equity
3. Framework for transportation equity analysis
Horizontal equity evaluates the distribution of outcomes among all indi-
As noted previously, there is no existing standard approach for assessing viduals or among spatially distributed population groups. In transportation
the equity performance of transportation systems. After comprehensively studies, measuring outcomes at the individual level is usually impractical
reviewing the literature identified above, we found that the majority of due to the number of people, the complexity of human travel behavior in
existing methods for transportation equity analysis can be generalized a multimodal transportation system, and the lack of individual-level demo-
into a three-step framework, as shown in Fig. 1. graphic data due to privacy issues. Although equity analyses based on
individual-level outcome measures is still rare, the emergence of activity-
Population measurement – defines the research scope based on the pop-
based travel demand models is beginning to allow such studies (e.g.,
ulation groups or individuals for whom the outcomes of transportation Gurram et al., 2019). Nonetheless, equity has usually been evaluated by
systems are measured and compared; it also quantifies the distribution comparing outcomes among population groups defined based on the spatial
of the study domain population among those groups. If horizontal eq- location of residence (though some studies have considered the location of
uity is evaluated, the distribution of the total population to area units, work, school, and other activities). For these analyses, a defined spatial unit

Fig. 1. Three-step framework of equity assessment.

3
Y. Guo et al. Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

is used as the level of analysis for which the aggregate person count is deter- Smaller scales could potentially bring benefits to the study. Geronimus
mined. Aggregated population data for spatial unit areas, such as census and Bound (1998) argued that less bias was introduced for smaller geo-
tracts, census blocks, or Traffic Analysis Zones (TAZs), are typically ob- graphic units. Also, by comparing the census tract and census block scale,
tained from the Census Bureau. Details of different census area units such Soobader et al. (2001) found that smaller geographic units better repre-
as the area size and available population information can be found via sented individual socioeconomic status proxies. Conversely, Tian et al.
https://www.census.gov. The selection of the census area unit (the level (2013) argued that the census tract is the appropriate scale for evaluating
of analysis) will be discussed in Section 4.3. relationships between health disparities and socioeconomic status because
population characteristics and socioeconomic status within a census tract
4.2. Population measurement for vertical equity are relatively homogenous by design. At this stage, it remains unclear
which scale is the most appropriate and more research is needed to evaluate
Vertical equity compares distributions of outcomes between population the choice of scales and their impacts on transportation equity evaluation.
groups defined by demographic characteristics, such as socioeconomic sta-
tus. Due to the lack of individual-level data, data aggregated to the same 5. Cost/benefit measurement
types of spatial units of analysis described above for study of horizontal
equity are also used for vertical equity analysis. This is enabled by the avail- Cost/benefit measurement quantifies the outcomes of the transporta-
ability of detailed information on population characteristics for many of tion system for the population groups chosen above. In this paper, we sum-
these spatial units. Common population characteristics that are used to de- marize methods for the measurement of costs and benefits relevant to three
fine population subgroups in vertical equity studies include race, ethnicity, important outcomes in transportation equity literature: accessibility, traffic
income, and education level. emissions, and safety.
Race and ethnicity are similar classifications that group people by com-
mon ancestry and physical characteristics. Groups often considered include 5.1. Accessibility-related cost/benefit measures
black/African American, white, and Hispanic (Tian et al., 2013); more de-
tailed division includes American Indian or Alaskan native, Asian or Pacific Accessibility is a fundamental outcome for evaluating the equity perfor-
Islander, multiracial, etc. (Stuart and Zeager, 2011). For education level, mance of a transportation system. Accessibility, such as people's access to
adults with less than a high school education (Buzzelli and Jerrett, 2007) jobs, health facilities, etc., reflects the extent to which a transportation sys-
are the most commonly used group in equity analyses. More detailed anal- tem enables individuals to reach activities or destinations by means of indi-
ysis have included additional groups defined as adults who have completed vidual of combined transport modes (Welch and Mishra, 2013). The
high school, those with some college education, and college graduates following is a summary of common methods for measuring accessibility-
(Harper et al., 2015). Age-based groups that distinguish children and related cost/benefits that have been used in equity analysis.
older adult populations have been used in some studies (Gurram et al., The simplest method for assessing accessibility-related costs/benefits
2015), as children and the elderly are usually more vulnerable to environ- used in equity analysis was used by Currie (2010) to analyze the public
mental risks such as air pollution (Buka et al., 2006; Simoni et al., 2015). transportation supply in Melbourne. This approach measures the popula-
To quantify each of the above groups for equity analysis, the percentage tion of a zone's accessibility to transportation facilities (e.g., bus stops,
of each group residing within each spatial unit is usually used. Conversely, train stations, tram stops, etc.) as the ratio of the service area for a type of
due to the continuous nature of income data, it is usually measured based transportation facility to the area of that zone. It accounts for the spatial
on the median household income for each spatial area (Buzzelli and coverage of a transportation system with its service level in a simple and
Jerrett, 2007; Kravetz and Noland, 2012; Tian et al., 2013; Sider et al., intuitive manner. Thus, it is called a “coverage-based measure” here. Due
2015); however categorization of average household income into quintiles to its simplicity, this measure has been applied to investigate horizontal
has also been used (Morency et al., 2012). and vertical equity of public transport systems (Delbosc and Currie, 2011;
In addition to the above commonly-used population characteristics, Ricciardi et al., 2015). Yet, two significant drawbacks exist in this approach.
some researchers also consider groups that are explicitly based on disad- First, although service frequency has been incorporated, this measure
vantaged status, such as unemployment rate (Buzzelli and Jerrett, 2007; cannot capture many details in a transportation system such as vehicle ca-
Sider et al., 2015), mobility need and ability (Currie, 2010), deprivation pacity and running speed. Thus, it cannot accurately reflect the quality of
index (also termed social disadvantage indicator) (Havard et al., 2009; service of a transportation system. Second, it measures only the
Sider et al., 2015). The deprivation index is a measure of cumulative disad- population's accessibility to a transportation system in its own zone rather
vantage that integrates various socioeconomic factors such as average than describing the ability to reach activities or destinations within a larger
household income, percentage of car ownership, unemployment rate, and study area. Thus, it fails to reach the goal of accessibility assessment, i.e.,
ethnicity. Such a combination of socioeconomic factors is claimed to be determining to what extent a transportation system enables people to
more representative of a population's disadvantage (Sider et al., 2015). In reach activities or destinations.
additional to the use of aggregated population data, individual-level To address the first drawback, a couple of studies (Mishra et al., 2012
sociodemographic data from travel demand modeling is gaining more at- and Welch and Mishra, 2013) proposed a refined measure that captures
tention, but its use for equity analyses has been limited to a few studies more details on the operations of a transportation system. They considered
(Gurram et al., 2015; Gurram, 2017; Gurram et al., 2018; Chen et al., 2019). various factors that represent the service quality of a transit system, such as
frequency, speed, distance, capacity, and required transfers. They also in-
4.3. Spatial scale cluded consideration of the multiple bidirectional nature of transit lines
passing through a single station. A metric called “connecting power” was
For equity assessment, it is crucial to select a proper scale for the spatial used to describe the service quality in both inbound and outbound direc-
unit chosen. However, very few studies in the transportation equity field tions, where the connecting power of a station was defined as the sum of
have explored the suitability of scale. Different scales can lead to inconsis- the connecting power of the inbound and outbound connecting power of
tent results in equity evaluation, which has been referred to as the modifi- all lines passing through this station. Finally, the connecting power of a
able area unit problem (MAUP) (Dark and Bram, 2007). For example, zone was estimated as the average of the connecting power of all stations
Tian et al. (2013) used census tract data to analyze the near-road traffic in that zone. Note that this measure of accessibility still adopts the concept
population throughout the U.S., while Rowangould (2013) used census of coverage, so it falls into the category of coverage-based measure.
block data and found that many of his results did not agree with Tian. He Welch and Mishra's revised approach overcomes the quality-of-service
explained that the difference might stem from different scales used for drawback of coverage-based measures; however, it cannot address the sec-
data analysis. ond drawback. Besides, coverage-based measures are built on service radii

4
Y. Guo et al. Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

of transit stations, so they cannot be used for modes without stations, such sparse and generally inadequate for intra-urban equity analyses. Hence,
as emerging transportation technologies including free-floating bike- dedicated fixed-site or mobile monitoring campaigns are often used to
sharing, free-floating car-sharing, and ride-sourcing. In light of these issues, capture the variability of traffic-related pollutant concentrations at high
some scholars have proposed reachability-based measures that indicate the spatial resolution. As good indicators for traffic pollution, NO2/NOx is
ability of a zone's population to reach activities or destinations in all other often sampled to represent traffic pollution levels. Different samplers such
zones within the study area given a monetary and (or) time budget. The as the Ogawa passive samplers (Stuart and Zeager, 2011) or IVL passive
basic idea of a reachability-based approach is to count how many zones samplers (Habermann et al., 2015) can be used to measure NO2 concentra-
the population within a specific zone can reach with the given budget; tions. As monitoring data are rarely available for each specific unit of
the more zones one can reach, the larger the accessibility. Intuitively speak- analysis needed, regression models are often applied to this data (as
ing, the accessibility between two zones decreases as the travel cost in- discussed in Section 5.2.1.4) to estimate pollutant concentrations at
creases. Any functions that satisfy this property can be applied. Common unsampled locations.
examples in the literature are the cumulative accessibility function (Golub
and Martens, 2014; El-Geneidy et al., 2016) and a negative exponential 5.2.1.2. Measures of nearby traffic. Nearby traffic characteristics such as traf-
function (Guzman et al., 2017). With this, we can formulate the accessibil- fic density can be used as surrogates for traffic pollution when no traffic-
ity of a zone as the sum of its accessibility to any other zone. related air pollution data are available. Studies have shown that traffic
density can influence NO2 concentration levels near roadways (Rijnders
5.2. Cost/benefit measures related to traffic emissions et al., 2001; Stuart and Zeager, 2011). For the calculation of traffic density
within a spatial unit, refer to (Tian et al., 2013). Other indicators of traffic
Traffic is a major source of air pollution worldwide. Traffic emissions in- pollution include highest annual average daily traffic (AADT), distance to
clude many types of regulated air pollutants, such as carbon monoxide a high-volume roadway, and number of roads (Health Effects Institute.
(CO), oxides of nitrogen (NOx), benzene, and particulate matter (PM), in- Panel on the Health Effects of Traffic-Related Air, 2010).
cluding fine particles (PM2.5) and elemental carbon (EC). Exposure to
traffic-related air pollution has adverse effects on human health; it can ex- 5.2.1.3. Modeling of traffic emissions and dispersion. Modeling of traffic emis-
acerbate asthma, and may also contribute to the development of childhood sions and their dispersion to estimate ambient pollutant concentrations can
asthma, other respiratory symptoms, impaired lung function, and cardio- provide highly-resolved estimates of traffic-related air pollution within a
vascular mortality and morbidity (Health Effects Institute. Panel on the research area. Generally, the process can be divided into three steps, as
Health Effects of Traffic-Related Air, 2010). Additionally, according to shown in Fig. 2:
U.S. EPA (2019), transportation contributed the largest portion (29%) of The first step estimates characteristics of roadway travel conditions
greenhouse gas (GHG) emissions in the U.S in 2017. Because GHGs contrib- such as vehicle volume and travel speeds on a road segment within a time
ute to climate change, which can disproportionately affect disadvantaged interval. These features are input to a traffic emissions model. Two methods
people, these emissions are also a health equity concern. are commonly used to estimate these values: travel demand modeling
However, assessment of impacts of traffic emissions on equity is chal- (e.g., activity-based travel demand model) (Hatzopoulou and Miller,
lenging because ambient pollution is a complex mixture of many emitted 2010; Sider et al., 2015; Yu and Stuart, 2016) or direct calculation based
pollutants. Secondary formation of some pollutants, through chemical reac- on available data (e.g., AADT) (Yu and Stuart, 2016). The second step esti-
tions in the atmosphere, also makes attribution of ambient pollutant levels mates link-based or area-based emission inventories through complex emis-
to traffic sources difficult. Furthermore, impacts of GHGs on populations sion models such as the USEPA MOVES model. The majority of studies have
are not related to direct exposure to emissions or resultant pollutant con- used a single pollutant such as NO2 as an indicator of traffic pollution when
centrations. Due to these difficulties, surrogates of traffic emissions, such producing emission inventories. In addition to the rate and location of emis-
as concentrations of specific pollutants that have large vehicular-source sions provided by inventory data, other factors such as winds and land use
contributions (e.g., CO, NOx, EC, PM2.5) and measures of traffic itself can impact the fate and transport of traffic emissions. Dispersion modeling
(such as traffic density), are commonly used in equity assessments of integrates these factors to produce a high-resolution map of pollutant con-
exposure-related equity impacts. Although no surrogate has been found to centrations throughout a study domain. CALPUFF, AERMOD and RLINE
be ideal (Health Effects Institute. Panel on the Health Effects of Traffic- are popular dispersion models used in the transportation context. Other
Related Air, 2010), nitrogen dioxide (NO2, a component of NOx) is often simplified proximity-based modeling methods, such as the use of rounded
used because its level has been found to be influenced by traffic counts buffer zones around road links to approximate the effects of dispersion
(Rijnders et al., 2001; Stuart and Zeager, 2011) and it also has greater spa- (Sider et al., 2015), have also been applied.
tial heterogeneity than some other traffic-related air pollutants (Jerrett
et al., 2005). 5.2.1.4. Land use regression (LUR) modeling. LUR is the newest addition to
Although direct measurement is possible, exposure to traffic emissions pollution modeling and has become very popular in the last 10 years. It is
in the transportation equity literature is typically estimated using two used to estimate the small-scale spatial variations in pollutant concentra-
steps. The first step estimates the distribution of traffic-related air pollution tions for intra-urban analyses. LUR models have the basic form:
(or surrogates) in the study area, and the second step estimates exposures
by matching pollution levels with humans, based on their location in Pollutant Concentration ¼ ðβ0 þ β1 X 1 þ β2 X 2 ∙∙∙βx X x Þ þ ɛ ð1Þ
space and/or time.
LUR models estimate the above βi coefficients by regressing measured
5.2.1. Measurement of traffic-related air pollution ambient pollution concentration levels against independent variables (Xi)
Generally, there are four methods for estimating levels of traffic-related such as land-use, nearby traffic counts, elevation, etc. Once the variable co-
air pollution that are regularly used in the transportation field. This section efficients are determined, the model can be applied to estimate pollution
will provide a brief overview of each method. levels for unsampled locations. The process of modeling has four steps: 1.

5.2.1.1. Monitoring of ambient pollutant concentrations. Monitoring is the


most straightforward way to measure ambient pollution. It uses samplers
or monitoring instruments to collect pollution samples at fixed sites or on
mobile vehicles. Concentrations can be determined through in-situ or labo-
ratory analysis. A few traffic-related pollutants (CO, NO2, PM2.5) are rou-
tinely monitored throughout the U.S., but monitoring networks are very Fig. 2. Traffic emissions and dispersion modeling process.

5
Y. Guo et al. Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

pollutant concentration sampling, 2. independent variable selection and 5.2.2. Measurement of exposure to traffic pollution
quantification, 3. regression modeling to determine the coefficients, and A person's total daily exposure to pollutants depends on where they
4. model validation and evaluation. travel, time spent in vehicles, being indoors and outdoors, ventilation of
Pollutant concentration sampling was described at Section 5.2.1.1. The buildings, etc. (Rowangould, 2013). Due to the natural complexity of travel
concentration of a single pollutant (e.g., NO2) is often used as the indicator behavior, individual or population exposures are usually simplified and es-
for traffic pollution and as the dependent variable for the regression model. timated by two approaches: fixed-location approach and activity-based
The independent variables can be unique for different cities and usually are approach.
characterized for buffers areas surrounding each concentration measure- A fixed-location approach estimates individual or population exposure
ment site location. In Ryan and LeMasters (2007), independent variables based on residential address, school location, etc., by estimating the con-
are grouped into four classes: road type, traffic count, elevation, and land centration at that location. Population data are commonly only available
cover. To assist the selection of independent variables, statistical ap- aggregated to a spatial area, such as residential demographic data for the
proaches such as AIC/BIC (Dirgawati et al., 2015) and stepwise linear re- census block level. Hence exposure is often determined at an aggregate spa-
gression (Wang et al., 2013) can be adopted. The Variance Inflation tial level. For population subgroups, group-average exposure can be esti-
Factor (VIF) has also been applied to ensure the final model is not affected mated by summing over all spatial areas the products of the subgroup
by multicollinearity (Dirgawati et al., 2015). After building the multivari- population and the pollution level specific to that spatial area, and dividing
able regression model, the performance can be evaluated by cross valida- the sum by the total subgroup population of the study area. If concentration
tion (Wang et al., 2013; Dirgawati et al., 2015) and root mean square data are time varying (e.g., hourly), then temporal statistics of exposure,
error (RMSE) (Brauer et al., 2003). The final model can be used to predict such as maximum hourly exposure or daily average exposure can also be es-
pollutant concentrations for unsampled locations with available predictor timated (Yu and Stuart, 2013). The disadvantage of using a fixed-location
data for those locations. A few studies (Eeftens et al., 2011; Cesaroni approach is that it does not capture population activity patterns.
et al., 2012; Wang et al., 2013) have explored the prediction power of The activity-based approach estimates exposures based on tracking in-
LUR modeling and found that it performs very well for long-term forecast- dividuals throughout their daily activities. Individual exposure can be esti-
ing, whereas hindcasting has lower accuracy. mated by summing the product of pollutant concentration and time spent at
each activity location. The activity locations of individuals can be obtained
5.2.1.5. Comparison of methods for measuring traffic-related air pollution. from the National Household Travel Survey (NHTS) (Gurram et al., 2015)
Overall, each method has its own advantages and disadvantages. When to or activity-based travel demand models (Hatzopoulou and Miller, 2010;
apply each method really depends on the scenario of study. Gurram, 2017; Gurram et al., 2018). Different from Gurram et al. (2015,
2019), Hatzopoulou and Miller (2010) did not consider exposures during
• Pollution monitoring: Direct measurement provides the pollution concen- travel; they argued that in-vehicle exposures are significantly different
tration level at a site. However, if the study area is very large (such as a from ambient air concentrations.
state or county), it is expensive and impractical to distribute enough sam- Comparing individual exposure results from an activity-based analysis
plers to capture the spatial variation of pollution concentrations. How- with those from a fixed-location analysis based on residential location,
ever, monitoring can be combined with other methods such as LUR to Gurram et al. (2015) suggested that the residence-based exposure approach
predict concentration level at unsampled locations. may misclassify the exposures of socially-disadvantaged population groups
• Measures of nearby traffic: This is the simplest approach to estimating less than those of others.
traffic pollution and relies on readily available data. However, the link be-
tween traffic density levels and pollutant concentrations is not fully un-
derstood (Rowangould, 2013), and measures of traffic do not capture 5.3. Cost/benefit measures related to traffic safety
the effects of meteorology, such as winds, temperature, etc. Furthermore,
using AADT to estimate traffic density can be biased because AADT data Studies of traffic safety have mainly focused on injuries or death caused
do not typically include traffic counts on small local roads and so cannot by motor vehicle accidents. Factors that can influence traffic injuries/casual-
represent all traffic counts within a measured area. ties have been studied extensively, including population socio-demographics
• Traffic emissions and dispersion modeling: Although modeling provides (e.g., race, income), physical development (e.g., employment density), and
estimates of traffic-related air pollution at user-defined resolution that is environmental factors (e.g., road density, traffic density) in an area
based on the established knowledge of air pollution physics and chemis- (Kravetz and Noland, 2012). The cost/benefit measures typically used to
try, its application requires advanced familiarity with those topics and re- quantify traffic safety for equity studies include the crash rate or counts of in-
lated mathematics, substantial amounts of specialized data, and juries and fatalities. For example, the crash rate can be quantified as the num-
significant computational resources. These characteristics have restricted ber of people from a specified population group injured in a census area
widespread use in the multiple fields that study impacts of traffic emis- divided by total population of that group in that area. Studies of equity be-
sions. tween races commonly adopt this measure. Examples of measures used for
• LUR: LUR allows high-resolution estimation of pollution using available injuries and fatalities have included simply the number of pedestrian injuries
data and statistical methods that are familiar to researchers in many in a census area or the number of crashes per 100,000 people. Unlike cost/
fields. However, most LUR studies to assess equity have used short-term benefit measurement approaches for accessibility and traffic pollution,
monitoring campaign data that ignore temporal variations in pollution which may involve formula calculations and modeling, the cost/benefit mea-
(Habermann et al., 2015). Also, there is an issue of the availability of pre- sures of traffic safety are mainly directly obtained from data sources that
dictor variable data. For example, the household density and industrial focus on pedestrian, cyclist, and driver injuries and casualties. Kravetz and
area data may not be available for all years of modeling interest or for Noland (2012) and Noland et al. (2013) extracted crash data from the
the same year that the pollution concentration was measured; this Plan4Safety database, which has comprehensive records of crashes occurring
might also influence the quality of the LUR model (Dirgawati et al., in New Jersey, including date of time, place, severity, vehicle actions and di-
2015). rections, environmental and surface conditions, occupants and pedestrians
involved, driver characteristics, etc. The geocoded crash data were aggre-
de Hoogh et al. (2014) compared the use of LUR and dispersion gated to the census block-group level for equity analysis. Similarly,
modeling of annual-average concentrations of a few traffic-related Steinbach et al. (2010) and Steinbach et al. (2016) obtained STATS19 child
pollutants against measured pollution data for 13 study areas. Results sug- injury data from the London Road Safety Unit that includes all reported cau-
gest that both methods are useful for studying small-scale variations of salities and traffic collisions in London. The crash data location was assigned
traffic pollution. to a lower super output area (LSOA) for equity analysis purposes.

6
Y. Guo et al. Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

However, frequently-used administrative data has the limitation of groups is not large, a map is not needed; the statistical metrics can be sum-
under-reporting, especially related to slight injuries. Police may focus on ac- marized in tables and compared directly (Stuart and Zeager, 2011; El-
cidents that legally must be reported but may ignore accidents not involv- Geneidy et al., 2016; Boarnet et al., 2017).
ing motor vehicles. For example, STATS19 data cover accidents on public Although the mapping approach can offer intuitive information, it can
highways but exclude single bicycle incidents and pedestrian falls be cumbersome to manually identify gaps in equity from the maps for all
(Aldred, 2018). As summarized by Ahmed et al. (2017), the percentage zones in a large metropolitan area. Thus, simplified approaches have been
error reporting (under-reporting) for developed countries is low for fatal in- proposed. Currie (2010) plotted the supply versus demand for each zone
juries, but it can be more than 40% for slight injury accidents. There are within the study area on a two-dimensional Cartesian coordinate graph.
many methods to rectify the data, the most popular of which is the In this way, zones with different degrees of inequality (between supply
capture-recapture method. Other methods including comparison with and demand) clustered in different regions in the graph and, thus, were
health sector data and probabilistic linkage. But it is not the focus of this very efficiently identified. For instance, zones that fell into the southeast
paper to discuss data enhancement. Kraemer and Benton (2015) adopted corner in the coordinate system had the highest demand but the lowest
the capture-recapture method to assemble the data, which combines data transit supply, which reflects the highest degree of inequality. Another
from two independent registries of fatal crashes of wheelchair users. One method is to combine the population and cost/benefit measures to obtain
registry used the National Highway Traffic Safety Administration a new measure (if possible) and then plot the combined measure. For in-
(NHTSA) Fatality Analysis Reporting System (FARS); preliminary assess- stance, to analyze transit provision with respect to social needs in Mel-
ment, however, indicates incompleteness of data. A second registry was bourne, Currie (2010) defined the need-gap of a zone as the difference
constructed by searching keywords in a LexisNexis U.S. newspaper data- between its transport provision and need index. This measure links trans-
base. The unmatched data from the news were sought in the FARS data- port provision and the population's heterogeneity in terms of their travel
base; almost all unmatched cases were identified in FARS, as those data needs and offers an opportunity to analyze the vertical equity with respect
failed to be coded as wheelchair users. Besides administrative safety data, to the mobility need and ability. A similar method was applied to investi-
Morency et al. (2012) used ambulance service data to collect information gate the disparity between transport provision and social exclusion in
on road traffic injuries, although ambulance service data naturally neglect Cali, Colombia, by taking transport social needs into account (Delmelle
more light injuries. Aldred (2018) extracted data from the National Travel and Casas, 2012). Later, Ricciardi et al. (2015) replaced need index values
Survey (NTS) in the UK, in which respondents were asked to recall if they with potential demands generated from the Distribution Fit Tool in Matlab
were involved in road accidents over the previous three years; details studio, based on which the equity distributions of three separate social dis-
such as injury, treatment, police involvement, and mode used were re- advantaged groups (i.e., older adults, low-income, and households without
corded. Although the data do not include geospatial information on acci- a car) were investigated.
dents, the detailed personal information (e.g., income, gender, age) and Note that although mismatch analysis is quite simple and intuitive, it
accident data still provide insights regarding equity issues. does not offer quantitative information on equity performance. Therefore,
Overall, while using transportation safety data, it is important to fully some quantitative analysis approaches based on statistical and econometric
understand the dataset to avoid bias caused by missing data. approaches or inequality index formulations have been proposed.

6. Equity assessment approaches 6.2. Statistical approaches

Various approaches for assessing equity in transportation system perfor- To quantify inequality for the purposes of assessment of both horizontal
mance have been proposed in the past few decades, which are broadly di- and vertical inequity, statistical and econometric measures are commonly
vided into three categories in this paper. The following subsections applied. For horizontal inequity, univariate measures of the dispersion of
discuss each equity assessment method. the distribution of a cost/benefit measure among geographic units (such
as census block groups) can be used. These include the variance or standard
6.1. Mismatch analysis deviation of the distribution, the range or inter-decile range, or the ratio of
a high to low value, such as the ratio of the 95th to 5th%ile value in the dis-
Mismatch analysis of simple descriptive statistics of the cost/benefit tribution. For example, Yu and Stuart (2013) evaluated the spatial distribu-
measures is a typical approach to study the equity performance of a trans- tions of NOx exposure in Hillsborough County Florida by the standard
portation system; its history dates back to the very earliest study measuring deviation and range. Delmelle and Casas (2012) applied the standard devi-
the performance of public transport in meeting the transport needs for dif- ation to quantify the inequality of public transit need/provision gap among
ferent demographic groups (Currie, 2004). Basically, this method presents spatial districts.
the distributions of the cost/benefit measures as maps or tables and then Methods such as analysis of variance (ANOVA), correlation analysis,
manually compares the distributions. With these maps or tables, an intui- and regression modeling are also widely used to assess vertical transporta-
tive understanding of the equity performance for each zone or population tion equity. Stuart and Zeager (2011) used bivariate linear correlations to
group can be obtained. investigate relationships between NO2 exposure and race-based percentage
A very simple approach following this idea is to map the statistics of enrollments of elementary school students. Sider et al. (2015) used one-way
population measures and of the cost/benefit measures in two different ANOVA to determine whether levels of exposure to traffic emissions were
maps (usually using GIS), where each spatial zone has one color on a significantly different between groups with different social disadvantage
scale from the lowest to the highest quintile of the distributions (Kaplan index values.
et al., 2014). The statistical metrics are usually the average values (of the One-way ANOVA and bivariate correlation analysis are straightforward
measures) for a zone but, in some situations, median, maximum, minimum, methods for exploring the relationship between a cost/benefit measure and
and standard deviation have also been used (El-Geneidy et al., 2016). This one socio-demographic factor (e.g., race). A more comprehensive approach
simple mapping approach can present much macroscopic information in a is to consider more factors that may influence the cost/benefit measure,
very intuitive and compact way such that it has been used extensively in through regression modeling with multiple explanatory variables. For ex-
assessing both horizontal and vertical equity. If one wants to assess horizon- ample, additional factors such as household income and education level
tal equity, the cost/benefit measures in each zone within the studied area have been found to contribute to overall inequality. In such models, the
need to be plotted (Golub and Martens, 2014). To assess vertical equity, sign of the coefficient on an explanatory variable provides the direction of
both the population characteristics and the cost/benefit measures can be the relationship between that factor and the cost/benefit measure, while
plotted in two maps and then the distributions of these two maps can be the magnitude of the coefficient is often used as a measure of the strength
compared (Kaplan et al., 2014). Note that when the number of zones or of the relationship (or level of inequality).

7
Y. Guo et al. Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

To study the equity of exposures to traffic emissions, logistic regression weigh costs/benefits for those at the low end of the cost/benefit distribu-
or spatial autoregressive (SAR) modeling have been applied. Buzzelli and tion. This parameter ranges from 0 to infinity, with 0 indicating no societal
Jerrett (2007) simplified exposure to a binary categorical variable, concern with inequality, and a higher value indicating more concern for the
e.g., low and high exposure, and used logistic regression to capture the fac- low cost/benefit (e.g., income) group. After comparing a variety of inequity
tors (number of dwelling, low income, low education, medium income, indicators for health benefits analysis, the Atkinson index was suggested by
race, etc.) that explain high and low exposures in Toronto. Buzzelli and Levy et al. (2006) as the best indicator for inequality evaluation in health
Jerrett (2007) and Havard et al. (2009) both found spatial correlation in benefit analysis. Levy et al. (2009) applied the Atkinson Index to quantify
the distributions of NO2 in their study area and applied SAR models to ac- the inequality benefits of different mobile source control strategies in
count for this factor. For traffic safety, negative binomial regression Boston.
(Kravetz and Noland, 2012; Harper et al., 2015) and Poisson regression In addition to borrowing concepts of inequality indicators from eco-
(Morency et al., 2012; Zhang and Lin, 2013) are commonly used, as traffic nomics, environmental and transportation researchers have also developed
accidents are infrequent discrete events whose frequency can typically de- inequality indicators that intuitively quantify disparities between popula-
scribed by using a Poisson or negative binomial distribution. For example, tions groups. Harner et al. (2002) developed several indicators, including
Morency et al. (2012) used a multilevel Poisson regression model to exam- the Comparative Environmental Risk Index (CERI), which quantifies
ine the disparity between wealthy and poor urban areas in traffic injuries at whether socio-demographically disadvantaged groups are more likely to
intersections, as well as the influence of roadway environmental factors be exposed to environmental hazards than the rest of population. For exam-
(such as population density and traffic estimates) on that relationship. ple, to compare pollution exposure between white and non-white popula-
Statistical and econometric approaches are particularly appropriate for tions, they used the following formula:
analyzing vertical equity because of their capacity for describing the rela-  
tionship between transportation costs/benefits and sociodemographic attri- at risk nonwhites
butes. Insights into horizontal equity can also be obtained using simple total MSA nonwhites
CERI nw ¼   ; ð2Þ
statistics of dispersion in the univariate distribution of the cost/benefit at risk whites
measures across geographic units of analysis. total MSA whites

6.3. Inequality indicator-based approaches where MSA is the Metropolitan Statistical Area. If CERInw is greater than 1,
non-whites are disproportionately exposed to environmental hazards, and
Although statistical approaches offer profound information on the rela- vice versa. The CERI index can also be used for other subgroups, such as
tionship between different sociodemographic attributes and transportation poor people. The CERI is equivalent to the concept of relative risk in epide-
costs/benefits, they do not provide an overall level of inequality that can be miology; it quantifies the relative risk of exposure for a subgroup. Similarly,
compared between cases. A solution to this drawback is the application of Stuart et al. (2009) developed a subgroup index of inequality that quantifies
inequality indices, which have been extensively used in economics, social the degree to which a specific subgroup is disproportionately exposed to
science, and public health. These offer a quantitative measure of the degree both costs (e.g., pollutant emission sources) and benefits (e.g., regulatory
of inequality in any variable, such as income, distributed over a population. monitoring) of an environmental management system. This subgroup in-
Popular inequality indicators include the Gini index, Atkinson index, Theil's equality index is formulated as:
entropy index, etc.  
In transportation, Gini index is the most extensively used indicator. The Zi
F i ¼ log ; ð3Þ
Gini index traditionally has been used to evaluate the distribution of wealth Ti
or income among a population. Delbosc and Currie (2011) first applied the
Gini index along with the Lorenz curve to analyze equity performance in where i indicates the subgroup (such as African Americans), Z is the fraction
transit supply in Melbourne. The Gini index ranges from 0 to 1, with 0 indi- that subgroup comprises of the population exposed to a particular hazard
cating perfect equality and 1 perfect inequality. Due to the computational (or benefit) level, and T is the overall fraction that subgroup comprises of
tractability and intuition of the Gini index, many studies have followed the larger study area population (such as a county). This quantitative
the pioneering work of Currie and used the Gini index as an overall index index is easy to interpret. Positive F indicates that members of the subgroup
of the equity performance in transportation systems (Welch and Mishra, (on average) are disproportionately exposed to the hazard (or benefit) at
2013; Kaplan et al., 2014; Guzman et al., 2017). The Gini index can be eas- the level considered; negative F means they are disproportionately unex-
ily adapted to various contexts and used for evaluating horizontal equity for posed at this level. Yu and Stuart (2013, 2016) and Gurram et al. (2015,
both the entire population and a targeted population group. 2019) have used this subgroup inequality index to study inequity in expo-
Although the Gini index has many good properties, it also has limita- sures to traffic-related air pollution in Tampa.
tions for equity analyses. For example, it is not decomposable into additive
attributions from subgroups or factors (Levy et al., 2006; World Bank, 7. Equity results and policy implications for traditional transportation
2005); this limitation can be detrimental to vertical equity analysis. Also, system
the implicit weighting of the Gini index's underlying social welfare func-
tion, and the value judgments implied for the cost/benefit distribution, As previously mentioned, equity can be evaluated from horizontal and
are somewhat arbitrary (Weymark, 1981). This can be limiting for policy vertical perspectives. Horizontal equity considers whether similar individ-
implications. For example, if a city wants to invest money to improve transit uals or groups are treated with same benefits/costs. Vertical equity con-
accessibility and decrease inequity, they may want to apply an equity in- siders whether different sociodemographic or need subgroups receive the
dictor that values improved accessibility in areas with the least prior transit same benefits/costs. As they evaluate the equity from different perspec-
development more highly than areas that are already very connected. To tives, practical implications are also different. If analysis of horizontal
address these limitations, other indices have been developed, such as equity shows significant inequality, policies could be targeted at areas
Atkinson index. with low accessibility or high traffic emissions, etc. If vertical equity indi-
The Atkinson index was first developed to quantify income inequality cates significant inequality among population subgroups, then actions
(Atkinson, 1970) and has been applied extensively in the environmental could be taken to enhance the benefits for disadvantaged populations.
justice contexts. Like the Gini index, the Atkinson index ranges between 0 However, horizontal and vertical equity are not necessarily correlated.
and 1, with 0 indicating complete equality and 1 indicating complete in- Boyce et al. (2016) measured horizontal and vertical equity of environmen-
equality. The Atkinson index is subgroup decomposable and incorporates tal pollution for all 50 U.S. states and did not identify a correlation between
an inequality aversion parameter that can be adjusted to more heavily horizontal equity and vertical equity. Similarly, disadvantaged population

8
Y. Guo et al. Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

groups identified from vertical equity analyses may not match with the dis- areas. For example, transportation planners can subsidize on-demand ser-
advantaged areas identified from horizontal equity analyses. For practical vices for low-income residents to improve their accessibility from home
applications, both horizontal and vertical equity should be measured to un- to bus/transit stations (Boarnet et al., 2017). While planning transit stations
derstand the overall equity performance of a system. While implementing a for deprived communities, planners should also ensure high connectivity of
project, equity is usually one of the goals in the project framework, and pol- stations rather than simply their accessibility (Welch and Mishra, 2013). In
icy makers need to make trade-offs between the two equity measures based terms of transit fares, transit planners can assess the impact of fare struc-
on the other project objectives, goals, and values. This section discusses the tures on accessibility for different areas in the city, based on their level of
equity issues and policy implications related with accessibility, traffic emis- socio-deprivation (El-Geneidy et al., 2016). This is important because it
sions, and safety. has been shown that people who have access to high quality public transit
are more likely to get employed and earn a higher average salary (Yi, 2006).
7.1. Transportation equity issues For inequity of exposure to traffic emissions, a significant contributor is
residential segregation, i.e., disadvantaged populations often live near
Title VI of Civil Right Act of 1964 mandates non-discrimination in pro- roadway sources (Yu and Stuart, 2016). Thus, it is suggested to reduce mo-
grams that receive federal funds, including transportation programs. None- bile emissions from areas where disadvantaged populations live and spend
theless, the performance of transportation systems remains inequitable for time (Yu and Stuart, 2013). Suggestions include promoting public transit
different socio-demographic groups. Litman and Brenman (2012) have crit- (Tian et al., 2013) and active transportation, setting low-emission zones
icized most transportation policies and practices as favoring automobiles in highly polluted areas, and setting tighter urban planning standards for es-
over other transportation modes, despite economically and socially disad- tablishing sensitive facilities near roadways (Carrier et al., 2017). For exam-
vantaged populations tending to heavily rely on non-motorized transport ple, California Senate Bill 352 prohibits siting a school within 500 ft. of a
modes. Also, there are disparities in public investment among different busy traffic corridor or freeway.
socio-demographic populations (Kravetz and Noland, 2012). Traffic safety results from the interaction between built environment
Several studies have found that disadvantaged groups enjoy the highest design and human behaviors (Ewing and Dumbaugh, 2009). In disadvan-
accessibility to public transportation in many cities, such as Inner Mel- taged communities, road infrastructure and facilities are often inferior to
bourne (Delbosc and Currie, 2011), Perth (Ricciardi et al., 2015), the San those in rich communities. However, people who live in poor areas are
Diego Metropolitan Area (Boarnet et al., 2017), the San Francisco Bay more likely to use active transportation (e.g., walking, biking) due to lack
Area (Golub and Martens, 2014) and Bogota (Guzman et al., 2017). How- of private vehicles (Kravetz and Noland, 2012). Thus, disadvantaged popu-
ever, because automobiles can bring better accessibility to activities and lations have more exposure to unsafe traffic, which results in more acci-
destinations over a larger area (Golub and Martens, 2014) and because dents. To improve safety conditions, policy makers should ensure that
high-income groups have higher car accessibility (Guzman et al., 2017), sidewalks and bike lanes in disadvantaged areas are good quality, well-
the accessibility of low-income and minority groups is lower than other maintained and well-connected, providing a safer environmental for active
groups when both public transport and automobiles are taken into account transportation.
(Kawabata, 2009; Grengs, 2010; Golub and Martens, 2014; El-Geneidy Besides the above discussions, the current scope of transportation eq-
et al., 2016). uity considerations also has some limitations. First, current transportation
In terms of exposures to direct traffic pollution, populations that are equity evaluations only consider a subset of all social equity issues
white and have high income have been found to be less exposed (Stuart (Litman and Brenman, 2012). For example, the ‘intangible’ transportation
et al., 2009; Tian et al., 2013; Yu and Stuart, 2013; Gurram et al., 2015). outcomes, such as walkability or livability, are overlooked. While ‘tangible’
Disadvantaged communities in Montreal were also found to be exposed to outcomes such as reduced traffic congestion, improved transit coverage, are
higher levels of pollution, but contribute very little emissions (Sider et al., easier to measure and present to the public; thus they have become the
2015). Although socially-advantaged groups have generally been shown focus of social equity considerations and are prioritized over the ‘intangible’
to enjoy greater health benefits of transportation, locality and other factors outcomes (Handy, 2008). For effective decision making, it is essential to in-
can cause variations in this overall trend. For example, Yu and Stuart tegrate all factors into a transportation equity analysis framework. Second,
(2016) evaluated exposures to multiple traffic-related pollutants in consideration of equity issues is challenged by the need to balance the di-
Tampa and found that white and high-income people were disproportion- verse goals of a project. Transportation planners need to carefully consider
ately exposed to high levels of some pollutants (acetaldehyde and formalde- the trade-offs between improving equity and the value of other social ben-
hyde), but low levels of other pollutants (NOx, benzene, and butadiene). efits provided by the transportation system, making the system more sus-
Buzzelli and Jerrett (2007) also found that areas of high income and dwell- tainable. Martens et al. (2012) suggest that planners should limit the gap
ing values in central Toronto were susceptible to exposure. between highest and lowest accessibility, meanwhile attempting to maxi-
For traffic safety, populations that have high income (Kravetz and mize the average benefits for the public.
Noland, 2012; Morency et al., 2012), high education (Harper et al.,
2015), and high white population density (Zhang and Lin, 2013; Yu,
2014) have been found to have fewer vehicle accident injuries and deaths. 8. Applications of the framework to emerging transportation
Furthermore, Harper et al. (2015) found that inequity in vehicle accident technologies
deaths for socially-disadvantaged populations worsened over time in their
study of trends from 1995 and 2010. In this paper, we focus on the emerging transportation technologies or
services of shared mobility (bike sharing and car sharing), ride hailing,
7.2. Policy implications and electric and autonomous vehicles. There is substantial excitement
that these new transportation modes could potentially reduce the equity
From the research outcomes, it has been shown that inequity remains a gap within the current transportation network. For example, bike sharing
prevalent issue. Minority or socially disadvantaged populations generally can be used as a feeder mode to public transit for people who do not live
suffer more from negative transportation impacts. An easy way for trans- in an accessible area. Furthermore, Bösch et al. (2018) have suggested
portation planners to tackle equity issues is to focus on areas with a high that if car sharing and carpooling are equipped with autonomous vehicle
share of people who belong to low-income, low-education, and minority technologies and are electric vehicles in the future, the cost of usage
groups (Gaffron, 2012; Tian et al., 2013). Many studies also propose could be reduced to less than $0.2 per mile. With such low costs and zero
other approaches to improve the equity of the system, as discussed below. direct emissions, they have the potential to significantly improve the mobil-
To improve equity of accessibility, it is essential for policy makers to ity and air quality of disadvantaged areas. However, given the gains that are
provide accessible, efficient and affordable public transit for disadvantaged possible, we must ensure the benefits are distributed in a fair manner.

9
Y. Guo et al. Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

Current studies on the equity performance of emerging transportation (i.e., bike/car surplus or shortage due to imbalanced travel demand),
technologies are still in their infancy; only limited research has been con- which have required various rebalancing strategies to be applied in prac-
ducted and is mainly focused on shared mobility. However, we found that tice. Thus, more sophisticated accessibility measurement approaches that
these equity assessment methods still follow the three-step framework capture the unique operational characteristics of station-based shared-
discussed above. In this section, we summarize the equity measurement mobility systems are needed. An attempt to address this gap was found in
methodologies that have been applied to emerging transportation technol- Mooney et al. (2019), where spatial equity in access to dockless bike-
ogies, along with the expected equity outcomes and policy implications. sharing was investigated. They defined the demand as the mean of inverse
of idle time in a neighborhood per day while the supply was defined as the
8.1. Equity assessment framework for emerging transportation technologies average number of bikes rebalanced in a neighborhood per day. Then they
plotted the supply versus demand for each neighborhood in a two-
The section summarizes equity assessment frameworks that have been dimensional Cartesian coordinate system, which is essentially the same as
applied for emerging transportation systems in terms of accessibility, traffic to the approach used by Currie (2010) for transit systems.
emissions and safety. For ride-hailing services, user profiles have also been analyzed to reveal
equity impacts (Smith, 2016), with the underlying assumption that only
8.1.1. Accessibility registered users have accessibility. Some researchers used ‘waiting time’
After surveying studies on equity evaluation of emerging transportation as a proxy for accessibility. Longer waiting time indicates the supply of ve-
technologies, we find that the applied methods also follow the three-step hicles is lower than demand and thus indicates lower accessibility. To mea-
framework. Table 1 presents the measurements used for each step in the sure the waiting time, Hughes and MacKenzie (2016) and Stark and
three-step framework from these studies. As shown in Table 1, the ap- Diakopoulos (2016) obtained ‘waiting time’ through the Uber API. Brown
proaches for measurement of the population and of inequity have all been (2018) conducted an audit study (i.e., they employed passengers to take
covered in the discussion of traditional transportation systems, while not Lyft and record the waiting times) to obtain this information. This approach
all accessibility measurement approaches are the same. aimed to measure the supply-demand gap of ride-hailing services, which is
The accessibility measurement methods for shared mobility systems similar to the need gap analysis in Currie (2010) for traditional transit sys-
(e.g., bike-sharing, car-sharing) can be divided into two categories. The tems, but with the gap measured by the waiting time. Other accessibility
first category treats registered system users as those having accessibility measurement approaches for traditional transportation systems can also
to bike-sharing services. Gavin et al. (2016) used an online survey to collect be potentially used for ride-hailing services as researchers did for shared-
demographic data of bike sharing users and analyzed equity based on user mobility systems. For example, to estimate the accessibility for ride hailing
profiles. Other studies used the number of times a bike was used per cus- services, an intuitive method would be to identify the distribution of trans-
tomer (which can be computed from the system's operational data) as a sur- portation network company (TNC) vehicles and passenger demand across
rogate accessibility measure, to evaluate equity among different the spatial space. However, the ride-hailing companies rarely share their
demographic groups (Ogilvie and Goodman, 2012; Wang and Akar, operational data with researchers, making it nearly impossible to assess eq-
2019). The second category defines the coverage/service area of a shared uity performance using this method and there is no research adopting this
mobility system using buffers around sharing stations, and defines those method currently.
who reside within the coverage/service area as having accessibility. Although autonomous vehicles (AVs) are still under development, many
These studies have been primarily focused on station-based systems. Due studies have evaluated their possible impacts on human travel behaviors
to the similarity between station-based shared mobility systems and tradi- (Zhang et al., 2015; Zmud and Sener, 2017) while few studies have quanti-
tional public transit systems, the coverage-based approach for the latter tatively studied AVs' impacts on equity of accessibility among populations.
has been applied for analyzing station-based shared-mobility systems The only relevant literature we found is Cohn et al. (2019). They adopted a
(Ursaki and Aultman-Hall, 2016; Gavin et al., 2016). However, this ap- travel demand modeling approach to estimate job accessibility, trip dura-
proach ignores the true supply of shared bikes due to factors such as the sta- tion, trip distance for disadvantaged communities and others, based on
tion size, number of bikes at a station and rebalancing strategies. Therefore, eight different AV scenarios. However, they assumed that each household
other studies have adopted the mean density or availability of shared bikes will own an AV, which weakens the generality of the results. For example,
or cars within the census area as the measure of accessibility (Shellooe, if AVs are entirely shared like ride hailing services, waiting time and AV
2013; Couch and Smalley, 2019; Mooney et al., 2019). This method follows density within a unit could also be a proxy for accessibility. However, if
a similar logic to the way Welch and Mishra (2013) refined accessibility AVs can serve as both private cars and shared mobility, more sophisticated
measurement for traditional public transportation systems. Yet, it fails to modeling approaches are needed to measure accessibility for AVs, which
consider temporal variations of bike/car supply at sharing stations has not been addressed in the existing literature. Further, though it is

Table 1
Studies of equity of accessibility for emerging transportation technologies.
Population measurement Cost/benefit measurement Inequity measurement Citation

Census block level demographic data Accessibility (coverage-based approach) T-student test (Vertical equity) (Ursaki and
Aultman-Hall, 2016)
Census block level demographic data Accessibility (coverage-based approach) Chi-square test (Vertical equity) (Gavin et al., 2016)
Bike sharing data, Census block level data, Bike drop offs by gender Negative binomial regression (Vertical equity) (Wang and Akar, 2019)
land use, time of day, weather, etc.
Census block level demographic data Spatial access for shared dockless bikes (Bike Spatial mismatch analysis, descriptive statistics (Mooney et al., 2019)
availability, rebalancing) (median), (Vertical equity)
Census tract level demographic data Counts of available bikes Poisson regression (Vertical equity) (Couch and Smalley,
2019)
Census block level demographic data Accessibility (coverage-based approach) Spatial mismatch (Vertical equity) (Bhuyan et al., 2019)
(Bike equity index)
Census tract level demographic data Car sharing density Bivariate regression model (Vertical equity) (Shellooe, 2013)
Disaggregated individual data from travel Tour based accessibility Gini index (Horizontal equity); (Chen et al., 2019)
demand model Subgroup inequality index (Vertical equity)
Disaggregated data from travel Job accessibility, trip duration, trip distance Descriptive statistics (average) (Vertical equity) (Cohn et al., 2019)
demand model

10
Y. Guo et al. Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

reasonable to use trip duration and trip distance to quantify accessibility, neighborhood environmental factors that could possibly explain the dispar-
different socio-demographic populations may perceive the importance of ities found (Kravetz and Noland, 2012). However, differences in pedes-
travel time and distance differently depending on trip purpose. With disag- trian/cyclist behaviors among socio-demographic population groups are
gregated data from a travel demand model, such factors could also be con- less understood. Future studies could also assess interaction between
sidered in an accessibility measure. human behaviors and the emerging transportation technologies (e.g., e-
scooter sharing).
8.1.2. Traffic emissions
Compared with accessibility, studies on the equity impacts of emerging 8.2. Equity outcomes and policy implications
transportation systems in terms of environmental outcomes are rare. Gener-
ally, these impacts could be considered from two perspectives: Similar to traditional transportation modes, shared mobility programs
• Net impact - Emerging transportation technologies may influence traffic are also considered unfair overall in terms of accessibility. Ursaki and
emissions and their distribution across space in multiple conflicting Aultman-Hall (2016) compared social and economic characteristics of cen-
ways. What are the net impacts and how are they distributed spatially sus block groups that were within a service area to those outside a service
and among different socio-demographic groups in society? area for eight U.S. bike-sharing programs. Results indicate that the percent
• User exposure - How does exposure to traffic emissions vary among differ- of white, college-educated, higher-income people within the service area
ent population groups while using emerging transportation technologies? was higher than that of the remaining groups in Chicago, Denver, Seattle,
and New York city. Similarly, an equity gap also exists in Baltimore
The first perspective is rather difficult to address as it requires the esti- (Bhuyan et al., 2019) and Seattle (Mooney et al., 2019), and for a car shar-
mation of the net effects on traffic emissions and their spatial distribution. ing program in New York (Shellooe, 2013). In terms of user profiles, Gavin
Take ride hailing as an example, a small percentage of ride hailing users re- et al. (2016) surveyed three anonymous cities and reported that users were
duce vehicle ownership after using ride hailing (Clewlow and Mishra, more likely to be residents who are male, young, white, affluent, and edu-
2017). However, ride hailing vehicle drivers also bring many vehicles cated. Several factors have been found to lead to unequal accessibility.
onto roads, causing substantial emissions. How much the net emissions Firstly, as shared mobility companies are usually in the private sector,
change due to both the additional vehicles and reduced car ownership is their profit-driven strategy usually overlooks equity and pushes the services
hard to estimate, partially because operational data for ride hailing compa- out of reach to many disadvantaged areas. Sometimes, even if the program
nies are not publicly available. A travel demand modeling approach may be places stations in those areas, they are not fully integrated with the main
able to capture these changes, but no studies have done it yet. With bike cluster of stations, which limits the usefulness for adjacent users (Schmitt,
sharing operation data, Zhang and Mi (2018) attempted to estimate the en- 2012). Also, limited internet, smartphone and credit/debit card access re-
vironmental benefits of a bike sharing program from a spatial-temporal per- strict the use of bike sharing programs for disadvantaged populations
spective. They assumed that bike sharing trip distances greater than 1 km (Shaheen et al., 2017). In terms of environmental equity, bike sharing riders
are energy saving (due to automobile substitution) and estimated saved pet- in disadvantaged communities have also been found to bear higher traffic
rol, CO2 and NOx. CO2 and NOx concentrations were estimated using exposures in Chicago, largely because those areas are close to major and
MOVES as the emission model and AERMOD as the dispersion model; busy roadways (Qian and Wu, 2019).
this is consistent with the approach used for traffic pollution measurement Current research shows that white, affluent and highly educated popu-
for traditional transportation systems. However, this study overlooks the lations usually enjoy most of the benefits of shared mobility services.
possibility that bike sharing trips may replace public transit (bus, metro), Hence, these services do not bring as many benefits to the people who
so the benefits could be overestimated. To extend this study, the benefits might need it the most. To promote a more equitable shared mobility sys-
could be mapped on census areas so that equity could be evaluated using tem, several actions could be taken. Shaheen et al. (2017) suggest that
spatial mismatch or inequality indicator approaches. the government could play a role by facilitating public and private partner-
The second perspective might be more relevant to bike-sharing and e- ships and knowledge sharing among government agencies, attracting pri-
scooter sharing as riders directly expose themselves to the traffic emissions vate sector partners by providing subsidies for meeting community equity
while using these modes. For vehicle drivers or passengers, in-vehicle expo- goals, performing pilot studies and evaluations prior to setting standards
sures can be significantly different from ambient air concentrations and regulations, ensuring policies are goal orientated rather than mitiga-
(Hatzopoulou and Miller, 2010). So far, only one paper has evaluated the tion orientated, standardizing data sharing requirements for all service pro-
environment equity of emerging transportation modes from a user expo- viders, and developing more rigorous equity metrics.
sure perspective. Qian and Wu (2019) assessed the equity of PM2.5 exposure Although emerging transportation technologies and services are ad-
for a bike sharing program in Chicago. The analysis process exactly follows vancing rapidly, it is essential for governments to plan ahead and take a pro-
the three-step framework discussed above, except for the population mea- active approach by implementing policy frameworks that can prevent
surement step, due to lack of user socio-demographic data. Instead, the au- inequity, rather than waiting for the diffusion of new technology to imple-
thors evaluated equity for trip types based on whether a trip began or ended ment the policy.
in disadvantaged areas. The PM2.5 concentration was obtained through air
quality modeling (MOVES and AERMOD) and exposure was estimated as 8.3. Research gaps
the product of time spent in each spatial unit and concentration for that
unit. Finally, equity for trip types was compared using descriptive statistics This section identifies several gaps in the current research related to as-
(boxplots). As previously mentioned, a similar study could provide addi- sessment of equity for emerging transportation systems.
tional insights if socio-demographic information of trip users were known.
• Comparison of the equity performance between emerging systems and
8.1.3. Safety baseline. Current studies on equity performance of emerging transporta-
To the best knowledge of the authors, there are no studies evaluating tion systems (e.g., bike sharing programs) solely focus on the emerging
the equity of safety outcomes for emerging transportation technologies. transportation systems, while the baseline disparity is usually ignored.
The lack of accident data relevant to emerging transportation technologies To truly reflect the impacts of a policy or program on equity, it is neces-
could be the largest hurdle. If such data were available, the three-step sary to compare the equity performance before and after the implementa-
framework could be applied to assess equity in a similar manner to tradi- tion of an emerging transportation technology.
tional transportation systems. Currently, many of the safety equity studies • Integrated assessment with respect to accessibility, environment, and
have explored the disparity of accidents between different socio- public health. Equity research on emerging transportation technologies
demographic groups. Some studies have also attempted to understand the has mainly focused on accessibility, while the equity of environmental

11
Y. Guo et al. Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

impact and of safety outcomes have not been frequently discussed, and no and autonomous vehicles have not yet received enough attention in
integrated assessment considers all these aspects together. Furthermore, the literature. There is also limited research on equity performant re-
there are several other environmental and public health related outcomes lated to environmental impact and safety outcomes.
of transportation programs that can have equity implications, such as im-
pacts on physical activity, noise, and climate and their subsequent health As research gaps indicate, equity evaluation of emerging transportation
effects on vulnerable populations. For practical implications, making technologies and services remains a challenging task. Not only do we need
decisions based on one aspect alone can be erroneous. For example, to consider the equity from an integrated perspective which needs expertise
ride-hailing can significantly impact several outcomes. It improves acces- from various disciplines (e.g., accessibility, environment and safety), but
sibility for people who do not have vehicles, but it also brings more vehi- also better understand people's travel behaviors (e.g., through survey
cles on the road; consequently more traffic emissions and traffic accidents data, operation data collection and analysis) and come up with meaningful
may occur. To assess whether the outcomes of this mode are equitable, cost/benefit measures under new transportation systems (e.g., through pol-
multiple aspects should be considered. However, no studies have been icy makers). Though individual level of travel data potentially offers a more
carried out to evaluate equity from an integrated perspective for both tra- accurate equity assessment, it can have high computational cost if the size
ditional and emerging transportation system, nor does research discuss of study area is very large, this may require collaboration with expert
the trade-offs and interactions among equity measures for the different from computer science field. Overall, interdisciplinary efforts are necessary
types of outcomes. to address these issues.
• Assessment under a multimodal transportation system context. Different
travel modes in a transportation system are interrelated with and affected 9. Conclusion
by each other, especially for many emerging transportation services. For
example, apart from providing stand-alone services for short-distance This paper summarizes a general framework that evaluates transporta-
travel, bike-sharing is also an effective solution to the first/last-mile prob- tion equity from an interdisciplinary perspective that considers accessibil-
lem for transit. In this context, analyzing a bike-sharing system without ity, environmental impact, and safety. The equity assessment framework
considering public transit is not ideal. Also, a multimodal perspective is is divided into three components—population measurement, cost/benefit
important, as it can evaluate the overall equity performance of a transpor- measurement, and equity measurement—which are commonly applied in
tation system instead of just one mode, which could be used to compare the literature. Furthermore, we examined the equity evaluation papers for
the overall equity performance before and after the implementation of emerging transportation technologies and found that the analysis method-
an emerging transportation technology. Hence, developing a methodol- ologies still follow the three-step framework, though the specific cost/ben-
ogy that can be applied in the context of a multimodal transportation sys- efit measurement approaches may vary. Further, we found the majority of
tem should be an interesting future direction. these studies have focused on equity impacts in terms of accessibility for
• Disaggregate measures with high-resolution inputs. Due to the lack of shared mobility. Rarely have researchers evaluted the equity performance
high-resolution data, equity assessment of traditional transportation sys- of other emering transportation systems, nor have they adequately assessed
tems has usually adopted aggregate information on both population and the equity of environmental and safety impacts.
cost/benefit measures, e.g., zone-level information. This practice can pro- Overall, this paper offers a comprehensive review of equity assessment
vide a macroscopic assessment of the equity performance of a transporta- methods in transportation systems from an integrated perspective consider-
tion system, but results in some errors. Many existing emerging ing accessibility, traffic emissions, and safety outcomes. The three-step tax-
transportation technologies are based on smartphone applications or onomy described here can serve as a convenient resource for transportation
website toolkits, e.g., Uber, Zipcar, Ofo, and some future technologies planning agencies to find the measurement approaches they desire more ef-
are believed to be the same (http://www.next-future-mobility.com/). ficiently. Research gaps identified also offer resources on future research di-
This operational mode opens an opportunity to collect individual-level rections for equity studies of emerging transportation systems.
population characteristics. Further, vehicles offering emerging transpor-
tation services are usually installed with GPS, which may offer Acknowledgement
individual-level trajectory information. With this, the proposition of dis-
aggregate assessment approaches might be an interesting topic. How- The authors gratefully acknowledge support provided by the Center for
ever, if the private sector does not share the operation data with Transportation, Environment, and Community Health (CTECH), University
researchers and planners, it will be hard to get a reliable equity esti- Transportation Centers sponsored by the US Department of Transportation
mation. The governmental and industry sectors need to collaborate through Grant No. 69A3551747119. The contents of this manuscript reflect
to fill this gap. Besides operational data, the output from travel de- the views of the authors, who are responsible for the facts and the accuracy
mand model also offers individual level details that help disaggregate of the information presented herein. The contents do not necessarily reflect
analysis. the official views or policies of US Department of Transportation.
• Consideration of the operational characteristics of emerging transpor-
tation services. Many operational details concerning the service qual- References
ity of an emerging transportation system are not the same as those in
traditional transportation services. For instance, in bike-sharing sys- Ahmed, A., et al., 2017. Errors in accident data, its types, causes and methods of rectification-
tems, rebalancing (redistributing bikes across the network using a analysis of the literature. Accident Analysis & Prevention.
Aldred, R., 2018. Inequalities in self-report road injury risk in Britain: a new analysis of Na-
fleet of vehicles) is an important operation that can affect service qual- tional Travel Survey data, focusing on pedestrian injuries (Journal of Transport &
ity. However, due to its uniqueness to such emerging services, effects Health).
on equity performance have not been studied for traditional transpor- Atkinson, A.B., 1970. On the measurement of inequality. J. Econ. Theory 2 (3), 244–263.
Bhuyan, I.A., et al., 2019. GIS-based equity gap analysis: case study of Baltimore bike share
tation systems. Thus, although measurements that can reflect the
program. Urban Sci. 3 (2), 42.
service quality of traditional public transit systems are present, meth- Boarnet, M.G., et al., 2017. First/last mile transit access as an equity planning issue. Transp.
odologies that can capture more operational details are needed for Res. A Policy Pract. 103, 296–310.
Bösch, P. M., et al. (2018). "Cost-based analysis of autonomous mobility services." Transport
emerging transportation technologies.
policy 64: 76–91%@ 0967-0070X.
• Assessment of other emerging technologies. As pointed out in Boyce, J. K., et al. (2016). "Measuring environmental inequality." Ecological Economics 124:
Section 8.1, only the equity performance of some bike sharing and 114–123%@ 0921–8009.
car sharing systems have been studied and most of them have focused Brauer, M., et al., 2003. Estimating long-term average particulate air pollution concentrations:
application of traffic indicators and geographic information systems. Epidemiology
on accessibility outcomes. The equity performance of other emerging 228–239.
transportation technologies such as electric vehicles, ride-sharing, Brown, A.E., 2018. Ridehail Revolution: Ridehail Travel and Equity in Los Angeles (UCLA).

12
Y. Guo et al. Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

Buck, D., 2013. Encouraging equitable access to public bikesharing systems. ITE Journal 83 Gurram, S., et al., 2019. Agent-based modeling to estimate exposures to urban air pollution
(3), 24–27. from transportation: exposure disparities and impacts of high-resolution data. Comput.
Buka, I., et al. (2006). "The effects of air pollution on the health of children." Paediatrics & Environ. Urban. Syst. 75, 22–34.
child health 11(8): 513–516%@ 1205–7088. Guzman, L.A., et al., 2017. Assessing equity in transport accessibility to work and study: the
Buzzelli, M., Jerrett, M., 2007. Geographies of susceptibility and exposure in the city: environ- Bogotá region. J. Transp. Geogr. 58, 236–246.
mental inequity of traffic-related air pollution in Toronto. Can. J. Reg. Sci. 30 (2). Habermann, M., et al., 2015. Land use regression as method to model air pollution. Previous
Carrier, M., et al. (2017). "School locations and road transportation nuisances in Montreal: an results for Gothenburg/Sweden. Procedia Engineering 115, 21–28.
environmental equity diagnosis." Transp. Policy 0967-070X. Handy, S. (2008). "Regional transportation planning in the US: an examination of changes in
Cesaroni, G., et al., 2012. Nitrogen dioxide levels estimated from land use regression models technical aspects of the planning process in response to changing goals." Transport policy
several years apart and association with mortality in a large cohort study. Environ. Health 15(2): 113–126%@ 0967-0070X.
11 (1), 48. Harner, J., et al., 2002. Urban environmental justice indices. Prof. Geogr. 54 (3), 318–331.
Chen, Z., et al. (2019). "Exploring the equity performance of bike-sharing systems with disag- Harper, S., et al., 2015. Trends in socioeconomic inequalities in motor vehicle accident deaths
gregated data: a story of southern Tampa." Transportation research part A: policy and in the United States, 1995–2010. Am. J. Epidemiol. 182 (7), 606–614.
practice 130: 529–545%@ 0965–8564. Hatzopoulou, M., Miller, E.J., 2010. Linking an activity-based travel demand model with traf-
Clewlow, R. R. and G. S. Mishra (2017). "Disruptive transportation: the adoption, utilization, fic emission and dispersion models: transport’s contribution to air pollution in Toronto.
and impacts of ride-hailing in the United States.". Transp. Res. Part D: Transp. Environ. 15 (6), 315–325.
Cohn, J., et al., 2019. Examining the equity impacts of autonomous vehicles: a travel demand Havard, S., et al., 2009. Traffic-related air pollution and socioeconomic status: a spatial auto-
model approach. Transportation research record: 0361198119836971%@ correlation study to assess environmental equity on a small-area scale. Epidemiology 20
0361198119830361–0361198119831981. (2), 223–230.
Couch, S. and H. K. Smalley (2019). "Encouraging equitable bikeshare: implications of docked Health Effects Institute. Panel on the Health Effects of Traffic-Related Air, P, 2010. Traffic-re-
and dockless models for spatial equity." arXiv preprint arXiv:1906.00129. lated Air Pollution: A Critical Review of the Literature on Emissions, Exposure, and
Currie, G., 2004. Gap analysis of public transport needs: measuring spatial distribution of pub- Health Effects. Institute, Health Effects.
lic transport needs and identifying gaps in the quality of public transport provision. Trans- (9/7/2017). "Title VI of the Civil Rights Act of 1964.". Retrieved from. https://www.transit.
portation Research Record: Journal of the Transportation Research Board 1895, dot.gov/title6.
137–146. Hughes, R. and D. MacKenzie (2016). "Transportation network company wait times in Greater
Currie, G., 2010. Quantifying spatial gaps in public transport supply based on social needs. Seattle, and relationship to socioeconomic indicators." Journal of Transport Geography
J. Transp. Geogr. 18 (1), 31–41. 56: 36–44%@ 0966–6923.
Dark, S.J., Bram, D., 2007. The modifiable areal unit problem (MAUP) in physical geography. Jerrett, M., et al., 2005. A review and evaluation of intraurban air pollution exposure models.
Prog. Phys. Geogr. 31 (5), 471–479. Journal of exposure science and environmental epidemiology 15 (2), 185.
de Hoogh, K., et al., 2014. Comparing land use regression and dispersion modelling to assess Kaplan, S., et al., 2014. Using connectivity for measuring equity in transit provision. J. Transp.
residential exposure to ambient air pollution for epidemiological studies. Environ. Int. 73, Geogr. 37, 82–92.
382–392. Karner, A., et al., 2016. Understanding regional disparities in public transit performance using
Delbosc, A., Currie, G., 2011. Using Lorenz curves to assess public transport equity. J. Transp. realtime transit data. NTC2015-MU.
Geogr. 19 (6), 1252–1259. Kawabata, M., 2009. Spatiotemporal dimensions of modal accessibility disparity in Boston
Delmelle, E.C., Casas, I., 2012. Evaluating the spatial equity of bus rapid transit-based acces- and San Francisco. Environ Plan A 41 (1), 183–198.
sibility patterns in a developing country: the case of Cali, Colombia. Transp. Policy 20, Kraemer, J.D., Benton, C.S., 2015. Disparities in road crash mortality among pedestrians using
36–46. wheelchairs in the USA: results of a capture–recapture analysis. BMJ Open 5 (11), e008396.
Department of Trannsportation of United States (1970) Nondiscrimination in Federally- Kravetz, D., Noland, R., 2012. Spatial analysis of income disparities in pedestrian safety in
Assisted Programs of The Department of Transportation—Effectuation of Title VI of the northern New Jersey: is there an environmental justice issue? Transportation Research
Civil Rights Act of 1964. Record: Journal of the Transportation Research Board(2320) 10–17.
Di Ciommo, F., Shiftan, Y., 2017. Transport Equity Analysis. Taylor & Francis. Levy, J.I., et al., 2006. Incorporating concepts of inequality and inequity into health benefits
Dirgawati, M., et al. (2015). "Development of land use regression models for predicting expo- analysis. Int. J. Equity Health 5 (1), 2.
sure to NO2 and NOx in metropolitan Perth, Western Australia." Environmental Model- Levy, J.I., et al., 2009. Evaluating efficiency-equality tradeoffs for mobile source control strat-
ling & Software 74: 258–267. egies in an urban area. Risk Anal. 29 (1), 34–47.
Eeftens, M., et al., 2011. Stability of measured and modelled spatial contrasts in NO2 over Li, X., et al., 2016. Design framework of large-scale one-way electric vehicle sharing systems: a
time. Occupational and environmental medicine: oem. 061135, 2010. continuum approximation model. Transp. Res. B Methodol. 88, 21–45.
El-Geneidy, A., et al., 2016. The cost of equity: assessing transit accessibility and social dispar- Litman, T. (2002). "Evaluating transportation equity." World Transport Policy & Practice 8(2):
ity using total travel cost. Transp. Res. A Policy Pract. 91, 302–316. 50–65.
Ewing, R. and E. Dumbaugh (2009). "The built environment and traffic safety: a review Litman, T. And M. Brenman (2012). “A new social equity agenda for sustainable transporta-
of empirical evidence." Journal of Planning Literature 23(4): 347–367%@ 0885– tion”, Transportation Research Board 91st annual meeting, Washington, D.C., January,
4122. paper 12-3916. Available from the Victoria Transport Policy Institute, Victoria, Canadá,
Federal Highway Administration of the United States [FHWA], 2012. Order 6640.23 at https://www.vtpi.org/equityagenda.pdf
(a) FHWA Actions to Address Environmental Justice in Minority Populations and Low- Macinko, J.A., Starfield, B., 2002. Annotated bibliography on equity in health, 1980–2001.
income Populations. Retrieved 12/19, 2019, from. https://www.fhwa.dot.gov/ Int. J. Equity Health 1 (1), 1.
legsregs/directives/orders/664023a.cfm. Martens, K., et al. (2012). "A justice-theoretic approach to the distribution of transportation
FHWA and Federal Transit Administration [FTA], 1999. Implementing Title VI Requirements benefits: implications for transportation planning practice in the United States." Transpor-
in Metropolitan and Statewide Planning. Retrieved 12/19, 2019, from. https://www. tation research part A: policy and practice 46(4): 684–695%@ 0965–8564.
fhwa.dot.gov/environment/environmental_justice/legislation/ej-10-7.cfm. Martin, E., et al., 2010. Impact of carsharing on household vehicle holdings: results from
FTA, 2012a. Circular 4702.1B Title VI Requirements and Guidelines for Federal Transit Ad- North American shared-use vehicle survey. Transp. Res. Rec. 2143 (1), 150–158.
ministration Recipients. from https://www.transit.dot.gov/regulations-and-guidance/ Mishra, S., et al., 2012. Performance indicators for public transit connectivity in multi-modal
fta-circulars/title-vi-requirements-and-guidelines-federal-transit. transportation networks. Transp. Res. A Policy Pract. 46 (7), 1066–1085.
FTA, 2012b. Environmental Justice Policy Guidance for Federal Transit Administration Recip- Mooney, S.J., et al., 2019. Freedom from the station: spatial equity in access to dockless bike
ients. Retrieved 12/19, 2019, from https://www.transit.dot.gov/regulations-and-guid- share. J. Transp. Geogr. 74, 91–96.
ance/fta-circulars/environmental-justice-policy-guidance-federal-transit. Morency, P., et al., 2012. Neighborhood social inequalities in road traffic injuries: the influ-
Gaffron, P. (2012). "Urban transport, environmental justice and human daily activity pat- ence of traffic volume and road design. Am. J. Public Health 102 (6), 1112–1119.
terns." Transport policy 20: 114–127%@ 0967-0070X. Noland, R.B., et al., 2013. Do lower income areas have more pedestrian casualties? Accid.
García-Valiñas, M.A., et al., 2005. More Income Equality or Not? An Empirical Analysis of In- Anal. Prev. 59, 337–345.
dividuals’ Preferences. Center for Research in Economics, Management and the Arts Ogilvie, F. and A. Goodman (2012). "Inequalities in usage of a public bicycle sharing scheme:
(CREMA) socio-demographic predictors of uptake and usage of the London (UK) cycle hire scheme."
Gavin, K., et al., 2016. A brief study exploring social equity within bicycle share programs. Preventive medicine 55(1): 40–45%@ 0091–7435.
Transportation Letters 8 (3), 177–180. Pal, A., Zhang, Y., 2017. Free-floating bike sharing: solving real-life large-scale static rebalancing
Geronimus, A.T., Bound, J., 1998. Use of census-based aggregate variables to proxy for socio- problems. Transportation Research Part C: Emerging Technologies 80, 92–116.
economic group: evidence from national samples. Am. J. Epidemiol. 148 (5), 475–486. Qian, X. and Y. Wu (2019). "Assessment for health equity of PM2. 5 exposure in bikeshare sys-
Golub, A., Martens, K., 2014. Using principles of justice to assess the modal equity of regional tems: the case of Divvy in Chicago." Journal of Transport & Health 14: 100596%@
transportation plans. J. Transp. Geogr. 41, 10–20. 102214–101405.
Grengs, J., 2010. Job accessibility and the modal mismatch in Detroit. J. Transp. Geogr. 18 Ricciardi, A. M., et al. (2015). "Exploring public transport equity between separate disadvantaged
(1), 42–54. cohorts: a case study in Perth, Australia." Journal of Transport Geography 43: 111–122.
Gurram, S., 2017. Understanding the linkages between urban transportation design and pop- Rijnders, E., et al., 2001. Personal and outdoor nitrogen dioxide concentrations in relation to
ulation exposure to traffic-related air pollution. application of an integrated transporta- degree of urbanization and traffic density. Environ. Health Perspect. 109 (Suppl. 3), 411.
tion and air pollution modeling framework to Tampa, FL. Rowangould, G.M., 2013. A census of the US near-roadway population: public health and en-
Gurram, S., et al., 2015. Impacts of travel activity and urbanicity on exposures to ambient ox- vironmental justice considerations. Transp. Res. Part D: Transp. Environ. 25, 59–67.
ides of nitrogen and on exposure disparities. Air Quality, Atmosphere & Health 8 (1), Ryan, P.H., LeMasters, G.K., 2007. A review of land-use regression models for characterizing
97–114. intraurban air pollution exposure. Inhal. Toxicol. 19 (sup1), 127–133.
Gurram, S., Stuart, A.L., Pinjari, A.R., 2018. Agent-based modeling to estimate exposures to Schmitt, A., 2012. Why isn't bike-share reaching more low-income people? Retrieved 11/16,
urban air pollution from transportation: exposure disparities and impacts of high-resolu- 2019, from https://usa.streetsblog.org/2012/10/03/why-isnt-bike-share-reaching-
tion data. Comput. Environ. Urban. Syst. (In review). more-low-income-people/

13
Y. Guo et al. Transportation Research Interdisciplinary Perspectives xxx (xxxx) xxx

Shaheen, S., et al. (2017). Travel Behavior: Shared Mobility and Transportation Equity. Wang, K. and G. Akar (2019). "Gender gap generators for bike share ridership: evidence from Citi
Shellooe, S.D., 2013. Wheels When Who Wants Them: Assessing Social Equity and Access Im- Bike system in New York City." Journal of Transport Geography 76: 1–9%@ 0966–6923.
plications of Carsharing in NYC. Wang, R., et al., 2013. Temporal stability of land use regression models for traffic-related air
Sider, T., et al., 2015. Smog and socioeconomics: an evaluation of equity in traffic-related air pollution. Atmos. Environ. 64, 312–319.
pollution generation and exposure. Environment and Planning B: Planning and Design 42 Welch, T.F., Mishra, S., 2013. A measure of equity for public transit connectivity. J. Transp.
(5), 870–887. Geogr. 33, 29–41.
Simoni, M., et al., 2015. Adverse effects of outdoor pollution in the elderly. Journal of thoracic Weymark, J. A. (1981). "Generalized Gini inequality indices." Mathematical Social Sciences 1
disease 7 (1), 34. (4): 409–430%@ 0165–4896.
Smith, A., 2016. Shared. The New Digital Economy, Pew Research Center, Collaborative and Woodcock, J., et al., 2014. Health effects of the London bicycle sharing system: health impact
On Demand. modelling study. Bmj g425, 348.
Soobader, M., et al., 2001. Using aggregate geographic data to proxy individual socioeco- World Bank, 2005. In Equality Measures (Introduction to Poverty Analysis).
nomic status: does size matter? Am. J. Public Health 91 (4), 632. Yi, C. (2006). "Impact of public transit on employment status: disaggregate analysis of Hous-
Stark, J., Diakopoulos, N., 2016. Uber Seems to Offer Better Service in Areas With More White ton, Texas." Transportation Research Record 1986(1): 137–144%@ 0361–1981.
People (That Raises Some Tough Questions). Yu, C.-Y., 2014. Environmental supports for walking/biking and traffic safety: income and
Steinbach, R., et al., 2010. Race’or place? Explaining ethnic variations in childhood pedestrian ethnicity disparities. Prev. Med. 67, 12–16.
injury rates in London. Health & place 16 (1), 34–42. Yu, H., Stuart, A.L., 2013. Spatiotemporal distributions of ambient oxides of nitrogen, with
Steinbach, R., et al. (2016). "Is ethnic density associated with risk of child pedestrian injury? A implications for exposure inequality and urban design. J. Air Waste Manage. Assoc. 63
comparison of inter-census changes in ethnic populations and injury rates." Ethnicity & (8), 943–955.
health 21(1): 1–19. U.S. EPA (2019). Fast Facts on Transportation Greenhouse Gas Yu, H., Stuart, A.L., 2016. Exposure and inequality for select urban air pollutants in the Tampa
Emissions. Bay area. Sci. Total Environ. 551, 474–483.
Stuart, A.L., Zeager, M., 2011. An inequality study of ambient nitrogen dioxide and traffic Yu, H., Stuart, A.L., 2017. Impacts of compact growth and electric vehicles on future air qual-
levels near elementary schools in the Tampa area. J. Environ. Manag. 92 (8), ity and urban exposures may be mixed. Sci. Total Environ. 576, 148–158.
1923–1930. Zhang, Y., Lin, G., 2013. Disparity surveillance of nonfatal motor vehicle crash injuries. Traffic
Stuart, A.L., et al., 2009. The social distribution of neighborhood-scale air pollution and mon- injury prevention 14 (7), 697–702.
itoring protection. J. Air Waste Manage. Assoc. 59 (5), 591–602. Zhang, Y. and Z. Mi (2018). "Environmental benefits of bike sharing: a big data-based analy-
Tian, N., et al., 2013. Evaluating socioeconomic and racial differences in traffic-related met- sis." Applied Energy 220: 296–301%@ 0306–2619.
rics in the United States using a GIS approach. Journal of exposure science and environ- Zhang, W., et al. (2015). "Exploring the impact of shared autonomous vehicles on urban
mental epidemiology 23 (2), 215. parking demand: an agent-based simulation approach." Sustainable Cities and Society
U.S. EPA (2019). Fast Facts on Transportation Greenhouse Gas Emissions. 19: 34–45%@ 2210–6707.
Ursaki, J., Aultman-Hall, L., 2016. Quantifying the equity of bikeshare access in US cities. Zmud, J. P. and I. N. Sener (2017). "Towards an understanding of the travel behavior impact of
95th Annual Meeting of the Transportation Research Board. DC, Washington. autonomous vehicles." Transportation research procedia 25: 2500–2519%@ 2352–1465.

14

Das könnte Ihnen auch gefallen